rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
postgresql/src/backend/optimizer/plan/initsplan.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

3225 lines
107 KiB
C
Raw Normal View History

Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*-------------------------------------------------------------------------
*
Change my-function-name-- to my_function_name, and optimizer renames.
1999-02-14 00:22:53 +01:00
* initsplan.c
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
* Target list, qualification, joininfo initialization routines
*
Update copyright for 2023 Backpatch-through: 11
2023-01-02 21:00:37 +01:00
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
Add: * Portions Copyright (c) 1996-2000, PostgreSQL, Inc to all files copyright Regents of Berkeley. Man, that's a lot of files.
2000-01-26 06:58:53 +01:00
* Portions Copyright (c) 1994, Regents of the University of California
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*
* IDENTIFICATION
Remove cvs keywords from all files.
2010-09-20 22:08:53 +02:00
* src/backend/optimizer/plan/initsplan.c
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*-------------------------------------------------------------------------
*/
Rewrite of planner statistics-gathering code. ANALYZE is now available as a separate statement (though it can still be invoked as part of VACUUM, too). pg_statistic redesigned to be more flexible about what statistics are stored. ANALYZE now collects a list of several of the most common values, not just one, plus a histogram (not just the min and max values). Random sampling is used to make the process reasonably fast even on very large tables. The number of values and histogram bins collected is now user-settable via an ALTER TABLE command. There is more still to do; the new stats are not being used everywhere they could be in the planner. But the remaining changes for this project should be localized, and the behavior is already better than before. A not-very-related change is that sorting now makes use of btree comparison routines if it can find one, rather than invoking '<' twice.
2001-05-07 02:43:27 +02:00
#include "postgres.h"
Expand partitioned table RTEs level by level, without flattening. Flattening the partitioning hierarchy at this stage makes various desirable optimizations difficult. The original use case for this patch was partition-wise join, which wants to match up the partitions in one partitioning hierarchy with those in another such hierarchy. However, it now seems that it will also be useful in making partition pruning work using the PartitionDesc rather than constraint exclusion, because with a flattened expansion, we have no easy way to figure out which PartitionDescs apply to which leaf tables in a multi-level partition hierarchy. As it turns out, we end up creating both rte->inh and !rte->inh RTEs for each intermediate partitioned table, just as we previously did for the root table. This seems unnecessary since the partitioned tables have no storage and are not scanned. We might want to go back and rejigger things so that no partitioned tables (including the parent) need !rte->inh RTEs, but that seems to require some adjustments not related to the core purpose of this patch. Ashutosh Bapat, reviewed by me and by Amit Langote. Some final adjustments by me. Discussion: http://postgr.es/m/CAFjFpRd=1venqLL7oGU=C1dEkuvk2DJgvF+7uKbnPHaum1mvHQ@mail.gmail.com
2017-09-14 21:41:08 +02:00
#include "catalog/pg_class.h"
Make the order of the header file includes consistent in backend modules. Similar to commits 7e735035f2 and dddf4cdc33, this commit makes the order of header file inclusion consistent for backend modules. In the passing, removed a couple of duplicate inclusions. Author: Vignesh C Reviewed-by: Kuntal Ghosh and Amit Kapila Discussion: https://postgr.es/m/CALDaNm2Sznv8RR6Ex-iJO6xAdsxgWhCoETkaYX=+9DW3q0QCfA@mail.gmail.com
2019-11-12 04:00:16 +01:00
#include "catalog/pg_type.h"
Make some small planner API cleanups. Move a few very simple node-creation and node-type-testing functions from the planner's clauses.c to nodes/makefuncs and nodes/nodeFuncs. There's nothing planner-specific about them, as evidenced by the number of other places that were using them. While at it, rename and_clause() etc to is_andclause() etc, to clarify that they are node-type-testing functions not node-creation functions. And use "static inline" implementations for the shortest ones. Also, modify flatten_join_alias_vars() and some subsidiary functions to take a Query not a PlannerInfo to define the join structure that Vars should be translated according to. They were only using the "parse" field of the PlannerInfo anyway, so this just requires removing one level of indirection. The advantage is that now parse_agg.c can use flatten_join_alias_vars() without the horrid kluge of creating an incomplete PlannerInfo, which will allow that file to be decoupled from relation.h in a subsequent patch. Discussion: https://postgr.es/m/11460.1548706639@sss.pgh.pa.us
2019-01-29 21:26:44 +01:00
#include "nodes/makefuncs.h"
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
#include "nodes/nodeFuncs.h"
Final cleanup.
1999-07-16 07:00:38 +02:00
#include "optimizer/clauses.h"
Improve planner's cost estimation in the presence of semijoins. If we have a semijoin, say SELECT * FROM x WHERE x1 IN (SELECT y1 FROM y) and we're estimating the cost of a parameterized indexscan on x, the number of repetitions of the indexscan should not be taken as the size of y; it'll really only be the number of distinct values of y1, because the only valid plan with y on the outside of a nestloop would require y to be unique-ified before joining it to x. Most of the time this doesn't make that much difference, but sometimes it can lead to drastically underestimating the cost of the indexscan and hence choosing a bad plan, as pointed out by David Kubečka. Fixing this is a bit difficult because parameterized indexscans are costed out quite early in the planning process, before we have the information that would be needed to call estimate_num_groups() and thereby estimate the number of distinct values of the join column(s). However we can move the code that extracts a semijoin RHS's unique-ification columns, so that it's done in initsplan.c rather than on-the-fly in create_unique_path(). That shouldn't make any difference speed-wise and it's really a bit cleaner too. The other bit of information we need is the size of the semijoin RHS, which is easy if it's a single relation (we make those estimates before considering indexscan costs) but problematic if it's a join relation. The solution adopted here is just to use the product of the sizes of the join component rels. That will generally be an overestimate, but since estimate_num_groups() only uses this input as a clamp, an overestimate shouldn't hurt us too badly. In any case we don't allow this new logic to produce a value larger than we would have chosen before, so that at worst an overestimate leaves us no wiser than we were before.
2015-03-12 02:21:00 +01:00
#include "optimizer/cost.h"
Speed up planning when partitions can be pruned at plan time. Previously, the planner created RangeTblEntry and RelOptInfo structs for every partition of a partitioned table, even though many of them might later be deemed uninteresting thanks to partition pruning logic. This incurred significant overhead when there are many partitions. Arrange to postpone creation of these data structures until after we've processed the query enough to identify restriction quals for the partitioned table, and then apply partition pruning before not after creation of each partition's data structures. In this way we need not open the partition relations at all for partitions that the planner has no real interest in. For queries that can be proven at plan time to access only a small number of partitions, this patch improves the practical maximum number of partitions from under 100 to perhaps a few thousand. Amit Langote, reviewed at various times by Dilip Kumar, Jesper Pedersen, Yoshikazu Imai, and David Rowley Discussion: https://postgr.es/m/9d7c5112-cb99-6a47-d3be-cf1ee6862a1d@lab.ntt.co.jp
2019-03-30 23:58:55 +01:00
#include "optimizer/inherit.h"
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
#include "optimizer/joininfo.h"
Refactor planner's header files. Create a new header optimizer/optimizer.h, which exposes just the planner functions that can be used "at arm's length", without need to access Paths or the other planner-internal data structures defined in nodes/relation.h. This is intended to provide the whole planner API seen by most of the rest of the system; although FDWs still need to use additional stuff, and more thought is also needed about just what selfuncs.c should rely on. The main point of doing this now is to limit the amount of new #include baggage that will be needed by "planner support functions", which I expect to introduce later, and which will be in relevant datatype modules rather than anywhere near the planner. This commit just moves relevant declarations into optimizer.h from other header files (a couple of which go away because everything got moved), and adjusts #include lists to match. There's further cleanup that could be done if we want to decide that some stuff being exposed by optimizer.h doesn't belong in the planner at all, but I'll leave that for another day. Discussion: https://postgr.es/m/11460.1548706639@sss.pgh.pa.us
2019-01-29 21:48:51 +01:00
#include "optimizer/optimizer.h"
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
#include "optimizer/pathnode.h"
New cost model for planning, incorporating a penalty for random page accesses versus sequential accesses, a (very crude) estimate of the effects of caching on random page accesses, and cost to evaluate WHERE- clause expressions. Export critical parameters for this model as SET variables. Also, create SET variables for the planner's enable flags (enable_seqscan, enable_indexscan, etc) so that these can be controlled more conveniently than via PGOPTIONS. Planner now estimates both startup cost (cost before retrieving first tuple) and total cost of each path, so it can optimize queries with LIMIT on a reasonable basis by interpolating between these costs. Same facility is a win for EXISTS(...) subqueries and some other cases. Redesign pathkey representation to achieve a major speedup in planning (I saw as much as 5X on a 10-way join); also minor changes in planner to reduce memory consumption by recycling discarded Path nodes and not constructing unnecessary lists. Minor cleanups to display more-plausible costs in some cases in EXPLAIN output. Initdb forced by change in interface to index cost estimation functions.
2000-02-15 21:49:31 +01:00
#include "optimizer/paths.h"
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
#include "optimizer/placeholder.h"
Final cleanup.
1999-07-16 07:00:38 +02:00
#include "optimizer/planmain.h"
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
#include "optimizer/planner.h"
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
#include "optimizer/prep.h"
Rewrite OR indexscan processing to be more flexible. We can now for the first time generate an OR indexscan for a two-column index when the WHERE condition is like 'col1 = foo AND (col2 = bar OR col2 = baz)' --- before, the OR had to be on the first column of the index or we'd not notice the possibility of using it. Some progress towards extracting OR indexscans from subclauses of an OR that references multiple relations, too, although this code is #ifdef'd out because it needs more work.
2004-01-04 01:07:32 +01:00
#include "optimizer/restrictinfo.h"
Tweak FOR UPDATE/SHARE error message wording (again) In commit 0ac5ad5134 I changed some error messages from "FOR UPDATE/SHARE" to a rather long gobbledygook which nobody liked. Then, in commit cb9b66d31 I changed them again, but the alternative chosen there was deemed suboptimal by Peter Eisentraut, who in message 1373937980.20441.8.camel@vanquo.pezone.net proposed an alternative involving a dynamically-constructed string based on the actual locking strength specified in the SQL command. This patch implements that suggestion.
2013-07-23 20:03:09 +02:00
#include "parser/analyze.h"
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
#include "rewrite/rewriteManip.h"
Final cleanup.
1999-07-16 07:00:38 +02:00
#include "utils/lsyscache.h"
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
#include "utils/typcache.h"
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/* These parameters are set by GUC */
int from_collapse_limit;
int join_collapse_limit;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* deconstruct_jointree requires multiple passes over the join tree, because we
* need to finish computing JoinDomains before we start distributing quals.
* As long as we have to do that, other information such as the relevant
* qualscopes might as well be computed in the first pass too.
*
* deconstruct_recurse recursively examines the join tree and builds a List
* (in depth-first traversal order) of JoinTreeItem structs, which are then
* processed iteratively by deconstruct_distribute. If there are outer
* joins, non-degenerate outer join clauses are processed in a third pass
* deconstruct_distribute_oj_quals.
*
* The JoinTreeItem structs themselves can be freed at the end of
* deconstruct_jointree, but do not modify or free their substructure,
* as the relid sets may also be pointed to by RestrictInfo and
* SpecialJoinInfo nodes.
*/
typedef struct JoinTreeItem
{
/* Fields filled during deconstruct_recurse: */
Node *jtnode; /* jointree node to examine */
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
JoinDomain *jdomain; /* join domain for its ON/WHERE clauses */
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
struct JoinTreeItem *jti_parent; /* JoinTreeItem for this node's
* parent, or NULL if it's the top */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
Relids qualscope; /* base+OJ Relids syntactically included in
* this jointree node */
Relids inner_join_rels; /* base+OJ Relids syntactically included
* in inner joins appearing at or below
* this jointree node */
Relids left_rels; /* if join node, Relids of the left side */
Relids right_rels; /* if join node, Relids of the right side */
Relids nonnullable_rels; /* if outer join, Relids of the
* non-nullable side */
/* Fields filled during deconstruct_distribute: */
SpecialJoinInfo *sjinfo; /* if outer join, its SpecialJoinInfo */
List *oj_joinclauses; /* outer join quals not yet distributed */
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
List *lateral_clauses; /* quals postponed from children due to
* lateral references */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
} JoinTreeItem;
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
static void extract_lateral_references(PlannerInfo *root, RelOptInfo *brel,
Index rtindex);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
static List *deconstruct_recurse(PlannerInfo *root, Node *jtnode,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
JoinDomain *parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *parent_jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
List **item_list);
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
static void deconstruct_distribute(PlannerInfo *root, JoinTreeItem *jtitem);
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
static void process_security_barrier_quals(PlannerInfo *root,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
int rti, JoinTreeItem *jtitem);
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
static void mark_rels_nulled_by_join(PlannerInfo *root, Index ojrelid,
Relids lower_rels);
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
static SpecialJoinInfo *make_outerjoininfo(PlannerInfo *root,
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Relids left_rels, Relids right_rels,
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
Relids inner_join_rels,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
JoinType jointype, Index ojrelid,
List *clause);
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
static void compute_semijoin_info(PlannerInfo *root, SpecialJoinInfo *sjinfo,
List *clause);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
static void deconstruct_distribute_oj_quals(PlannerInfo *root,
List *jtitems,
JoinTreeItem *jtitem);
static void distribute_quals_to_rels(PlannerInfo *root, List *clauses,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
SpecialJoinInfo *sjinfo,
Index security_level,
Relids qualscope,
Relids ojscope,
Relids outerjoin_nonnullable,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
Relids incompatible_relids,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bool allow_equivalence,
bool has_clone,
bool is_clone,
List **postponed_oj_qual_list);
Remove planner's private fields from Query struct, and put them into a new PlannerInfo struct, which is passed around instead of the bare Query in all the planning code. This commit is essentially just a code-beautification exercise, but it does open the door to making larger changes to the planner data structures without having to muck with the widely-known Query struct.
2005-06-06 00:32:58 +02:00
static void distribute_qual_to_rels(PlannerInfo *root, Node *clause,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
SpecialJoinInfo *sjinfo,
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
Index security_level,
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Relids qualscope,
Relids ojscope,
Fix planning of non-strict equivalence clauses above outer joins. If a potential equivalence clause references a variable from the nullable side of an outer join, the planner needs to take care that derived clauses are not pushed to below the outer join; else they may use the wrong value for the variable. (The problem arises only with non-strict clauses, since if an upper clause can be proven strict then the outer join will get simplified to a plain join.) The planner attempted to prevent this type of error by checking that potential equivalence clauses aren't outerjoin-delayed as a whole, but actually we have to check each side separately, since the two sides of the clause will get moved around separately if it's treated as an equivalence. Bugs of this type can be demonstrated as far back as 7.4, even though releases before 8.3 had only a very ad-hoc notion of equivalence clauses. In addition, we neglected to account for the possibility that such clauses might have nonempty nullable_relids even when not outerjoin-delayed; so the equivalence-class machinery lacked logic to compute correct nullable_relids values for clauses it constructs. This oversight was harmless before 9.2 because we were only using RestrictInfo.nullable_relids for OR clauses; but as of 9.2 it could result in pushing constructed equivalence clauses to incorrect places. (This accounts for bug #7604 from Bill MacArthur.) Fix the first problem by adding a new test check_equivalence_delay() in distribute_qual_to_rels, and fix the second one by adding code in equivclass.c and called functions to set correct nullable_relids for generated clauses. Although I believe the second part of this is not currently necessary before 9.2, I chose to back-patch it anyway, partly to keep the logic similar across branches and partly because it seems possible we might find other reasons why we need valid values of nullable_relids in the older branches. Add regression tests illustrating these problems. In 9.0 and up, also add test cases checking that we can push constants through outer joins, since we've broken that optimization before and I nearly broke it again with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
Relids outerjoin_nonnullable,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
Relids incompatible_relids,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bool allow_equivalence,
bool has_clone,
bool is_clone,
List **postponed_oj_qual_list);
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
static bool check_redundant_nullability_qual(PlannerInfo *root, Node *clause);
Fix some issues with wrong placement of pseudo-constant quals. initsplan.c figured that it could push Var-free qual clauses to the top of the current JoinDomain, which is okay in the abstract. But if the current domain is inside some outer join, and we later commute an inside-the-domain outer join with one outside it, we end up placing the pushed-up qual clause incorrectly. In distribute_qual_to_rels, avoid this by using the syntactic scope of the qual clause; with the exception that if we're in the top-level join domain we can still use the full query relid set, ensuring the resulting gating Result node goes to the top of the plan. (This is approximately as smart as the pre-v16 code was. Perhaps we can do better later, but it's not clear that such cases are worth a lot of sweat.) In process_implied_equality, we don't have a clear notion of syntactic scope, but we do have the results of SpecialJoinInfo construction. Thumb through those and remove any lower outer joins that might get commuted to above the join domain. Again, we can make an exception for the top-level join domain. It'd be possible to work harder here (for example, by keeping outer joins that aren't shown as potentially commutable), but I'm going to stop here for the moment. This issue has convinced me that the current representation of join domains probably needs further refinement, so I'm disinclined to write inessential dependent logic just yet. In passing, tighten the qualscope passed to process_implied_equality by generate_base_implied_equalities_no_const; there's no need for it to be larger than the rel we are currently considering. Tom Lane and Richard Guo, per report from Tender Wang. Discussion: https://postgr.es/m/CAHewXNk9eJ35ru5xATWioTV4+xZPHptjy9etdcNPjUfY9RQ+uQ@mail.gmail.com
2023-02-22 18:39:07 +01:00
static Relids get_join_domain_min_rels(PlannerInfo *root, Relids domain_relids);
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
static void check_mergejoinable(RestrictInfo *restrictinfo);
static void check_hashjoinable(RestrictInfo *restrictinfo);
Change the name of the Result Cache node to Memoize "Result Cache" was never a great name for this node, but nobody managed to come up with another name that anyone liked enough. That was until David Johnston mentioned "Node Memoization", which Tom Lane revised to just "Memoize". People seem to like "Memoize", so let's do the rename. Reviewed-by: Justin Pryzby Discussion: https://postgr.es/m/20210708165145.GG1176@momjian.us Backpatch-through: 14, where Result Cache was introduced
2021-07-14 02:43:58 +02:00
static void check_memoizable(RestrictInfo *restrictinfo);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*****************************************************************************
*
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
* JOIN TREES
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*****************************************************************************/
/*
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
* add_base_rels_to_query
*
* Scan the query's jointree and create baserel RelOptInfos for all
Build "other rels" of appendrel baserels in a separate step. Up to now, otherrel RelOptInfos were built at the same time as baserel RelOptInfos, thanks to recursion in build_simple_rel(). However, nothing in query_planner's preprocessing cares at all about otherrels, only baserels, so we don't really need to build them until just before we enter make_one_rel. This has two benefits: * create_lateral_join_info did a lot of extra work to propagate lateral-reference information from parents to the correct children. But if we delay creation of the children till after that, it's trivial (and much harder to break, too). * Since we have all the restriction quals correctly assigned to parent appendrels by this point, it'll be possible to do plan-time pruning and never make child RelOptInfos at all for partitions that can be pruned away. That's not done here, but will be later on. Amit Langote, reviewed at various times by Dilip Kumar, Jesper Pedersen, Yoshikazu Imai, and David Rowley Discussion: https://postgr.es/m/9d7c5112-cb99-6a47-d3be-cf1ee6862a1d@lab.ntt.co.jp
2019-03-26 23:21:10 +01:00
* the base relations (e.g., table, subquery, and function RTEs)
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
* appearing in the jointree.
Fix planner and rewriter to follow SQL semantics for tables that are mentioned in FROM but not elsewhere in the query: such tables should be joined over anyway. Aside from being more standards-compliant, this allows removal of some very ugly hacks for COUNT(*) processing. Also, allow HAVING clause without aggregate functions, since SQL does. Clean up CREATE RULE statement-list syntax the same way Bruce just fixed the main stmtmulti production. CAUTION: addition of a field to RangeTblEntry nodes breaks stored rules; you will have to initdb if you have any rules.
1999-10-07 06:23:24 +02:00
*
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
* The initial invocation must pass root->parse->jointree as the value of
* jtnode. Internally, the function recurses through the jointree.
*
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
* At the end of this process, there should be one baserel RelOptInfo for
Build "other rels" of appendrel baserels in a separate step. Up to now, otherrel RelOptInfos were built at the same time as baserel RelOptInfos, thanks to recursion in build_simple_rel(). However, nothing in query_planner's preprocessing cares at all about otherrels, only baserels, so we don't really need to build them until just before we enter make_one_rel. This has two benefits: * create_lateral_join_info did a lot of extra work to propagate lateral-reference information from parents to the correct children. But if we delay creation of the children till after that, it's trivial (and much harder to break, too). * Since we have all the restriction quals correctly assigned to parent appendrels by this point, it'll be possible to do plan-time pruning and never make child RelOptInfos at all for partitions that can be pruned away. That's not done here, but will be later on. Amit Langote, reviewed at various times by Dilip Kumar, Jesper Pedersen, Yoshikazu Imai, and David Rowley Discussion: https://postgr.es/m/9d7c5112-cb99-6a47-d3be-cf1ee6862a1d@lab.ntt.co.jp
2019-03-26 23:21:10 +01:00
* every non-join RTE that is used in the query. Some of the baserels
* may be appendrel parents, which will require additional "otherrel"
* RelOptInfos for their member rels, but those are added later.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
void
Remove planner's private fields from Query struct, and put them into a new PlannerInfo struct, which is passed around instead of the bare Query in all the planning code. This commit is essentially just a code-beautification exercise, but it does open the door to making larger changes to the planner data structures without having to muck with the widely-known Query struct.
2005-06-06 00:32:58 +02:00
add_base_rels_to_query(PlannerInfo *root, Node *jtnode)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
if (jtnode == NULL)
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
return;
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
if (IsA(jtnode, RangeTblRef))
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
int varno = ((RangeTblRef *) jtnode)->rtindex;
pgindent run. Make it all clean.
2001-03-22 05:01:46 +01:00
Abstract logic to allow for multiple kinds of child rels. Currently, the only type of child relation is an "other member rel", which is the child of a baserel, but in the future joins and even upper relations may have child rels. To facilitate that, introduce macros that test to test for particular RelOptKind values, and use them in various places where they help to clarify the sense of a test. (For example, a test may allow RELOPT_OTHER_MEMBER_REL either because it intends to allow child rels, or because it intends to allow simple rels.) Also, remove find_childrel_top_parent, which will not work for a child rel that is not a baserel. Instead, add a new RelOptInfo member top_parent_relids to track the same kind of information in a more generic manner. Ashutosh Bapat, slightly tweaked by me. Review and testing of the patch set from which this was taken by Rajkumar Raghuwanshi and Rafia Sabih. Discussion: http://postgr.es/m/CA+TgmoagTnF2yqR3PT2rv=om=wJiZ4-A+ATwdnriTGku1CLYxA@mail.gmail.com
2017-04-04 04:41:31 +02:00
(void) build_simple_rel(root, varno, NULL);
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
else if (IsA(jtnode, FromExpr))
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
{
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
FromExpr *f = (FromExpr *) jtnode;
Reimplement the linked list data structure used throughout the backend. In the past, we used a 'Lispy' linked list implementation: a "list" was merely a pointer to the head node of the list. The problem with that design is that it makes lappend() and length() linear time. This patch fixes that problem (and others) by maintaining a count of the list length and a pointer to the tail node along with each head node pointer. A "list" is now a pointer to a structure containing some meta-data about the list; the head and tail pointers in that structure refer to ListCell structures that maintain the actual linked list of nodes. The function names of the list API have also been changed to, I hope, be more logically consistent. By default, the old function names are still available; they will be disabled-by-default once the rest of the tree has been updated to use the new API names.
2004-05-26 06:41:50 +02:00
ListCell *l;
Minor cleanups and code beautification; eliminate some routines that are now dead code.
1999-08-10 05:00:15 +02:00
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
foreach(l, f->fromlist)
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
add_base_rels_to_query(root, lfirst(l));
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
else if (IsA(jtnode, JoinExpr))
{
JoinExpr *j = (JoinExpr *) jtnode;
pgindent run.
2002-09-04 22:31:48 +02:00
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
add_base_rels_to_query(root, j->larg);
add_base_rels_to_query(root, j->rarg);
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
else
Error message editing in backend/optimizer, backend/rewrite.
2003-07-25 02:01:09 +02:00
elog(ERROR, "unrecognized node type: %d",
(int) nodeTag(jtnode));
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Build "other rels" of appendrel baserels in a separate step. Up to now, otherrel RelOptInfos were built at the same time as baserel RelOptInfos, thanks to recursion in build_simple_rel(). However, nothing in query_planner's preprocessing cares at all about otherrels, only baserels, so we don't really need to build them until just before we enter make_one_rel. This has two benefits: * create_lateral_join_info did a lot of extra work to propagate lateral-reference information from parents to the correct children. But if we delay creation of the children till after that, it's trivial (and much harder to break, too). * Since we have all the restriction quals correctly assigned to parent appendrels by this point, it'll be possible to do plan-time pruning and never make child RelOptInfos at all for partitions that can be pruned away. That's not done here, but will be later on. Amit Langote, reviewed at various times by Dilip Kumar, Jesper Pedersen, Yoshikazu Imai, and David Rowley Discussion: https://postgr.es/m/9d7c5112-cb99-6a47-d3be-cf1ee6862a1d@lab.ntt.co.jp
2019-03-26 23:21:10 +01:00
/*
* add_other_rels_to_query
* create "otherrel" RelOptInfos for the children of appendrel baserels
*
* At the end of this process, there should be RelOptInfos for all relations
* that will be scanned by the query.
*/
void
add_other_rels_to_query(PlannerInfo *root)
{
int rti;
for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
RelOptInfo *rel = root->simple_rel_array[rti];
RangeTblEntry *rte = root->simple_rte_array[rti];
/* there may be empty slots corresponding to non-baserel RTEs */
if (rel == NULL)
continue;
/* Ignore any "otherrels" that were already added. */
if (rel->reloptkind != RELOPT_BASEREL)
continue;
/* If it's marked as inheritable, look for children. */
if (rte->inh)
Speed up planning when partitions can be pruned at plan time. Previously, the planner created RangeTblEntry and RelOptInfo structs for every partition of a partitioned table, even though many of them might later be deemed uninteresting thanks to partition pruning logic. This incurred significant overhead when there are many partitions. Arrange to postpone creation of these data structures until after we've processed the query enough to identify restriction quals for the partitioned table, and then apply partition pruning before not after creation of each partition's data structures. In this way we need not open the partition relations at all for partitions that the planner has no real interest in. For queries that can be proven at plan time to access only a small number of partitions, this patch improves the practical maximum number of partitions from under 100 to perhaps a few thousand. Amit Langote, reviewed at various times by Dilip Kumar, Jesper Pedersen, Yoshikazu Imai, and David Rowley Discussion: https://postgr.es/m/9d7c5112-cb99-6a47-d3be-cf1ee6862a1d@lab.ntt.co.jp
2019-03-30 23:58:55 +01:00
expand_inherited_rtentry(root, rel, rte, rti);
Build "other rels" of appendrel baserels in a separate step. Up to now, otherrel RelOptInfos were built at the same time as baserel RelOptInfos, thanks to recursion in build_simple_rel(). However, nothing in query_planner's preprocessing cares at all about otherrels, only baserels, so we don't really need to build them until just before we enter make_one_rel. This has two benefits: * create_lateral_join_info did a lot of extra work to propagate lateral-reference information from parents to the correct children. But if we delay creation of the children till after that, it's trivial (and much harder to break, too). * Since we have all the restriction quals correctly assigned to parent appendrels by this point, it'll be possible to do plan-time pruning and never make child RelOptInfos at all for partitions that can be pruned away. That's not done here, but will be later on. Amit Langote, reviewed at various times by Dilip Kumar, Jesper Pedersen, Yoshikazu Imai, and David Rowley Discussion: https://postgr.es/m/9d7c5112-cb99-6a47-d3be-cf1ee6862a1d@lab.ntt.co.jp
2019-03-26 23:21:10 +01:00
}
}
Deduce equality constraints that are implied by transitivity of mergejoinable qual clauses, and add them to the query quals. For example, WHERE a = b AND b = c will cause us to add AND a = c. This is necessary to ensure that it's safe to use these variables as interchangeable sort keys, which is something 7.0 knows how to do. Should provide a useful improvement in planning ability, too.
2000-07-24 05:11:01 +02:00
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
/*****************************************************************************
*
* TARGET LISTS
*
*****************************************************************************/
/*
* build_base_rel_tlists
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
* Add targetlist entries for each var needed in the query's final tlist
Delay creation of subplan tlist until after create_plan(). Once upon a time it was necessary for grouping_planner() to determine the tlist it wanted from the scan/join plan subtree before it called query_planner(), because query_planner() would actually make a Plan using that. But we refactored things a long time ago to delay construction of the Plan tree till later, so there's no need to build that tlist until (and indeed unless) we're ready to plaster it onto the Plan. The only thing query_planner() cares about is what Vars are going to be needed for the tlist, and it can perfectly well get that by looking at the real tlist rather than some masticated version. Well, actually, there is one minor glitch in that argument, which is that make_subplanTargetList also adds Vars appearing only in HAVING to the tlist it produces. So now we have to account for HAVING explicitly in build_base_rel_tlists. But that just adds a few lines of code, and I doubt it moves the needle much on processing time; we might be doing pull_var_clause() twice on the havingQual, but before we had it scanning dummy tlist entries instead. This is a very small down payment on rationalizing grouping_planner enough so it can be refactored.
2016-01-08 02:23:57 +01:00
* (and HAVING clause, if any) to the appropriate base relations.
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
*
* We mark such vars as needed by "relation 0" to ensure that they will
* propagate up through all join plan steps.
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
*/
void
Remove planner's private fields from Query struct, and put them into a new PlannerInfo struct, which is passed around instead of the bare Query in all the planning code. This commit is essentially just a code-beautification exercise, but it does open the door to making larger changes to the planner data structures without having to muck with the widely-known Query struct.
2005-06-06 00:32:58 +02:00
build_base_rel_tlists(PlannerInfo *root, List *final_tlist)
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
{
Fix estimate_num_groups() to not fail on PlaceHolderVars, per report from Stefan Kaltenbrunner. The most reasonable behavior (at least for the near term) seems to be to ignore the PlaceHolderVar and examine its argument instead. In support of this, change the API of pull_var_clause() to allow callers to request recursion into PlaceHolderVars. Currently estimate_num_groups() is the only customer for that behavior, but where there's one there may be others.
2009-04-19 21:46:33 +02:00
List *tlist_vars = pull_var_clause((Node *) final_tlist,
Refactor pull_var_clause's API to make it less tedious to extend. In commit 1d97c19a0f748e94 and later c1d9579dd8bf3c92, we extended pull_var_clause's API by adding enum-type arguments. That's sort of a pain to maintain, though, because it means every time we add a new behavior we must touch every last one of the call sites, even if there's a reasonable default behavior that most of them could use. Let's switch over to using a bitmask of flags, instead; that seems more maintainable and might save a nanosecond or two as well. This commit changes no behavior in itself, though I'm going to follow it up with one that does add a new behavior. In passing, remove flatten_tlist(), which has not been used since 9.1 and would otherwise need the same API changes. Removing these enums means that optimizer/tlist.h no longer needs to depend on optimizer/var.h. Changing that caused a number of C files to need addition of #include "optimizer/var.h" (probably we can thank old runs of pgrminclude for that); but on balance it seems like a good change anyway.
2016-03-10 21:52:58 +01:00
PVC_RECURSE_AGGREGATES |
Give pull_var_clause() reject/recurse/return behavior for WindowFuncs too. All along, this function should have treated WindowFuncs in a manner similar to Aggrefs, ie with an option whether or not to recurse into them. By not considering the case, it was always recursing, which is OK for most callers (although I suspect that the case in prepare_sort_from_pathkeys might represent a bug). But now we need return-without-recursing behavior as well. There are also more than a few callers that should never see a WindowFunc, and now we'll get some error checking on that.
2016-03-10 22:23:40 +01:00
PVC_RECURSE_WINDOWFUNCS |
Fix estimate_num_groups() to not fail on PlaceHolderVars, per report from Stefan Kaltenbrunner. The most reasonable behavior (at least for the near term) seems to be to ignore the PlaceHolderVar and examine its argument instead. In support of this, change the API of pull_var_clause() to allow callers to request recursion into PlaceHolderVars. Currently estimate_num_groups() is the only customer for that behavior, but where there's one there may be others.
2009-04-19 21:46:33 +02:00
PVC_INCLUDE_PLACEHOLDERS);
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
if (tlist_vars != NIL)
{
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
add_vars_to_targetlist(root, tlist_vars, bms_make_singleton(0));
Use the new List API function names throughout the backend, and disable the list compatibility API by default. While doing this, I decided to keep the llast() macro around and introduce llast_int() and llast_oid() variants.
2004-05-31 01:40:41 +02:00
list_free(tlist_vars);
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
}
Delay creation of subplan tlist until after create_plan(). Once upon a time it was necessary for grouping_planner() to determine the tlist it wanted from the scan/join plan subtree before it called query_planner(), because query_planner() would actually make a Plan using that. But we refactored things a long time ago to delay construction of the Plan tree till later, so there's no need to build that tlist until (and indeed unless) we're ready to plaster it onto the Plan. The only thing query_planner() cares about is what Vars are going to be needed for the tlist, and it can perfectly well get that by looking at the real tlist rather than some masticated version. Well, actually, there is one minor glitch in that argument, which is that make_subplanTargetList also adds Vars appearing only in HAVING to the tlist it produces. So now we have to account for HAVING explicitly in build_base_rel_tlists. But that just adds a few lines of code, and I doubt it moves the needle much on processing time; we might be doing pull_var_clause() twice on the havingQual, but before we had it scanning dummy tlist entries instead. This is a very small down payment on rationalizing grouping_planner enough so it can be refactored.
2016-01-08 02:23:57 +01:00
/*
Give pull_var_clause() reject/recurse/return behavior for WindowFuncs too. All along, this function should have treated WindowFuncs in a manner similar to Aggrefs, ie with an option whether or not to recurse into them. By not considering the case, it was always recursing, which is OK for most callers (although I suspect that the case in prepare_sort_from_pathkeys might represent a bug). But now we need return-without-recursing behavior as well. There are also more than a few callers that should never see a WindowFunc, and now we'll get some error checking on that.
2016-03-10 22:23:40 +01:00
* If there's a HAVING clause, we'll need the Vars it uses, too. Note
* that HAVING can contain Aggrefs but not WindowFuncs.
Delay creation of subplan tlist until after create_plan(). Once upon a time it was necessary for grouping_planner() to determine the tlist it wanted from the scan/join plan subtree before it called query_planner(), because query_planner() would actually make a Plan using that. But we refactored things a long time ago to delay construction of the Plan tree till later, so there's no need to build that tlist until (and indeed unless) we're ready to plaster it onto the Plan. The only thing query_planner() cares about is what Vars are going to be needed for the tlist, and it can perfectly well get that by looking at the real tlist rather than some masticated version. Well, actually, there is one minor glitch in that argument, which is that make_subplanTargetList also adds Vars appearing only in HAVING to the tlist it produces. So now we have to account for HAVING explicitly in build_base_rel_tlists. But that just adds a few lines of code, and I doubt it moves the needle much on processing time; we might be doing pull_var_clause() twice on the havingQual, but before we had it scanning dummy tlist entries instead. This is a very small down payment on rationalizing grouping_planner enough so it can be refactored.
2016-01-08 02:23:57 +01:00
*/
if (root->parse->havingQual)
{
List *having_vars = pull_var_clause(root->parse->havingQual,
Refactor pull_var_clause's API to make it less tedious to extend. In commit 1d97c19a0f748e94 and later c1d9579dd8bf3c92, we extended pull_var_clause's API by adding enum-type arguments. That's sort of a pain to maintain, though, because it means every time we add a new behavior we must touch every last one of the call sites, even if there's a reasonable default behavior that most of them could use. Let's switch over to using a bitmask of flags, instead; that seems more maintainable and might save a nanosecond or two as well. This commit changes no behavior in itself, though I'm going to follow it up with one that does add a new behavior. In passing, remove flatten_tlist(), which has not been used since 9.1 and would otherwise need the same API changes. Removing these enums means that optimizer/tlist.h no longer needs to depend on optimizer/var.h. Changing that caused a number of C files to need addition of #include "optimizer/var.h" (probably we can thank old runs of pgrminclude for that); but on balance it seems like a good change anyway.
2016-03-10 21:52:58 +01:00
PVC_RECURSE_AGGREGATES |
Delay creation of subplan tlist until after create_plan(). Once upon a time it was necessary for grouping_planner() to determine the tlist it wanted from the scan/join plan subtree before it called query_planner(), because query_planner() would actually make a Plan using that. But we refactored things a long time ago to delay construction of the Plan tree till later, so there's no need to build that tlist until (and indeed unless) we're ready to plaster it onto the Plan. The only thing query_planner() cares about is what Vars are going to be needed for the tlist, and it can perfectly well get that by looking at the real tlist rather than some masticated version. Well, actually, there is one minor glitch in that argument, which is that make_subplanTargetList also adds Vars appearing only in HAVING to the tlist it produces. So now we have to account for HAVING explicitly in build_base_rel_tlists. But that just adds a few lines of code, and I doubt it moves the needle much on processing time; we might be doing pull_var_clause() twice on the havingQual, but before we had it scanning dummy tlist entries instead. This is a very small down payment on rationalizing grouping_planner enough so it can be refactored.
2016-01-08 02:23:57 +01:00
PVC_INCLUDE_PLACEHOLDERS);
if (having_vars != NIL)
{
add_vars_to_targetlist(root, having_vars,
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
bms_make_singleton(0));
Delay creation of subplan tlist until after create_plan(). Once upon a time it was necessary for grouping_planner() to determine the tlist it wanted from the scan/join plan subtree before it called query_planner(), because query_planner() would actually make a Plan using that. But we refactored things a long time ago to delay construction of the Plan tree till later, so there's no need to build that tlist until (and indeed unless) we're ready to plaster it onto the Plan. The only thing query_planner() cares about is what Vars are going to be needed for the tlist, and it can perfectly well get that by looking at the real tlist rather than some masticated version. Well, actually, there is one minor glitch in that argument, which is that make_subplanTargetList also adds Vars appearing only in HAVING to the tlist it produces. So now we have to account for HAVING explicitly in build_base_rel_tlists. But that just adds a few lines of code, and I doubt it moves the needle much on processing time; we might be doing pull_var_clause() twice on the havingQual, but before we had it scanning dummy tlist entries instead. This is a very small down payment on rationalizing grouping_planner enough so it can be refactored.
2016-01-08 02:23:57 +01:00
list_free(having_vars);
}
}
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
}
/*
* add_vars_to_targetlist
* For each variable appearing in the list, add it to the owning
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
* relation's targetlist if not already present, and mark the variable
* as being needed for the indicated join (or for final output if
* where_needed includes "relation 0").
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
*
* The list may also contain PlaceHolderVars. These don't necessarily
* have a single owning relation; we keep their attr_needed info in
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
* root->placeholder_list instead. Find or create the associated
* PlaceHolderInfo entry, and update its ph_needed.
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
*/
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
void
Fix nested PlaceHolderVar expressions that appear only in targetlists. A PlaceHolderVar's expression might contain another, lower-level PlaceHolderVar. If the outer PlaceHolderVar is used, the inner one certainly will be also, and so we have to make sure that both of them get into the placeholder_list with correct ph_may_need values during the initial pre-scan of the query (before deconstruct_jointree starts). We did this correctly for PlaceHolderVars appearing in the query quals, but overlooked the issue for those appearing in the top-level targetlist; with the result that nested placeholders referenced only in the targetlist did not work correctly, as illustrated in bug #6154. While at it, add some error checking to find_placeholder_info to ensure that we don't try to create new placeholders after it's too late to do so; they have to all be created before deconstruct_jointree starts. Back-patch to 8.4 where the PlaceHolderVar mechanism was introduced.
2011-08-09 06:48:51 +02:00
add_vars_to_targetlist(PlannerInfo *root, List *vars,
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
Relids where_needed)
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
{
Reimplement the linked list data structure used throughout the backend. In the past, we used a 'Lispy' linked list implementation: a "list" was merely a pointer to the head node of the list. The problem with that design is that it makes lappend() and length() linear time. This patch fixes that problem (and others) by maintaining a count of the list length and a pointer to the tail node along with each head node pointer. A "list" is now a pointer to a structure containing some meta-data about the list; the head and tail pointers in that structure refer to ListCell structures that maintain the actual linked list of nodes. The function names of the list API have also been changed to, I hope, be more logically consistent. By default, the old function names are still available; they will be disabled-by-default once the rest of the tree has been updated to use the new API names.
2004-05-26 06:41:50 +02:00
ListCell *temp;
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
Assert(!bms_is_empty(where_needed));
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
foreach(temp, vars)
{
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
Node *node = (Node *) lfirst(temp);
if (IsA(node, Var))
{
Var *var = (Var *) node;
RelOptInfo *rel = find_base_rel(root, var->varno);
int attno = var->varattno;
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
if (bms_is_subset(where_needed, rel->relids))
continue;
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
Assert(attno >= rel->min_attr && attno <= rel->max_attr);
attno -= rel->min_attr;
if (rel->attr_needed[attno] == NULL)
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* Variable not yet requested, so add to rel's targetlist.
*
* The value available at the rel's scan level has not been
* nulled by any outer join, so drop its varnullingrels.
* (We'll put those back as we climb up the join tree.)
*/
var = copyObject(var);
var->varnullingrels = NULL;
rel->reltarget->exprs = lappend(rel->reltarget->exprs, var);
Add an explicit representation of the output targetlist to Paths. Up to now, there's been an assumption that all Paths for a given relation compute the same output column set (targetlist). However, there are good reasons to remove that assumption. For example, an indexscan on an expression index might be able to return the value of an expensive function "for free". While we have the ability to generate such a plan today in simple cases, we don't have a way to model that it's cheaper than a plan that computes the function from scratch, nor a way to create such a plan in join cases (where the function computation would normally happen at the topmost join node). Also, we need this so that we can have Paths representing post-scan/join steps, where the targetlist may well change from one step to the next. Therefore, invent a "struct PathTarget" representing the columns we expect a plan step to emit. It's convenient to include the output tuple width and tlist evaluation cost in this struct, and there will likely be additional fields in future. While Path nodes that actually do have custom outputs will need their own PathTargets, it will still be true that most Paths for a given relation will compute the same tlist. To reduce the overhead added by this patch, keep a "default PathTarget" in RelOptInfo, and allow Paths that compute that column set to just point to their parent RelOptInfo's reltarget. (In the patch as committed, actually every Path is like that, since we do not yet have any cases of custom PathTargets.) I took this opportunity to provide some more-honest costing of PlaceHolderVar evaluation. Up to now, the assumption that "scan/join reltargetlists have cost zero" was applied not only to Vars, where it's reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval cost of a PlaceHolderVar's expression to the first plan level where it can be computed, by including it in the PathTarget cost field and adding that to the cost estimates for Paths. This isn't perfect yet but it's much better than before, and there is a way forward to improve it more. This costing change affects the join order chosen for a couple of the regression tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
/* reltarget cost and width will be computed later */
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
}
rel->attr_needed[attno] = bms_add_members(rel->attr_needed[attno],
where_needed);
}
else if (IsA(node, PlaceHolderVar))
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
{
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
PlaceHolderVar *phv = (PlaceHolderVar *) node;
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
PlaceHolderInfo *phinfo = find_placeholder_info(root, phv);
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
phinfo->ph_needed = bms_add_members(phinfo->ph_needed,
where_needed);
Restructure building of join relation targetlists so that a join plan node emits only those vars that are actually needed above it in the plan tree. (There were comments in the code suggesting that this was done at some point in the dim past, but for a long time we have just made join nodes emit everything that either input emitted.) Aside from being marginally more efficient, this fixes the problem noted by Peter Eisentraut where a join above an IN-implemented-as-join might fail, because the subplan targetlist constructed in the latter case didn't meet the expectation of including everything. Along the way, fix some places that were O(N^2) in the targetlist length. This is not all the trouble spots for wide queries by any means, but it's a step forward.
2003-06-30 01:05:05 +02:00
}
Add a concept of "placeholder" variables to the planner. These are variables that represent some expression that we desire to compute below the top level of the plan, and then let that value "bubble up" as though it were a plain Var (ie, a column value). The immediate application is to allow sub-selects to be flattened even when they are below an outer join and have non-nullable output expressions. Formerly we couldn't flatten because such an expression wouldn't properly go to NULL when evaluated above the outer join. Now, we wrap it in a PlaceHolderVar and arrange for the actual evaluation to occur below the outer join. When the resulting Var bubbles up through the join, it will be set to NULL if necessary, yielding the correct results. This fixes a planner limitation that's existed since 7.1. In future we might want to use this mechanism to re-introduce some form of Hellerstein's "expensive functions" optimization, ie place the evaluation of an expensive function at the most suitable point in the plan tree.
2008-10-21 22:42:53 +02:00
else
elog(ERROR, "unrecognized node type: %d", (int) nodeTag(node));
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
}
}
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
/*****************************************************************************
*
* LATERAL REFERENCES
*
*****************************************************************************/
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
/*
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
* find_lateral_references
* For each LATERAL subquery, extract all its references to Vars and
* PlaceHolderVars of the current query level, and make sure those values
* will be available for evaluation of the subquery.
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
*
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
* While later planning steps ensure that the Var/PHV source rels are on the
* outside of nestloops relative to the LATERAL subquery, we also need to
* ensure that the Vars/PHVs propagate up to the nestloop join level; this
* means setting suitable where_needed values for them.
*
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
* Note that this only deals with lateral references in unflattened LATERAL
* subqueries. When we flatten a LATERAL subquery, its lateral references
* become plain Vars in the parent query, but they may have to be wrapped in
* PlaceHolderVars if they need to be forced NULL by outer joins that don't
* also null the LATERAL subquery. That's all handled elsewhere.
*
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
* This has to run before deconstruct_jointree, since it might result in
Remove ph_may_need from PlaceHolderInfo, with attendant simplifications. The planner logic that attempted to make a preliminary estimate of the ph_needed levels for PlaceHolderVars seems to be completely broken by lateral references. Fortunately, the potential join order optimization that this code supported seems to be of relatively little value in practice; so let's just get rid of it rather than trying to fix it. Getting rid of this allows fairly substantial simplifications in placeholder.c, too, so planning in such cases should be a bit faster. Issue noted while pursuing bugs reported by Jeremy Evans and Antonin Houska, though this doesn't in itself fix either of their reported cases. What this does do is prevent an Assert crash in the kind of query illustrated by the added regression test. (I'm not sure that the plan for that query is stable enough across platforms to be usable as a regression test output ... but we'll soon find out from the buildfarm.) Back-patch to 9.3. The problem case can't arise without LATERAL, so no need to touch older branches.
2013-08-15 00:38:32 +02:00
* creation of PlaceHolderInfos.
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
*/
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
void
find_lateral_references(PlannerInfo *root)
{
Index rti;
/* We need do nothing if the query contains no LATERAL RTEs */
if (!root->hasLateralRTEs)
return;
/*
* Examine all baserels (the rel array has been set up by now).
*/
for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
RelOptInfo *brel = root->simple_rel_array[rti];
/* there may be empty slots corresponding to non-baserel RTEs */
if (brel == NULL)
continue;
Assert(brel->relid == rti); /* sanity check on array */
/*
* This bit is less obvious than it might look. We ignore appendrel
* otherrels and consider only their parent baserels. In a case where
* a LATERAL-containing UNION ALL subquery was pulled up, it is the
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
* otherrel that is actually going to be in the plan. However, we
* want to mark all its lateral references as needed by the parent,
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
* because it is the parent's relid that will be used for join
* planning purposes. And the parent's RTE will contain all the
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
* lateral references we need to know, since the pulled-up member is
* nothing but a copy of parts of the original RTE's subquery. We
* could visit the parent's children instead and transform their
* references back to the parent's relid, but it would be much more
* complicated for no real gain. (Important here is that the child
* members have not yet received any processing beyond being pulled
* up.) Similarly, in appendrels created by inheritance expansion,
* it's sufficient to look at the parent relation.
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
*/
/* ignore RTEs that are "other rels" */
if (brel->reloptkind != RELOPT_BASEREL)
continue;
extract_lateral_references(root, brel, rti);
}
}
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
static void
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
extract_lateral_references(PlannerInfo *root, RelOptInfo *brel, Index rtindex)
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
{
RangeTblEntry *rte = root->simple_rte_array[rtindex];
List *vars;
List *newvars;
Relids where_needed;
ListCell *lc;
/* No cross-references are possible if it's not LATERAL */
if (!rte->lateral)
return;
/* Fetch the appropriate variables */
Redesign tablesample method API, and do extensive code review. The original implementation of TABLESAMPLE modeled the tablesample method API on index access methods, which wasn't a good choice because, without specialized DDL commands, there's no way to build an extension that can implement a TSM. (Raw inserts into system catalogs are not an acceptable thing to do, because we can't undo them during DROP EXTENSION, nor will pg_upgrade behave sanely.) Instead adopt an API more like procedural language handlers or foreign data wrappers, wherein the only SQL-level support object needed is a single handler function identified by having a special return type. This lets us get rid of the supporting catalog altogether, so that no custom DDL support is needed for the feature. Adjust the API so that it can support non-constant tablesample arguments (the original coding assumed we could evaluate the argument expressions at ExecInitSampleScan time, which is undesirable even if it weren't outright unsafe), and discourage sampling methods from looking at invisible tuples. Make sure that the BERNOULLI and SYSTEM methods are genuinely repeatable within and across queries, as required by the SQL standard, and deal more honestly with methods that can't support that requirement. Make a full code-review pass over the tablesample additions, and fix assorted bugs, omissions, infelicities, and cosmetic issues (such as failure to put the added code stanzas in a consistent ordering). Improve EXPLAIN's output of tablesample plans, too. Back-patch to 9.5 so that we don't have to support the original API in production.
2015-07-25 20:39:00 +02:00
if (rte->rtekind == RTE_RELATION)
vars = pull_vars_of_level((Node *) rte->tablesample, 0);
else if (rte->rtekind == RTE_SUBQUERY)
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
vars = pull_vars_of_level((Node *) rte->subquery, 1);
else if (rte->rtekind == RTE_FUNCTION)
Support multi-argument UNNEST(), and TABLE() syntax for multiple functions. This patch adds the ability to write TABLE( function1(), function2(), ...) as a single FROM-clause entry. The result is the concatenation of the first row from each function, followed by the second row from each function, etc; with NULLs inserted if any function produces fewer rows than others. This is believed to be a much more useful behavior than what Postgres currently does with multiple SRFs in a SELECT list. This syntax also provides a reasonable way to combine use of column definition lists with WITH ORDINALITY: put the column definition list inside TABLE(), where it's clear that it doesn't control the ordinality column as well. Also implement SQL-compliant multiple-argument UNNEST(), by turning UNNEST(a,b,c) into TABLE(unnest(a), unnest(b), unnest(c)). The SQL standard specifies TABLE() with only a single function, not multiple functions, and it seems to require an implicit UNNEST() which is not what this patch does. There may be something wrong with that reading of the spec, though, because if it's right then the spec's TABLE() is just a pointless alternative spelling of UNNEST(). After further review of that, we might choose to adopt a different syntax for what this patch does, but in any case this functionality seems clearly worthwhile. Andrew Gierth, reviewed by Zoltán Böszörményi and Heikki Linnakangas, and significantly revised by me
2013-11-22 01:37:02 +01:00
vars = pull_vars_of_level((Node *) rte->functions, 0);
Support XMLTABLE query expression XMLTABLE is defined by the SQL/XML standard as a feature that allows turning XML-formatted data into relational form, so that it can be used as a <table primary> in the FROM clause of a query. This new construct provides significant simplicity and performance benefit for XML data processing; what in a client-side custom implementation was reported to take 20 minutes can be executed in 400ms using XMLTABLE. (The same functionality was said to take 10 seconds using nested PostgreSQL XPath function calls, and 5 seconds using XMLReader under PL/Python). The implemented syntax deviates slightly from what the standard requires. First, the standard indicates that the PASSING clause is optional and that multiple XML input documents may be given to it; we make it mandatory and accept a single document only. Second, we don't currently support a default namespace to be specified. This implementation relies on a new executor node based on a hardcoded method table. (Because the grammar is fixed, there is no extensibility in the current approach; further constructs can be implemented on top of this such as JSON_TABLE, but they require changes to core code.) Author: Pavel Stehule, Álvaro Herrera Extensively reviewed by: Craig Ringer Discussion: https://postgr.es/m/CAFj8pRAgfzMD-LoSmnMGybD0WsEznLHWap8DO79+-GTRAPR4qA@mail.gmail.com
2017-03-08 16:39:37 +01:00
else if (rte->rtekind == RTE_TABLEFUNC)
vars = pull_vars_of_level((Node *) rte->tablefunc, 0);
More fixes for planner's handling of LATERAL. Re-allow subquery pullup for LATERAL subqueries, except when the subquery is below an outer join and contains lateral references to relations outside that outer join. If we pull up in such a case, we risk introducing lateral cross-references into outer joins' ON quals, which is something the code is entirely unprepared to cope with right now; and I'm not sure it'll ever be worth coping with. Support lateral refs in VALUES (this seems to be the only additional path type that needs such support as a consequence of re-allowing subquery pullup). Put in a slightly hacky fix for joinpath.c's refusal to consider parameterized join paths even when there cannot be any unparameterized ones. This was causing "could not devise a query plan for the given query" failures in queries involving more than two FROM items. Put in an even more hacky fix for distribute_qual_to_rels() being unhappy with join quals that contain references to rels outside their syntactic scope; which is to say, disable that test altogether. Need to think about how to preserve some sort of debugging cross-check here, while not expending more cycles than befits a debugging cross-check.
2012-08-12 22:01:26 +02:00
else if (rte->rtekind == RTE_VALUES)
vars = pull_vars_of_level((Node *) rte->values_lists, 0);
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
else
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
{
Assert(false);
return; /* keep compiler quiet */
}
if (vars == NIL)
return; /* nothing to do */
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
/* Copy each Var (or PlaceHolderVar) and adjust it to match our level */
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
newvars = NIL;
foreach(lc, vars)
{
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
Node *node = (Node *) lfirst(lc);
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
node = copyObject(node);
if (IsA(node, Var))
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
{
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
Var *var = (Var *) node;
/* Adjustment is easy since it's just one node */
var->varlevelsup = 0;
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
}
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
else if (IsA(node, PlaceHolderVar))
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
{
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
PlaceHolderVar *phv = (PlaceHolderVar *) node;
int levelsup = phv->phlevelsup;
/* Have to work harder to adjust the contained expression too */
if (levelsup != 0)
IncrementVarSublevelsUp(node, -levelsup, 0);
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
/*
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
* If we pulled the PHV out of a subquery RTE, its expression
* needs to be preprocessed. subquery_planner() already did this
* for level-zero PHVs in function and values RTEs, though.
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
*/
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
if (levelsup > 0)
phv->phexpr = preprocess_phv_expression(root, phv->phexpr);
Another round of planner fixes for LATERAL. Formerly, subquery pullup had no need to examine other entries in the range table, since they could not contain any references to the subquery being pulled up. That's no longer true with LATERAL, so now we need to be able to visit rangetable subexpressions to replace Vars referencing the pulled-up subquery. Also, this means that extract_lateral_references must be unsurprised at encountering lateral PlaceHolderVars, since such might be created when pulling up a subquery that's underneath an outer join with respect to the lateral reference.
2012-08-18 20:10:17 +02:00
}
else
Assert(false);
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
newvars = lappend(newvars, node);
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
}
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
list_free(vars);
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
/*
* We mark the Vars as being "needed" at the LATERAL RTE. This is a bit
* of a cheat: a more formal approach would be to mark each one as needed
* at the join of the LATERAL RTE with its source RTE. But it will work,
* and it's much less tedious than computing a separate where_needed for
* each Var.
*/
where_needed = bms_make_singleton(rtindex);
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
/*
* Push Vars into their source relations' targetlists, and PHVs into
* root->placeholder_list.
*/
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
add_vars_to_targetlist(root, newvars, where_needed);
Implement SQL-standard LATERAL subqueries. This patch implements the standard syntax of LATERAL attached to a sub-SELECT in FROM, and also allows LATERAL attached to a function in FROM, since set-returning function calls are expected to be one of the principal use-cases. The main change here is a rewrite of the mechanism for keeping track of which relations are visible for column references while the FROM clause is being scanned. The parser "namespace" lists are no longer lists of bare RTEs, but are lists of ParseNamespaceItem structs, which carry an RTE pointer as well as some visibility-controlling flags. Aside from supporting LATERAL correctly, this lets us get rid of the ancient hacks that required rechecking subqueries and JOIN/ON and function-in-FROM expressions for invalid references after they were initially parsed. Invalid column references are now always correctly detected on sight. In passing, remove assorted parser error checks that are now dead code by virtue of our having gotten rid of add_missing_from, as well as some comments that are obsolete for the same reason. (It was mainly add_missing_from that caused so much fudging here in the first place.) The planner support for this feature is very minimal, and will be improved in future patches. It works well enough for testing purposes, though. catversion bump forced due to new field in RangeTblEntry.
2012-08-08 01:02:54 +02:00
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
/* Remember the lateral references for create_lateral_join_info */
brel->lateral_vars = newvars;
}
/*
* create_lateral_join_info
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
* Fill in the per-base-relation direct_lateral_relids, lateral_relids
* and lateral_referencers sets.
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
*/
void
create_lateral_join_info(PlannerInfo *root)
{
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
bool found_laterals = false;
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
Index rti;
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
ListCell *lc;
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
/* We need do nothing if the query contains no LATERAL RTEs */
if (!root->hasLateralRTEs)
return;
Fix thinko in outer-join removal. If we have a RestrictInfo that mentions both the removal-candidate relation and the outer join's relid, then that is a pushed-down condition not a join condition, so it should be grounds for deciding that we can't remove the outer join. In commit 2489d76c4, I'd blindly included the OJ's relid into "joinrelids" as per the new standard convention, but the checks of attr_needed and ph_needed should only allow the join's input rels to be mentioned. Having done that, the check for references in pushed-down quals a few lines further down should be redundant. I left it in place as an Assert, though. While researching this I happened across a couple of comments that worried about the effects of update_placeholder_eval_levels. That's gone as of b448f1c8d, so we can remove some worry. Per bug #17769 from Robins Tharakan. The submitted test case triggers this more or less accidentally because we flatten out a LATERAL sub-select after we've done join strength reduction; if we did that in the other order, this problem would be masked because the outer join would get simplified to an inner join. To ensure that the committed test case will continue to test what it means to even if we make that happen someday, use a test clause involving COALESCE(), which will prevent us from using it to do join strength reduction. Patch by me, but thanks to Richard Guo for initial investigation. Discussion: https://postgr.es/m/17769-e4f7a5c9d84a80a7@postgresql.org
2023-02-04 21:19:54 +01:00
/* We'll need to have the ph_eval_at values for PlaceHolderVars */
Assert(root->placeholdersFrozen);
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
/*
* Examine all baserels (the rel array has been set up by now).
*/
for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
RelOptInfo *brel = root->simple_rel_array[rti];
Relids lateral_relids;
/* there may be empty slots corresponding to non-baserel RTEs */
if (brel == NULL)
continue;
Assert(brel->relid == rti); /* sanity check on array */
/* ignore RTEs that are "other rels" */
if (brel->reloptkind != RELOPT_BASEREL)
continue;
lateral_relids = NULL;
/* consider each laterally-referenced Var or PHV */
foreach(lc, brel->lateral_vars)
{
Node *node = (Node *) lfirst(lc);
if (IsA(node, Var))
{
Var *var = (Var *) node;
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
found_laterals = true;
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
lateral_relids = bms_add_member(lateral_relids,
var->varno);
}
else if (IsA(node, PlaceHolderVar))
{
PlaceHolderVar *phv = (PlaceHolderVar *) node;
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
PlaceHolderInfo *phinfo = find_placeholder_info(root, phv);
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
found_laterals = true;
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
lateral_relids = bms_add_members(lateral_relids,
phinfo->ph_eval_at);
}
else
Assert(false);
}
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
/* We now have all the simple lateral refs from this rel */
brel->direct_lateral_relids = lateral_relids;
brel->lateral_relids = bms_copy(lateral_relids);
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
}
/*
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
* Now check for lateral references within PlaceHolderVars, and mark their
* eval_at rels as having lateral references to the source rels.
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
*
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
* For a PHV that is due to be evaluated at a baserel, mark its source(s)
* as direct lateral dependencies of the baserel (adding onto the ones
* recorded above). If it's due to be evaluated at a join, mark its
* source(s) as indirect lateral dependencies of each baserel in the join,
* ie put them into lateral_relids but not direct_lateral_relids. This is
* appropriate because we can't put any such baserel on the outside of a
* join to one of the PHV's lateral dependencies, but on the other hand we
* also can't yet join it directly to the dependency.
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
*/
foreach(lc, root->placeholder_list)
{
PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc);
Relids eval_at = phinfo->ph_eval_at;
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
int varno;
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
if (phinfo->ph_lateral == NULL)
continue; /* PHV is uninteresting if no lateral refs */
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
found_laterals = true;
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
if (bms_get_singleton_member(eval_at, &varno))
{
/* Evaluation site is a baserel */
RelOptInfo *brel = find_base_rel(root, varno);
brel->direct_lateral_relids =
bms_add_members(brel->direct_lateral_relids,
phinfo->ph_lateral);
brel->lateral_relids =
bms_add_members(brel->lateral_relids,
phinfo->ph_lateral);
}
else
{
/* Evaluation site is a join */
varno = -1;
while ((varno = bms_next_member(eval_at, varno)) >= 0)
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
RelOptInfo *brel = find_base_rel_ignore_join(root, varno);
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
if (brel == NULL)
continue; /* ignore outer joins in eval_at */
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
brel->lateral_relids = bms_add_members(brel->lateral_relids,
phinfo->ph_lateral);
}
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
}
}
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
/*
* If we found no actual lateral references, we're done; but reset the
* hasLateralRTEs flag to avoid useless work later.
*/
if (!found_laterals)
{
root->hasLateralRTEs = false;
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
return;
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
}
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
/*
Get rid of the planner's LateralJoinInfo data structure. I originally modeled this data structure on SpecialJoinInfo, but after commit acfcd45cacb6df23 that looks like a pretty poor decision. All we really need is relid sets identifying laterally-referenced rels; and most of the time, what we want to know about includes indirect lateral references, a case the LateralJoinInfo data was unsuited to compute with any efficiency. The previous commit redefined RelOptInfo.lateral_relids as the transitive closure of lateral references, so that it easily supports checking indirect references. For the places where we really do want just direct references, add a new RelOptInfo field direct_lateral_relids, which is easily set up as a copy of lateral_relids before we perform the transitive closure calculation. Then we can just drop lateral_info_list and LateralJoinInfo and the supporting code. This makes the planner's handling of lateral references noticeably more efficient, and shorter too. Such a change can't be back-patched into stable branches for fear of breaking extensions that might be looking at the planner's data structures; but it seems not too late to push it into 9.5, so I've done so.
2015-12-11 21:52:16 +01:00
* Calculate the transitive closure of the lateral_relids sets, so that
* they describe both direct and indirect lateral references. If relation
* X references Y laterally, and Y references Z laterally, then we will
* have to scan X on the inside of a nestloop with Z, so for all intents
* and purposes X is laterally dependent on Z too.
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
*
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
* This code is essentially Warshall's algorithm for transitive closure.
* The outer loop considers each baserel, and propagates its lateral
* dependencies to those baserels that have a lateral dependency on it.
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
*/
for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
RelOptInfo *brel = root->simple_rel_array[rti];
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
Relids outer_lateral_relids;
Index rti2;
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
if (brel == NULL || brel->reloptkind != RELOPT_BASEREL)
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
continue;
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
/* need not consider baserel further if it has no lateral refs */
outer_lateral_relids = brel->lateral_relids;
if (outer_lateral_relids == NULL)
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
continue;
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
/* else scan all baserels */
for (rti2 = 1; rti2 < root->simple_rel_array_size; rti2++)
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
{
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
RelOptInfo *brel2 = root->simple_rel_array[rti2];
if (brel2 == NULL || brel2->reloptkind != RELOPT_BASEREL)
continue;
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
/* if brel2 has lateral ref to brel, propagate brel's refs */
if (bms_is_member(rti, brel2->lateral_relids))
brel2->lateral_relids = bms_add_members(brel2->lateral_relids,
outer_lateral_relids);
Fix planner problems with LATERAL references in PlaceHolderVars. The planner largely failed to consider the possibility that a PlaceHolderVar's expression might contain a lateral reference to a Var coming from somewhere outside the PHV's syntactic scope. We had a previous report of a problem in this area, which I tried to fix in a quick-hack way in commit 4da6439bd8553059766011e2a42c6e39df08717f, but Antonin Houska pointed out that there were still some problems, and investigation turned up other issues. This patch largely reverts that commit in favor of a more thoroughly thought-through solution. The new theory is that a PHV's ph_eval_at level cannot be higher than its original syntactic level. If it contains lateral references, those don't change the ph_eval_at level, but rather they create a lateral-reference requirement for the ph_eval_at join relation. The code in joinpath.c needs to handle that. Another issue is that createplan.c wasn't handling nested PlaceHolderVars properly. In passing, push knowledge of lateral-reference checks for join clauses into join_clause_is_movable_to. This is mainly so that FDWs don't need to deal with it. This patch doesn't fix the original join-qual-placement problem reported by Jeremy Evans (and indeed, one of the new regression test cases shows the wrong answer because of that). But the PlaceHolderVar problems need to be fixed before that issue can be addressed, so committing this separately seems reasonable.
2013-08-18 02:22:37 +02:00
}
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
}
/*
* Now that we've identified all lateral references, mark each baserel
* with the set of relids of rels that reference it laterally (possibly
* indirectly) --- that is, the inverse mapping of lateral_relids.
*/
for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
RelOptInfo *brel = root->simple_rel_array[rti];
Relids lateral_relids;
int rti2;
if (brel == NULL || brel->reloptkind != RELOPT_BASEREL)
continue;
/* Nothing to do at rels with no lateral refs */
lateral_relids = brel->lateral_relids;
Remove local optimizations of empty Bitmapsets into null pointers. These are all dead code now that it's done centrally. Patch by me; thanks to Nathan Bossart and Richard Guo for review. Discussion: https://postgr.es/m/1159933.1677621588@sss.pgh.pa.us
2023-03-02 18:01:47 +01:00
if (bms_is_empty(lateral_relids))
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
continue;
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
Remove local optimizations of empty Bitmapsets into null pointers. These are all dead code now that it's done centrally. Patch by me; thanks to Nathan Bossart and Richard Guo for review. Discussion: https://postgr.es/m/1159933.1677621588@sss.pgh.pa.us
2023-03-02 18:01:47 +01:00
/* No rel should have a lateral dependency on itself */
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
Assert(!bms_is_member(rti, lateral_relids));
/* Mark this rel's referencees */
rti2 = -1;
while ((rti2 = bms_next_member(lateral_relids, rti2)) >= 0)
{
RelOptInfo *brel2 = root->simple_rel_array[rti2];
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
if (brel2 == NULL)
continue; /* must be an OJ */
Assert(brel2->reloptkind == RELOPT_BASEREL);
Still more fixes for planner's handling of LATERAL references. More fuzz testing by Andreas Seltenreich exposed that the planner did not cope well with chains of lateral references. If relation X references Y laterally, and Y references Z laterally, then we will have to scan X on the inside of a nestloop with Z, so for all intents and purposes X is laterally dependent on Z too. The planner did not understand this and would generate intermediate joins that could not be used. While that was usually harmless except for wasting some planning cycles, under the right circumstances it would lead to "failed to build any N-way joins" or "could not devise a query plan" planner failures. To fix that, convert the existing per-relation lateral_relids and lateral_referencers relid sets into their transitive closures; that is, they now show all relations on which a rel is directly or indirectly laterally dependent. This not only fixes the chained-reference problem but allows some of the relevant tests to be made substantially simpler and faster, since they can be reduced to simple bitmap manipulations instead of searches of the LateralJoinInfo list. Also, when a PlaceHolderVar that is due to be evaluated at a join contains lateral references, we should treat those references as indirect lateral dependencies of each of the join's base relations. This prevents us from trying to join any individual base relations to the lateral reference source before the join is formed, which again cannot work. Andreas' testing also exposed another oversight in the "dangerous PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting unsafe join paths in joinpath.c is insufficient, because in some cases we will end up rejecting *all* possible paths for a particular join, again leading to "could not devise a query plan" failures. The restriction has to be known also to join_is_legal and its cohort functions, so that they will not select a join for which that will happen. I chose to move the supporting logic into joinrels.c where the latter functions are. Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
brel2->lateral_referencers =
bms_add_member(brel2->lateral_referencers, rti);
}
}
Fix up planner infrastructure to support LATERAL properly. This patch takes care of a number of problems having to do with failure to choose valid join orders and incorrect handling of lateral references pulled up from subqueries. Notable changes: * Add a LateralJoinInfo data structure similar to SpecialJoinInfo, to represent join ordering constraints created by lateral references. (I first considered extending the SpecialJoinInfo structure, but the semantics are different enough that a separate data structure seems better.) Extend join_is_legal() and related functions to prevent trying to form unworkable joins, and to ensure that we will consider joins that satisfy lateral references even if the joins would be clauseless. * Fill in the infrastructure needed for the last few types of relation scan paths to support parameterization. We'd have wanted this eventually anyway, but it is necessary now because a relation that gets pulled up out of a UNION ALL subquery may acquire a reltargetlist containing lateral references, meaning that its paths *have* to be parameterized whether or not we have any code that can push join quals down into the scan. * Compute data about lateral references early in query_planner(), and save in RelOptInfo nodes, to avoid repetitive calculations later. * Assorted corner-case bug fixes. There's probably still some bugs left, but this is a lot closer to being real than it was before.
2012-08-27 04:48:55 +02:00
}
Restructure representation of join alias variables. An explicit JOIN now has an RTE of its own, and references to its outputs now are Vars referencing the JOIN RTE, rather than CASE-expressions. This allows reverse-listing in ruleutils.c to use the correct alias easily, rather than painfully reverse-engineering the alias namespace as it used to do. Also, nested FULL JOINs work correctly, because the result of the inner joins are simple Vars that the planner can cope with. This fixes a bug reported a couple times now, notably by Tatsuo on 18-Nov-01. The alias Vars are expanded into COALESCE expressions where needed at the very end of planning, rather than during parsing. Also, beginnings of support for showing plan qualifier expressions in EXPLAIN. There are probably still cases that need work. initdb forced due to change of stored-rule representation.
2002-03-12 01:52:10 +01:00
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*****************************************************************************
*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* JOIN TREE PROCESSING
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*****************************************************************************/
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
/*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* deconstruct_jointree
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
* Recursively scan the query's join tree for WHERE and JOIN/ON qual
Simplify the planner's join clause management by storing join clauses of a relation in a flat 'joininfo' list. The former arrangement grouped the join clauses according to the set of unjoined relids used in each; however, profiling on test cases involving lots of joins proves that that data structure is a net loss. It takes more time to group the join clauses together than is saved by avoiding duplicate tests later. It doesn't help any that there are usually not more than one or two clauses per group ...
2005-06-09 06:19:00 +02:00
* clauses, and add these to the appropriate restrictinfo and joininfo
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* lists belonging to base RelOptInfos. Also, add SpecialJoinInfo nodes
* to root->join_info_list for any outer joins appearing in the query tree.
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* Return a "joinlist" data structure showing the join order decisions
* that need to be made by make_one_rel().
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* The "joinlist" result is a list of items that are either RangeTblRef
* jointree nodes or sub-joinlists. All the items at the same level of
* joinlist must be joined in an order to be determined by make_one_rel()
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* (note that legal orders may be constrained by SpecialJoinInfo nodes).
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* A sub-joinlist represents a subproblem to be planned separately. Currently
* sub-joinlists arise only from FULL OUTER JOIN or when collapsing of
* subproblems is stopped by join_collapse_limit or from_collapse_limit.
*/
List *
deconstruct_jointree(PlannerInfo *root)
{
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
List *result;
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
JoinDomain *top_jdomain;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
List *item_list = NIL;
ListCell *lc;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
/*
* After this point, no more PlaceHolderInfos may be made, because
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
* make_outerjoininfo requires all active placeholders to be present in
* root->placeholder_list while we crawl up the join tree.
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
*/
root->placeholdersFrozen = true;
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* Fetch the already-created top-level join domain for the query */
top_jdomain = linitial_node(JoinDomain, root->join_domains);
top_jdomain->jd_relids = NULL; /* filled during deconstruct_recurse */
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/* Start recursion at top of jointree */
Assert(root->parse->jointree != NULL &&
IsA(root->parse->jointree, FromExpr));
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* These are filled as we scan the jointree */
root->all_baserels = NULL;
root->outer_join_rels = NULL;
Compute correct em_nullable_relids in get_eclass_for_sort_expr(). Bug #8591 from Claudio Freire demonstrates that get_eclass_for_sort_expr must be able to compute valid em_nullable_relids for any new equivalence class members it creates. I'd worried about this in the commit message for db9f0e1d9a4a0842c814a464cdc9758c3f20b96c, but claimed that it wasn't a problem because multi-member ECs should already exist when it runs. That is transparently wrong, though, because this function is also called by initialize_mergeclause_eclasses, which runs during deconstruct_jointree. The example given in the bug report (which the new regression test item is based upon) fails because the COALESCE() expression is first seen by initialize_mergeclause_eclasses rather than process_equivalence. Fixing this requires passing the appropriate nullable_relids set to get_eclass_for_sort_expr, and it requires new code to compute that set for top-level expressions such as ORDER BY, GROUP BY, etc. We store the top-level nullable_relids in a new field in PlannerInfo to avoid computing it many times. In the back branches, I've added the new field at the end of the struct to minimize ABI breakage for planner plugins. There doesn't seem to be a good alternative to changing get_eclass_for_sort_expr's API signature, though. There probably aren't any third-party extensions calling that function directly; moreover, if there are, they probably need to think about what to pass for nullable_relids anyway. Back-patch to 9.2, like the previous patch in this area.
2013-11-15 22:46:18 +01:00
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Perform the initial scan of the jointree */
result = deconstruct_recurse(root, (Node *) root->parse->jointree,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
top_jdomain, NULL,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
&item_list);
/* Now we can form the value of all_query_rels, too */
root->all_query_rels = bms_union(root->all_baserels, root->outer_join_rels);
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* ... which should match what we computed for the top join domain */
Assert(bms_equal(root->all_query_rels, top_jdomain->jd_relids));
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Now scan all the jointree nodes again, and distribute quals */
foreach(lc, item_list)
{
JoinTreeItem *jtitem = (JoinTreeItem *) lfirst(lc);
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
deconstruct_distribute(root, jtitem);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
}
/*
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* If there were any special joins then we may have some postponed LEFT
* JOIN clauses to deal with.
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
*/
if (root->join_info_list)
{
foreach(lc, item_list)
{
JoinTreeItem *jtitem = (JoinTreeItem *) lfirst(lc);
if (jtitem->oj_joinclauses != NIL)
deconstruct_distribute_oj_quals(root, item_list, jtitem);
}
}
/* Don't need the JoinTreeItems any more */
list_free_deep(item_list);
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
return result;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
}
/*
* deconstruct_recurse
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* One recursion level of deconstruct_jointree's initial jointree scan.
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
* jtnode is the jointree node to examine, and parent_domain is the
* enclosing join domain. (We must add all base+OJ relids appearing
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* here or below to parent_domain.) parent_jtitem is the JoinTreeItem
* for the parent jointree node, or NULL at the top of the recursion.
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
*
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* item_list is an in/out parameter: we add a JoinTreeItem struct to
* that list for each jointree node, in depth-first traversal order.
* (Hence, after each call, the last list item corresponds to its jtnode.)
*
* Return value is the appropriate joinlist for this jointree node.
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*/
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
static List *
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
deconstruct_recurse(PlannerInfo *root, Node *jtnode,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
JoinDomain *parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *parent_jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
List **item_list)
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
{
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
List *joinlist;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
JoinTreeItem *jtitem;
Assert(jtnode != NULL);
/* Make the new JoinTreeItem, but don't add it to item_list yet */
jtitem = palloc0_object(JoinTreeItem);
jtitem->jtnode = jtnode;
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem->jti_parent = parent_jtitem;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
if (IsA(jtnode, RangeTblRef))
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
{
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
int varno = ((RangeTblRef *) jtnode)->rtindex;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Fill all_baserels as we encounter baserel jointree nodes */
root->all_baserels = bms_add_member(root->all_baserels, varno);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* This node belongs to parent_domain */
jtitem->jdomain = parent_domain;
parent_domain->jd_relids = bms_add_member(parent_domain->jd_relids,
varno);
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
/* qualscope is just the one RTE */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->qualscope = bms_make_singleton(varno);
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
/* A single baserel does not create an inner join */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->inner_join_rels = NULL;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
joinlist = list_make1(jtnode);
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
}
else if (IsA(jtnode, FromExpr))
{
FromExpr *f = (FromExpr *) jtnode;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
int remaining;
Reimplement the linked list data structure used throughout the backend. In the past, we used a 'Lispy' linked list implementation: a "list" was merely a pointer to the head node of the list. The problem with that design is that it makes lappend() and length() linear time. This patch fixes that problem (and others) by maintaining a count of the list length and a pointer to the tail node along with each head node pointer. A "list" is now a pointer to a structure containing some meta-data about the list; the head and tail pointers in that structure refer to ListCell structures that maintain the actual linked list of nodes. The function names of the list API have also been changed to, I hope, be more logically consistent. By default, the old function names are still available; they will be disabled-by-default once the rest of the tree has been updated to use the new API names.
2004-05-26 06:41:50 +02:00
ListCell *l;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* This node belongs to parent_domain, as do its children */
jtitem->jdomain = parent_domain;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
/*
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* Recurse to handle child nodes, and compute output joinlist. We
* collapse subproblems into a single joinlist whenever the resulting
* joinlist wouldn't exceed from_collapse_limit members. Also, always
* collapse one-element subproblems, since that won't lengthen the
* joinlist anyway.
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*/
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->qualscope = NULL;
jtitem->inner_join_rels = NULL;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
joinlist = NIL;
remaining = list_length(f->fromlist);
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
foreach(l, f->fromlist)
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
JoinTreeItem *sub_item;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
List *sub_joinlist;
int sub_members;
sub_joinlist = deconstruct_recurse(root, lfirst(l),
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
sub_item = (JoinTreeItem *) llast(*item_list);
jtitem->qualscope = bms_add_members(jtitem->qualscope,
sub_item->qualscope);
jtitem->inner_join_rels = sub_item->inner_join_rels;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
sub_members = list_length(sub_joinlist);
remaining--;
if (sub_members <= 1 ||
list_length(joinlist) + sub_members + remaining <= from_collapse_limit)
joinlist = list_concat(joinlist, sub_joinlist);
else
joinlist = lappend(joinlist, sub_joinlist);
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
}
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
/*
* A FROM with more than one list element is an inner join subsuming
* all below it, so we should report inner_join_rels = qualscope. If
* there was exactly one element, we should (and already did) report
* whatever its inner_join_rels were. If there were no elements (is
In the planner, replace an empty FROM clause with a dummy RTE. The fact that "SELECT expression" has no base relations has long been a thorn in the side of the planner. It makes it hard to flatten a sub-query that looks like that, or is a trivial VALUES() item, because the planner generally uses relid sets to identify sub-relations, and such a sub-query would have an empty relid set if we flattened it. prepjointree.c contains some baroque logic that works around this in certain special cases --- but there is a much better answer. We can replace an empty FROM clause with a dummy RTE that acts like a table of one row and no columns, and then there are no such corner cases to worry about. Instead we need some logic to get rid of useless dummy RTEs, but that's simpler and covers more cases than what was there before. For really trivial cases, where the query is just "SELECT expression" and nothing else, there's a hazard that adding the extra RTE makes for a noticeable slowdown; even though it's not much processing, there's not that much for the planner to do overall. However testing says that the penalty is very small, close to the noise level. In more complex queries, this is able to find optimizations that we could not find before. The new RTE type is called RTE_RESULT, since the "scan" plan type it gives rise to is a Result node (the same plan we produced for a "SELECT expression" query before). To avoid confusion, rename the old ResultPath path type to GroupResultPath, reflecting that it's only used in degenerate grouping cases where we know the query produces just one grouped row. (It wouldn't work to unify the two cases, because there are different rules about where the associated quals live during query_planner.) Note: although this touches readfuncs.c, I don't think a catversion bump is required, because the added case can't occur in stored rules, only plans. Patch by me, reviewed by David Rowley and Mark Dilger Discussion: https://postgr.es/m/15944.1521127664@sss.pgh.pa.us
2019-01-28 23:54:10 +01:00
* that still possible?) the initialization before the loop fixed it.
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
*/
if (list_length(f->fromlist) > 1)
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->inner_join_rels = jtitem->qualscope;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
else if (IsA(jtnode, JoinExpr))
{
JoinExpr *j = (JoinExpr *) jtnode;
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
JoinDomain *child_domain,
*fj_domain;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
JoinTreeItem *left_item,
*right_item;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
List *leftjoinlist,
*rightjoinlist;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
switch (j->jointype)
{
case JOIN_INNER:
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* This node belongs to parent_domain, as do its children */
jtitem->jdomain = parent_domain;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Recurse */
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
leftjoinlist = deconstruct_recurse(root, j->larg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
left_item = (JoinTreeItem *) llast(*item_list);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
rightjoinlist = deconstruct_recurse(root, j->rarg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
right_item = (JoinTreeItem *) llast(*item_list);
/* Compute qualscope etc */
jtitem->qualscope = bms_union(left_item->qualscope,
right_item->qualscope);
jtitem->inner_join_rels = jtitem->qualscope;
jtitem->left_rels = left_item->qualscope;
jtitem->right_rels = right_item->qualscope;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
/* Inner join adds no restrictions for quals */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->nonnullable_rels = NULL;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
break;
case JOIN_LEFT:
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
case JOIN_ANTI:
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* Make new join domain for my quals and the RHS */
child_domain = makeNode(JoinDomain);
child_domain->jd_relids = NULL; /* filled by recursion */
root->join_domains = lappend(root->join_domains, child_domain);
jtitem->jdomain = child_domain;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Recurse */
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
leftjoinlist = deconstruct_recurse(root, j->larg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
left_item = (JoinTreeItem *) llast(*item_list);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
rightjoinlist = deconstruct_recurse(root, j->rarg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
child_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
right_item = (JoinTreeItem *) llast(*item_list);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* Compute join domain contents, qualscope etc */
parent_domain->jd_relids =
bms_add_members(parent_domain->jd_relids,
child_domain->jd_relids);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->qualscope = bms_union(left_item->qualscope,
right_item->qualscope);
/* caution: ANTI join derived from SEMI will lack rtindex */
if (j->rtindex != 0)
{
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain->jd_relids =
bms_add_member(parent_domain->jd_relids,
j->rtindex);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->qualscope = bms_add_member(jtitem->qualscope,
j->rtindex);
root->outer_join_rels = bms_add_member(root->outer_join_rels,
j->rtindex);
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
mark_rels_nulled_by_join(root, j->rtindex,
right_item->qualscope);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
}
jtitem->inner_join_rels = bms_union(left_item->inner_join_rels,
right_item->inner_join_rels);
jtitem->left_rels = left_item->qualscope;
jtitem->right_rels = right_item->qualscope;
jtitem->nonnullable_rels = left_item->qualscope;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
break;
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
case JOIN_SEMI:
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* This node belongs to parent_domain, as do its children */
jtitem->jdomain = parent_domain;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Recurse */
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
leftjoinlist = deconstruct_recurse(root, j->larg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
left_item = (JoinTreeItem *) llast(*item_list);
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
rightjoinlist = deconstruct_recurse(root, j->rarg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
right_item = (JoinTreeItem *) llast(*item_list);
/* Compute qualscope etc */
jtitem->qualscope = bms_union(left_item->qualscope,
right_item->qualscope);
/* SEMI join never has rtindex, so don't add to anything */
Assert(j->rtindex == 0);
jtitem->inner_join_rels = bms_union(left_item->inner_join_rels,
right_item->inner_join_rels);
jtitem->left_rels = left_item->qualscope;
jtitem->right_rels = right_item->qualscope;
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
/* Semi join adds no restrictions for quals */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->nonnullable_rels = NULL;
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
break;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
case JOIN_FULL:
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* The FULL JOIN's quals need their very own domain */
fj_domain = makeNode(JoinDomain);
root->join_domains = lappend(root->join_domains, fj_domain);
jtitem->jdomain = fj_domain;
/* Recurse, giving each side its own join domain */
child_domain = makeNode(JoinDomain);
child_domain->jd_relids = NULL; /* filled by recursion */
root->join_domains = lappend(root->join_domains, child_domain);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
leftjoinlist = deconstruct_recurse(root, j->larg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
child_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
left_item = (JoinTreeItem *) llast(*item_list);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
fj_domain->jd_relids = bms_copy(child_domain->jd_relids);
child_domain = makeNode(JoinDomain);
child_domain->jd_relids = NULL; /* filled by recursion */
root->join_domains = lappend(root->join_domains, child_domain);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
rightjoinlist = deconstruct_recurse(root, j->rarg,
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
child_domain,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
item_list);
right_item = (JoinTreeItem *) llast(*item_list);
/* Compute qualscope etc */
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
fj_domain->jd_relids = bms_add_members(fj_domain->jd_relids,
child_domain->jd_relids);
parent_domain->jd_relids = bms_add_members(parent_domain->jd_relids,
fj_domain->jd_relids);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->qualscope = bms_union(left_item->qualscope,
right_item->qualscope);
Assert(j->rtindex != 0);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
parent_domain->jd_relids = bms_add_member(parent_domain->jd_relids,
j->rtindex);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->qualscope = bms_add_member(jtitem->qualscope,
j->rtindex);
root->outer_join_rels = bms_add_member(root->outer_join_rels,
j->rtindex);
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
mark_rels_nulled_by_join(root, j->rtindex,
left_item->qualscope);
mark_rels_nulled_by_join(root, j->rtindex,
right_item->qualscope);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->inner_join_rels = bms_union(left_item->inner_join_rels,
right_item->inner_join_rels);
jtitem->left_rels = left_item->qualscope;
jtitem->right_rels = right_item->qualscope;
Be smart about outer-join qualifications that mention only one side of the join, per recent discussion on pgsql-sql. Not clear that this will come up often in real queries, but it's not any more expensive to do it right, so we may as well do it right.
2003-03-03 00:46:34 +01:00
/* each side is both outer and inner */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->nonnullable_rels = jtitem->qualscope;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
break;
default:
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
/* JOIN_RIGHT was eliminated during reduce_outer_joins() */
Error message editing in backend/optimizer, backend/rewrite.
2003-07-25 02:01:09 +02:00
elog(ERROR, "unrecognized join type: %d",
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
(int) j->jointype);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
leftjoinlist = rightjoinlist = NIL; /* keep compiler quiet */
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
break;
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* Compute the output joinlist. We fold subproblems together except
* at a FULL JOIN or where join_collapse_limit would be exceeded.
*/
if (j->jointype == JOIN_FULL)
{
/* force the join order exactly at this node */
joinlist = list_make1(list_make2(leftjoinlist, rightjoinlist));
}
else if (list_length(leftjoinlist) + list_length(rightjoinlist) <=
join_collapse_limit)
{
/* OK to combine subproblems */
joinlist = list_concat(leftjoinlist, rightjoinlist);
}
else
{
/* can't combine, but needn't force join order above here */
Node *leftpart,
*rightpart;
/* avoid creating useless 1-element sublists */
if (list_length(leftjoinlist) == 1)
leftpart = (Node *) linitial(leftjoinlist);
else
leftpart = (Node *) leftjoinlist;
if (list_length(rightjoinlist) == 1)
rightpart = (Node *) linitial(rightjoinlist);
else
rightpart = (Node *) rightjoinlist;
joinlist = list_make2(leftpart, rightpart);
}
}
else
{
elog(ERROR, "unrecognized node type: %d",
(int) nodeTag(jtnode));
joinlist = NIL; /* keep compiler quiet */
}
/* Finally, we can add the new JoinTreeItem to item_list */
*item_list = lappend(*item_list, jtitem);
return joinlist;
}
/*
* deconstruct_distribute
* Process one jointree node in phase 2 of deconstruct_jointree processing.
*
* Distribute quals of the node to appropriate restriction and join lists.
* In addition, entries will be added to root->join_info_list for outer joins.
*/
static void
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
deconstruct_distribute(PlannerInfo *root, JoinTreeItem *jtitem)
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
{
Node *jtnode = jtitem->jtnode;
if (IsA(jtnode, RangeTblRef))
{
int varno = ((RangeTblRef *) jtnode)->rtindex;
/* Deal with any securityQuals attached to the RTE */
if (root->qual_security_level > 0)
process_security_barrier_quals(root,
varno,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
}
else if (IsA(jtnode, FromExpr))
{
FromExpr *f = (FromExpr *) jtnode;
/*
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* Process any lateral-referencing quals that were postponed to this
* level by children.
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
*/
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
distribute_quals_to_rels(root, jtitem->lateral_clauses,
jtitem,
NULL,
root->qual_security_level,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
jtitem->qualscope,
NULL, NULL, NULL,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
true, false, false,
NULL);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* Now process the top-level quals.
*/
distribute_quals_to_rels(root, (List *) f->quals,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
NULL,
root->qual_security_level,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
jtitem->qualscope,
NULL, NULL, NULL,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
true, false, false,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
NULL);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
}
else if (IsA(jtnode, JoinExpr))
{
JoinExpr *j = (JoinExpr *) jtnode;
Relids ojscope;
List *my_quals;
SpecialJoinInfo *sjinfo;
List **postponed_oj_qual_list;
Fix bogus handling of "postponed" lateral quals. When pulling a "postponed" qual from a LATERAL subquery up into the quals of an outer join, we must make sure that the postponed qual is included in those seen by make_outerjoininfo(). Otherwise we might compute a too-small min_lefthand or min_righthand for the outer join, leading to "JOIN qualification cannot refer to other relations" failures from distribute_qual_to_rels. Subtler errors in the created plan seem possible, too, if the extra qual would only affect join ordering constraints. Per bug #9041 from David Leverton. Back-patch to 9.3.
2014-01-30 20:51:16 +01:00
/*
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* Include lateral-referencing quals postponed from children in
* my_quals, so that they'll be handled properly in
* make_outerjoininfo. (This is destructive to
* jtitem->lateral_clauses, but we won't use that again.)
Fix bogus handling of "postponed" lateral quals. When pulling a "postponed" qual from a LATERAL subquery up into the quals of an outer join, we must make sure that the postponed qual is included in those seen by make_outerjoininfo(). Otherwise we might compute a too-small min_lefthand or min_righthand for the outer join, leading to "JOIN qualification cannot refer to other relations" failures from distribute_qual_to_rels. Subtler errors in the created plan seem possible, too, if the extra qual would only affect join ordering constraints. Per bug #9041 from David Leverton. Back-patch to 9.3.
2014-01-30 20:51:16 +01:00
*/
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
my_quals = list_concat(jtitem->lateral_clauses,
(List *) j->quals);
Fix bogus handling of "postponed" lateral quals. When pulling a "postponed" qual from a LATERAL subquery up into the quals of an outer join, we must make sure that the postponed qual is included in those seen by make_outerjoininfo(). Otherwise we might compute a too-small min_lefthand or min_righthand for the outer join, leading to "JOIN qualification cannot refer to other relations" failures from distribute_qual_to_rels. Subtler errors in the created plan seem possible, too, if the extra qual would only affect join ordering constraints. Per bug #9041 from David Leverton. Back-patch to 9.3.
2014-01-30 20:51:16 +01:00
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/*
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* For an OJ, form the SpecialJoinInfo now, so that we can pass it to
* distribute_qual_to_rels. We must compute its ojscope too.
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
*
* Semijoins are a bit of a hybrid: we build a SpecialJoinInfo, but we
* want ojscope = NULL for distribute_qual_to_rels.
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
*/
if (j->jointype != JOIN_INNER)
{
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
sjinfo = make_outerjoininfo(root,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->left_rels,
jtitem->right_rels,
jtitem->inner_join_rels,
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
j->jointype,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
j->rtindex,
Fix bogus handling of "postponed" lateral quals. When pulling a "postponed" qual from a LATERAL subquery up into the quals of an outer join, we must make sure that the postponed qual is included in those seen by make_outerjoininfo(). Otherwise we might compute a too-small min_lefthand or min_righthand for the outer join, leading to "JOIN qualification cannot refer to other relations" failures from distribute_qual_to_rels. Subtler errors in the created plan seem possible, too, if the extra qual would only affect join ordering constraints. Per bug #9041 from David Leverton. Back-patch to 9.3.
2014-01-30 20:51:16 +01:00
my_quals);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
jtitem->sjinfo = sjinfo;
Get rid of the rather fuzzily defined FlattenedSubLink node type in favor of making pull_up_sublinks() construct a full-blown JoinExpr tree representation of IN/EXISTS SubLinks that it is able to convert to semi or anti joins. This makes pull_up_sublinks() a shade more complex, but the gain in semantic clarity is worth it. I still have more to do in this area to address the previously-discussed problems, but this commit in itself fixes at least one bug in HEAD, as shown by added regression test case.
2009-02-25 04:30:38 +01:00
if (j->jointype == JOIN_SEMI)
ojscope = NULL;
else
ojscope = bms_union(sjinfo->min_lefthand,
sjinfo->min_righthand);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
}
else
{
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
sjinfo = NULL;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
ojscope = NULL;
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* If it's a left join with a join clause that is strict for the LHS,
* then we need to postpone handling of any non-degenerate join
* clauses, in case the join is able to commute with another left join
* per identity 3. (Degenerate clauses need not be postponed, since
* they will drop down below this join anyway.)
*/
if (j->jointype == JOIN_LEFT && sjinfo->lhs_strict)
Fix thinko in qual distribution. deconstruct_distribute tweaks the outer join scope (ojscope) it passes to distribute_qual_to_rels when considering an outer join qual that's above potentially-commutable outer joins. However, if the current join is *not* potentially commutable, we shouldn't do that. The argument that distribute_qual_to_rels will not do something wrong with the bogus ojscope falls flat if we don't pass it non-null postponed_oj_qual_list. Moreover, there's no need to play games in this case since we aren't going to commute anything. Per SQLSmith testing by Robins Tharakan. Discussion: https://postgr.es/m/CAEP4nAw74k4b-=93gmfCNX3MOY3y4uPxqbk_MnCVEpdsqHJVsg@mail.gmail.com
2023-02-04 23:40:35 +01:00
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
postponed_oj_qual_list = &jtitem->oj_joinclauses;
Fix thinko in qual distribution. deconstruct_distribute tweaks the outer join scope (ojscope) it passes to distribute_qual_to_rels when considering an outer join qual that's above potentially-commutable outer joins. However, if the current join is *not* potentially commutable, we shouldn't do that. The argument that distribute_qual_to_rels will not do something wrong with the bogus ojscope falls flat if we don't pass it non-null postponed_oj_qual_list. Moreover, there's no need to play games in this case since we aren't going to commute anything. Per SQLSmith testing by Robins Tharakan. Discussion: https://postgr.es/m/CAEP4nAw74k4b-=93gmfCNX3MOY3y4uPxqbk_MnCVEpdsqHJVsg@mail.gmail.com
2023-02-04 23:40:35 +01:00
/*
* Add back any commutable lower OJ relids that were removed from
* min_lefthand or min_righthand, else the ojscope cross-check in
* distribute_qual_to_rels will complain. Since we are postponing
* processing of non-degenerate clauses, this addition doesn't
* affect anything except that cross-check. Real clause
* positioning decisions will be made later, when we revisit the
* postponed clauses.
*/
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
ojscope = bms_add_members(ojscope, sjinfo->commute_below_l);
ojscope = bms_add_members(ojscope, sjinfo->commute_below_r);
Fix thinko in qual distribution. deconstruct_distribute tweaks the outer join scope (ojscope) it passes to distribute_qual_to_rels when considering an outer join qual that's above potentially-commutable outer joins. However, if the current join is *not* potentially commutable, we shouldn't do that. The argument that distribute_qual_to_rels will not do something wrong with the bogus ojscope falls flat if we don't pass it non-null postponed_oj_qual_list. Moreover, there's no need to play games in this case since we aren't going to commute anything. Per SQLSmith testing by Robins Tharakan. Discussion: https://postgr.es/m/CAEP4nAw74k4b-=93gmfCNX3MOY3y4uPxqbk_MnCVEpdsqHJVsg@mail.gmail.com
2023-02-04 23:40:35 +01:00
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
else
postponed_oj_qual_list = NULL;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Process the JOIN's qual clauses */
distribute_quals_to_rels(root, my_quals,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
sjinfo,
root->qual_security_level,
jtitem->qualscope,
ojscope, jtitem->nonnullable_rels,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
NULL, /* incompatible_relids */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
true, /* allow_equivalence */
false, false, /* not clones */
postponed_oj_qual_list);
/* And add the SpecialJoinInfo to join_info_list */
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
if (sjinfo)
root->join_info_list = lappend(root->join_info_list, sjinfo);
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
else
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
{
Error message editing in backend/optimizer, backend/rewrite.
2003-07-25 02:01:09 +02:00
elog(ERROR, "unrecognized node type: %d",
(int) nodeTag(jtnode));
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
}
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
/*
* process_security_barrier_quals
* Transfer security-barrier quals into relation's baserestrictinfo list.
*
* The rewriter put any relevant security-barrier conditions into the RTE's
* securityQuals field, but it's now time to copy them into the rel's
* baserestrictinfo.
*
* In inheritance cases, we only consider quals attached to the parent rel
* here; they will be valid for all children too, so it's okay to consider
* them for purposes like equivalence class creation. Quals attached to
* individual child rels will be dealt with during path creation.
*/
static void
process_security_barrier_quals(PlannerInfo *root,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
int rti, JoinTreeItem *jtitem)
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
{
RangeTblEntry *rte = root->simple_rte_array[rti];
Index security_level = 0;
ListCell *lc;
/*
* Each element of the securityQuals list has been preprocessed into an
* implicitly-ANDed list of clauses. All the clauses in a given sublist
* should get the same security level, but successive sublists get higher
* levels.
*/
foreach(lc, rte->securityQuals)
{
List *qualset = (List *) lfirst(lc);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* We cheat to the extent of passing ojscope = qualscope rather than
* its more logical value of NULL. The only effect this has is to
* force a Var-free qual to be evaluated at the rel rather than being
* pushed up to top of tree, which we don't want.
*/
distribute_quals_to_rels(root, qualset,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
NULL,
security_level,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem->qualscope,
jtitem->qualscope,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
NULL,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
NULL,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
true,
false, false, /* not clones */
NULL);
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
security_level++;
}
/* Assert that qual_security_level is higher than anything we just used */
Assert(security_level <= root->qual_security_level);
}
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
/*
* mark_rels_nulled_by_join
* Fill RelOptInfo.nulling_relids of baserels nulled by this outer join
*
* Inputs:
* ojrelid: RT index of the join RTE (must not be 0)
* lower_rels: the base+OJ Relids syntactically below nullable side of join
*/
static void
mark_rels_nulled_by_join(PlannerInfo *root, Index ojrelid,
Relids lower_rels)
{
int relid = -1;
while ((relid = bms_next_member(lower_rels, relid)) > 0)
{
RelOptInfo *rel = root->simple_rel_array[relid];
if (rel == NULL) /* must be an outer join */
{
Assert(bms_is_member(relid, root->outer_join_rels));
continue;
}
rel->nulling_relids = bms_add_member(rel->nulling_relids, ojrelid);
}
}
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
/*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* make_outerjoininfo
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* Build a SpecialJoinInfo for the current outer join
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
*
* Inputs:
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* left_rels: the base+OJ Relids syntactically on outer side of join
* right_rels: the base+OJ Relids syntactically on inner side of join
* inner_join_rels: base+OJ Relids participating in inner joins below this one
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* jointype: what it says (must always be LEFT, FULL, SEMI, or ANTI)
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* ojrelid: RT index of the join RTE (0 for SEMI, which isn't in the RT list)
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* clause: the outer join's join condition (in implicit-AND format)
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
*
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* The node should eventually be appended to root->join_info_list, but we
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* do not do that here.
Repair bug in 8.2's new logic for planning outer joins: we have to allow joins that overlap an outer join's min_righthand but aren't fully contained in it, to support joining within the RHS after having performed an outer join that can commute with this one. Aside from the direct fix in make_join_rel(), fix has_join_restriction() and GEQO's desirable_join() to consider this possibility. Per report from Ian Harding.
2007-02-13 03:31:03 +01:00
*
* Note: we assume that this function is invoked bottom-up, so that
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
* root->join_info_list already contains entries for all outer joins that are
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
* syntactically below this one.
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*/
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
static SpecialJoinInfo *
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
make_outerjoininfo(PlannerInfo *root,
Relids left_rels, Relids right_rels,
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
Relids inner_join_rels,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
JoinType jointype, Index ojrelid,
List *clause)
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
{
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
SpecialJoinInfo *sjinfo = makeNode(SpecialJoinInfo);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Relids clause_relids;
Relids strict_relids;
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
Relids min_lefthand;
Relids min_righthand;
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
Relids commute_below_l;
Relids commute_below_r;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
ListCell *l;
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
/*
* We should not see RIGHT JOIN here because left/right were switched
* earlier
*/
Assert(jointype != JOIN_INNER);
Assert(jointype != JOIN_RIGHT);
Put back plan-time check for trying to apply SELECT FOR UPDATE/SHARE to a relation on the nullable side of an outer join. I had removed this during the outer join planning rewrite a few months ago ... I think I intended to put it somewhere else, but forgot ...
2006-09-08 19:49:13 +02:00
/*
Improve concurrency of foreign key locking This patch introduces two additional lock modes for tuples: "SELECT FOR KEY SHARE" and "SELECT FOR NO KEY UPDATE". These don't block each other, in contrast with already existing "SELECT FOR SHARE" and "SELECT FOR UPDATE". UPDATE commands that do not modify the values stored in the columns that are part of the key of the tuple now grab a SELECT FOR NO KEY UPDATE lock on the tuple, allowing them to proceed concurrently with tuple locks of the FOR KEY SHARE variety. Foreign key triggers now use FOR KEY SHARE instead of FOR SHARE; this means the concurrency improvement applies to them, which is the whole point of this patch. The added tuple lock semantics require some rejiggering of the multixact module, so that the locking level that each transaction is holding can be stored alongside its Xid. Also, multixacts now need to persist across server restarts and crashes, because they can now represent not only tuple locks, but also tuple updates. This means we need more careful tracking of lifetime of pg_multixact SLRU files; since they now persist longer, we require more infrastructure to figure out when they can be removed. pg_upgrade also needs to be careful to copy pg_multixact files over from the old server to the new, or at least part of multixact.c state, depending on the versions of the old and new servers. Tuple time qualification rules (HeapTupleSatisfies routines) need to be careful not to consider tuples with the "is multi" infomask bit set as being only locked; they might need to look up MultiXact values (i.e. possibly do pg_multixact I/O) to find out the Xid that updated a tuple, whereas they previously were assured to only use information readily available from the tuple header. This is considered acceptable, because the extra I/O would involve cases that would previously cause some commands to block waiting for concurrent transactions to finish. Another important change is the fact that locking tuples that have previously been updated causes the future versions to be marked as locked, too; this is essential for correctness of foreign key checks. This causes additional WAL-logging, also (there was previously a single WAL record for a locked tuple; now there are as many as updated copies of the tuple there exist.) With all this in place, contention related to tuples being checked by foreign key rules should be much reduced. As a bonus, the old behavior that a subtransaction grabbing a stronger tuple lock than the parent (sub)transaction held on a given tuple and later aborting caused the weaker lock to be lost, has been fixed. Many new spec files were added for isolation tester framework, to ensure overall behavior is sane. There's probably room for several more tests. There were several reviewers of this patch; in particular, Noah Misch and Andres Freund spent considerable time in it. Original idea for the patch came from Simon Riggs, after a problem report by Joel Jacobson. Most code is from me, with contributions from Marti Raudsepp, Alexander Shulgin, Noah Misch and Andres Freund. This patch was discussed in several pgsql-hackers threads; the most important start at the following message-ids: AANLkTimo9XVcEzfiBR-ut3KVNDkjm2Vxh+t8kAmWjPuv@mail.gmail.com 1290721684-sup-3951@alvh.no-ip.org 1294953201-sup-2099@alvh.no-ip.org 1320343602-sup-2290@alvh.no-ip.org 1339690386-sup-8927@alvh.no-ip.org 4FE5FF020200002500048A3D@gw.wicourts.gov 4FEAB90A0200002500048B7D@gw.wicourts.gov
2013-01-23 16:04:59 +01:00
* Presently the executor cannot support FOR [KEY] UPDATE/SHARE marking of
Put back plan-time check for trying to apply SELECT FOR UPDATE/SHARE to a relation on the nullable side of an outer join. I had removed this during the outer join planning rewrite a few months ago ... I think I intended to put it somewhere else, but forgot ...
2006-09-08 19:49:13 +02:00
* rels appearing on the nullable side of an outer join. (It's somewhat
* unclear what that would mean, anyway: what should we mark when a result
* row is generated from no element of the nullable relation?) So,
Improve concurrency of foreign key locking This patch introduces two additional lock modes for tuples: "SELECT FOR KEY SHARE" and "SELECT FOR NO KEY UPDATE". These don't block each other, in contrast with already existing "SELECT FOR SHARE" and "SELECT FOR UPDATE". UPDATE commands that do not modify the values stored in the columns that are part of the key of the tuple now grab a SELECT FOR NO KEY UPDATE lock on the tuple, allowing them to proceed concurrently with tuple locks of the FOR KEY SHARE variety. Foreign key triggers now use FOR KEY SHARE instead of FOR SHARE; this means the concurrency improvement applies to them, which is the whole point of this patch. The added tuple lock semantics require some rejiggering of the multixact module, so that the locking level that each transaction is holding can be stored alongside its Xid. Also, multixacts now need to persist across server restarts and crashes, because they can now represent not only tuple locks, but also tuple updates. This means we need more careful tracking of lifetime of pg_multixact SLRU files; since they now persist longer, we require more infrastructure to figure out when they can be removed. pg_upgrade also needs to be careful to copy pg_multixact files over from the old server to the new, or at least part of multixact.c state, depending on the versions of the old and new servers. Tuple time qualification rules (HeapTupleSatisfies routines) need to be careful not to consider tuples with the "is multi" infomask bit set as being only locked; they might need to look up MultiXact values (i.e. possibly do pg_multixact I/O) to find out the Xid that updated a tuple, whereas they previously were assured to only use information readily available from the tuple header. This is considered acceptable, because the extra I/O would involve cases that would previously cause some commands to block waiting for concurrent transactions to finish. Another important change is the fact that locking tuples that have previously been updated causes the future versions to be marked as locked, too; this is essential for correctness of foreign key checks. This causes additional WAL-logging, also (there was previously a single WAL record for a locked tuple; now there are as many as updated copies of the tuple there exist.) With all this in place, contention related to tuples being checked by foreign key rules should be much reduced. As a bonus, the old behavior that a subtransaction grabbing a stronger tuple lock than the parent (sub)transaction held on a given tuple and later aborting caused the weaker lock to be lost, has been fixed. Many new spec files were added for isolation tester framework, to ensure overall behavior is sane. There's probably room for several more tests. There were several reviewers of this patch; in particular, Noah Misch and Andres Freund spent considerable time in it. Original idea for the patch came from Simon Riggs, after a problem report by Joel Jacobson. Most code is from me, with contributions from Marti Raudsepp, Alexander Shulgin, Noah Misch and Andres Freund. This patch was discussed in several pgsql-hackers threads; the most important start at the following message-ids: AANLkTimo9XVcEzfiBR-ut3KVNDkjm2Vxh+t8kAmWjPuv@mail.gmail.com 1290721684-sup-3951@alvh.no-ip.org 1294953201-sup-2099@alvh.no-ip.org 1320343602-sup-2290@alvh.no-ip.org 1339690386-sup-8927@alvh.no-ip.org 4FE5FF020200002500048A3D@gw.wicourts.gov 4FEAB90A0200002500048B7D@gw.wicourts.gov
2013-01-23 16:04:59 +01:00
* complain if any nullable rel is FOR [KEY] UPDATE/SHARE.
Put back plan-time check for trying to apply SELECT FOR UPDATE/SHARE to a relation on the nullable side of an outer join. I had removed this during the outer join planning rewrite a few months ago ... I think I intended to put it somewhere else, but forgot ...
2006-09-08 19:49:13 +02:00
*
* You might be wondering why this test isn't made far upstream in the
* parser. It's because the parser hasn't got enough info --- consider
* FOR UPDATE applied to a view. Only after rewriting and flattening do
* we know whether the view contains an outer join.
Re-implement EvalPlanQual processing to improve its performance and eliminate a lot of strange behaviors that occurred in join cases. We now identify the "current" row for every joined relation in UPDATE, DELETE, and SELECT FOR UPDATE/SHARE queries. If an EvalPlanQual recheck is necessary, we jam the appropriate row into each scan node in the rechecking plan, forcing it to emit only that one row. The former behavior could rescan the whole of each joined relation for each recheck, which was terrible for performance, and what's much worse could result in duplicated output tuples. Also, the original implementation of EvalPlanQual could not re-use the recheck execution tree --- it had to go through a full executor init and shutdown for every row to be tested. To avoid this overhead, I've associated a special runtime Param with each LockRows or ModifyTable plan node, and arranged to make every scan node below such a node depend on that Param. Thus, by signaling a change in that Param, the EPQ machinery can just rescan the already-built test plan. This patch also adds a prohibition on set-returning functions in the targetlist of SELECT FOR UPDATE/SHARE. This is needed to avoid the duplicate-output-tuple problem. It seems fairly reasonable since the other restrictions on SELECT FOR UPDATE are meant to ensure that there is a unique correspondence between source tuples and result tuples, which an output SRF destroys as much as anything else does.
2009-10-26 03:26:45 +01:00
*
* We use the original RowMarkClause list here; the PlanRowMark list would
* list everything.
Put back plan-time check for trying to apply SELECT FOR UPDATE/SHARE to a relation on the nullable side of an outer join. I had removed this during the outer join planning rewrite a few months ago ... I think I intended to put it somewhere else, but forgot ...
2006-09-08 19:49:13 +02:00
*/
foreach(l, root->parse->rowMarks)
{
RowMarkClause *rc = (RowMarkClause *) lfirst(l);
if (bms_is_member(rc->rti, right_rels) ||
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
(jointype == JOIN_FULL && bms_is_member(rc->rti, left_rels)))
Put back plan-time check for trying to apply SELECT FOR UPDATE/SHARE to a relation on the nullable side of an outer join. I had removed this during the outer join planning rewrite a few months ago ... I think I intended to put it somewhere else, but forgot ...
2006-09-08 19:49:13 +02:00
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
Tweak FOR UPDATE/SHARE error message wording (again) In commit 0ac5ad5134 I changed some error messages from "FOR UPDATE/SHARE" to a rather long gobbledygook which nobody liked. Then, in commit cb9b66d31 I changed them again, but the alternative chosen there was deemed suboptimal by Peter Eisentraut, who in message 1373937980.20441.8.camel@vanquo.pezone.net proposed an alternative involving a dynamically-constructed string based on the actual locking strength specified in the SQL command. This patch implements that suggestion.
2013-07-23 20:03:09 +02:00
/*------
translator: %s is a SQL row locking clause such as FOR UPDATE */
errmsg("%s cannot be applied to the nullable side of an outer join",
LCS_asString(rc->strength))));
Put back plan-time check for trying to apply SELECT FOR UPDATE/SHARE to a relation on the nullable side of an outer join. I had removed this during the outer join planning rewrite a few months ago ... I think I intended to put it somewhere else, but forgot ...
2006-09-08 19:49:13 +02:00
}
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
sjinfo->syn_lefthand = left_rels;
sjinfo->syn_righthand = right_rels;
sjinfo->jointype = jointype;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
sjinfo->ojrelid = ojrelid;
/* these fields may get added to later: */
sjinfo->commute_above_l = NULL;
sjinfo->commute_above_r = NULL;
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
sjinfo->commute_below_l = NULL;
sjinfo->commute_below_r = NULL;
Improve planner's cost estimation in the presence of semijoins. If we have a semijoin, say SELECT * FROM x WHERE x1 IN (SELECT y1 FROM y) and we're estimating the cost of a parameterized indexscan on x, the number of repetitions of the indexscan should not be taken as the size of y; it'll really only be the number of distinct values of y1, because the only valid plan with y on the outside of a nestloop would require y to be unique-ified before joining it to x. Most of the time this doesn't make that much difference, but sometimes it can lead to drastically underestimating the cost of the indexscan and hence choosing a bad plan, as pointed out by David Kubečka. Fixing this is a bit difficult because parameterized indexscans are costed out quite early in the planning process, before we have the information that would be needed to call estimate_num_groups() and thereby estimate the number of distinct values of the join column(s). However we can move the code that extracts a semijoin RHS's unique-ification columns, so that it's done in initsplan.c rather than on-the-fly in create_unique_path(). That shouldn't make any difference speed-wise and it's really a bit cleaner too. The other bit of information we need is the size of the semijoin RHS, which is easy if it's a single relation (we make those estimates before considering indexscan costs) but problematic if it's a join relation. The solution adopted here is just to use the product of the sizes of the join component rels. That will generally be an overestimate, but since estimate_num_groups() only uses this input as a clamp, an overestimate shouldn't hurt us too badly. In any case we don't allow this new logic to produce a value larger than we would have chosen before, so that at worst an overestimate leaves us no wiser than we were before.
2015-03-12 02:21:00 +01:00
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
compute_semijoin_info(root, sjinfo, clause);
Repair planner bug introduced in 8.2 by ability to rearrange outer joins: in cases where a sub-SELECT inserts a WHERE clause between two outer joins, that clause may prevent us from re-ordering the two outer joins. The code was considering only the joins' own ON-conditions in determining reordering safety, which is not good enough. Add a "delay_upper_joins" flag to OuterJoinInfo to flag that we have detected such a clause and higher-level outer joins shouldn't be permitted to commute with this one. (This might seem overly coarse, but given the current rules for OJ reordering, it's sufficient AFAICT.) The failure case is actually pretty narrow: it needs a WHERE clause within the RHS of a left join that checks the RHS of a lower left join, but is not strict for that RHS (else we'd have simplified the lower join to a plain join). Even then no failure will be manifest unless the planner chooses to rearrange the join order. Per bug report from Adam Terrey.
2007-05-23 01:23:58 +02:00
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/* If it's a full join, no need to be very smart */
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
if (jointype == JOIN_FULL)
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
{
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
sjinfo->min_lefthand = bms_copy(left_rels);
sjinfo->min_righthand = bms_copy(right_rels);
sjinfo->lhs_strict = false; /* don't care about this */
return sjinfo;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
}
/*
* Retrieve all relids mentioned within the join clause.
*/
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
clause_relids = pull_varnos(root, (Node *) clause);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/*
* For which relids is the clause strict, ie, it cannot succeed if the
* rel's columns are all NULL?
*/
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
strict_relids = find_nonnullable_rels((Node *) clause);
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/* Remember whether the clause is strict for any LHS relations */
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
sjinfo->lhs_strict = bms_overlap(strict_relids, left_rels);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/*
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
* Required LHS always includes the LHS rels mentioned in the clause. We
* may have to add more rels based on lower outer joins; see below.
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
*/
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
min_lefthand = bms_intersect(clause_relids, left_rels);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/*
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
* Similarly for required RHS. But here, we must also include any lower
* inner joins, to ensure we don't try to commute with any of them.
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
*/
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
min_righthand = bms_int_members(bms_union(clause_relids, inner_join_rels),
right_rels);
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Fix some planner issues with degenerate outer join clauses. An outer join clause that didn't actually reference the RHS (perhaps only after constant-folding) could confuse the join order enforcement logic, leading to wrong query results. Also, nested occurrences of such things could trigger an Assertion that on reflection seems incorrect. Per fuzz testing by Andreas Seltenreich. The practical use of such cases seems thin enough that it's not too surprising we've not heard field reports about it. This has been broken for a long time, so back-patch to all active branches.
2015-08-02 02:57:41 +02:00
/*
* Now check previous outer joins for ordering restrictions.
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
*
* commute_below_l and commute_below_r accumulate the relids of lower
* outer joins that we think this one can commute with. These decisions
* are just tentative within this loop, since we might find an
* intermediate outer join that prevents commutation. Surviving relids
* will get merged into the SpecialJoinInfo structs afterwards.
Fix some planner issues with degenerate outer join clauses. An outer join clause that didn't actually reference the RHS (perhaps only after constant-folding) could confuse the join order enforcement logic, leading to wrong query results. Also, nested occurrences of such things could trigger an Assertion that on reflection seems incorrect. Per fuzz testing by Andreas Seltenreich. The practical use of such cases seems thin enough that it's not too surprising we've not heard field reports about it. This has been broken for a long time, so back-patch to all active branches.
2015-08-02 02:57:41 +02:00
*/
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
commute_below_l = commute_below_r = NULL;
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
foreach(l, root->join_info_list)
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
{
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
SpecialJoinInfo *otherinfo = (SpecialJoinInfo *) lfirst(l);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bool have_unsafe_phvs;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Fix planner failure with full join in RHS of left join. Given a left join containing a full join in its righthand side, with the left join's joinclause referencing only one side of the full join (in a non-strict fashion, so that the full join doesn't get simplified), the planner could fail with "failed to build any N-way joins" or related errors. This happened because the full join was seen as overlapping the left join's RHS, and then recent changes within join_is_legal() caused that function to conclude that the full join couldn't validly be formed. Rather than try to rejigger join_is_legal() yet more to allow this, I think it's better to fix initsplan.c so that the required join order is explicit in the SpecialJoinInfo data structure. The previous coding there essentially ignored full joins, relying on the fact that we don't flatten them in the joinlist data structure to preserve their ordering. That's sufficient to prevent a wrong plan from being formed, but as this example shows, it's not sufficient to ensure that the right plan will be formed. We need to work a bit harder to ensure that the right plan looks sane according to the SpecialJoinInfos. Per bug #14105 from Vojtech Rylko. This was apparently induced by commit 8703059c6 (though now that I've seen it, I wonder whether there are related cases that could have failed before that); so back-patch to all active branches. Unfortunately, that patch also went into 9.0, so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
/*
* A full join is an optimization barrier: we can't associate into or
* out of it. Hence, if it overlaps either LHS or RHS of the current
* rel, expand that side's min relset to cover the whole full join.
*/
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
if (otherinfo->jointype == JOIN_FULL)
Fix planner failure with full join in RHS of left join. Given a left join containing a full join in its righthand side, with the left join's joinclause referencing only one side of the full join (in a non-strict fashion, so that the full join doesn't get simplified), the planner could fail with "failed to build any N-way joins" or related errors. This happened because the full join was seen as overlapping the left join's RHS, and then recent changes within join_is_legal() caused that function to conclude that the full join couldn't validly be formed. Rather than try to rejigger join_is_legal() yet more to allow this, I think it's better to fix initsplan.c so that the required join order is explicit in the SpecialJoinInfo data structure. The previous coding there essentially ignored full joins, relying on the fact that we don't flatten them in the joinlist data structure to preserve their ordering. That's sufficient to prevent a wrong plan from being formed, but as this example shows, it's not sufficient to ensure that the right plan will be formed. We need to work a bit harder to ensure that the right plan looks sane according to the SpecialJoinInfos. Per bug #14105 from Vojtech Rylko. This was apparently induced by commit 8703059c6 (though now that I've seen it, I wonder whether there are related cases that could have failed before that); so back-patch to all active branches. Unfortunately, that patch also went into 9.0, so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
{
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
Assert(otherinfo->ojrelid != 0);
Fix planner failure with full join in RHS of left join. Given a left join containing a full join in its righthand side, with the left join's joinclause referencing only one side of the full join (in a non-strict fashion, so that the full join doesn't get simplified), the planner could fail with "failed to build any N-way joins" or related errors. This happened because the full join was seen as overlapping the left join's RHS, and then recent changes within join_is_legal() caused that function to conclude that the full join couldn't validly be formed. Rather than try to rejigger join_is_legal() yet more to allow this, I think it's better to fix initsplan.c so that the required join order is explicit in the SpecialJoinInfo data structure. The previous coding there essentially ignored full joins, relying on the fact that we don't flatten them in the joinlist data structure to preserve their ordering. That's sufficient to prevent a wrong plan from being formed, but as this example shows, it's not sufficient to ensure that the right plan will be formed. We need to work a bit harder to ensure that the right plan looks sane according to the SpecialJoinInfos. Per bug #14105 from Vojtech Rylko. This was apparently induced by commit 8703059c6 (though now that I've seen it, I wonder whether there are related cases that could have failed before that); so back-patch to all active branches. Unfortunately, that patch also went into 9.0, so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
if (bms_overlap(left_rels, otherinfo->syn_lefthand) ||
bms_overlap(left_rels, otherinfo->syn_righthand))
{
min_lefthand = bms_add_members(min_lefthand,
otherinfo->syn_lefthand);
min_lefthand = bms_add_members(min_lefthand,
otherinfo->syn_righthand);
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
min_lefthand = bms_add_member(min_lefthand,
otherinfo->ojrelid);
Fix planner failure with full join in RHS of left join. Given a left join containing a full join in its righthand side, with the left join's joinclause referencing only one side of the full join (in a non-strict fashion, so that the full join doesn't get simplified), the planner could fail with "failed to build any N-way joins" or related errors. This happened because the full join was seen as overlapping the left join's RHS, and then recent changes within join_is_legal() caused that function to conclude that the full join couldn't validly be formed. Rather than try to rejigger join_is_legal() yet more to allow this, I think it's better to fix initsplan.c so that the required join order is explicit in the SpecialJoinInfo data structure. The previous coding there essentially ignored full joins, relying on the fact that we don't flatten them in the joinlist data structure to preserve their ordering. That's sufficient to prevent a wrong plan from being formed, but as this example shows, it's not sufficient to ensure that the right plan will be formed. We need to work a bit harder to ensure that the right plan looks sane according to the SpecialJoinInfos. Per bug #14105 from Vojtech Rylko. This was apparently induced by commit 8703059c6 (though now that I've seen it, I wonder whether there are related cases that could have failed before that); so back-patch to all active branches. Unfortunately, that patch also went into 9.0, so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
}
if (bms_overlap(right_rels, otherinfo->syn_lefthand) ||
bms_overlap(right_rels, otherinfo->syn_righthand))
{
min_righthand = bms_add_members(min_righthand,
otherinfo->syn_lefthand);
min_righthand = bms_add_members(min_righthand,
otherinfo->syn_righthand);
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
min_righthand = bms_add_member(min_righthand,
otherinfo->ojrelid);
Fix planner failure with full join in RHS of left join. Given a left join containing a full join in its righthand side, with the left join's joinclause referencing only one side of the full join (in a non-strict fashion, so that the full join doesn't get simplified), the planner could fail with "failed to build any N-way joins" or related errors. This happened because the full join was seen as overlapping the left join's RHS, and then recent changes within join_is_legal() caused that function to conclude that the full join couldn't validly be formed. Rather than try to rejigger join_is_legal() yet more to allow this, I think it's better to fix initsplan.c so that the required join order is explicit in the SpecialJoinInfo data structure. The previous coding there essentially ignored full joins, relying on the fact that we don't flatten them in the joinlist data structure to preserve their ordering. That's sufficient to prevent a wrong plan from being formed, but as this example shows, it's not sufficient to ensure that the right plan will be formed. We need to work a bit harder to ensure that the right plan looks sane according to the SpecialJoinInfos. Per bug #14105 from Vojtech Rylko. This was apparently induced by commit 8703059c6 (though now that I've seen it, I wonder whether there are related cases that could have failed before that); so back-patch to all active branches. Unfortunately, that patch also went into 9.0, so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
}
/* Needn't do anything else with the full join */
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
continue;
Fix planner failure with full join in RHS of left join. Given a left join containing a full join in its righthand side, with the left join's joinclause referencing only one side of the full join (in a non-strict fashion, so that the full join doesn't get simplified), the planner could fail with "failed to build any N-way joins" or related errors. This happened because the full join was seen as overlapping the left join's RHS, and then recent changes within join_is_legal() caused that function to conclude that the full join couldn't validly be formed. Rather than try to rejigger join_is_legal() yet more to allow this, I think it's better to fix initsplan.c so that the required join order is explicit in the SpecialJoinInfo data structure. The previous coding there essentially ignored full joins, relying on the fact that we don't flatten them in the joinlist data structure to preserve their ordering. That's sufficient to prevent a wrong plan from being formed, but as this example shows, it's not sufficient to ensure that the right plan will be formed. We need to work a bit harder to ensure that the right plan looks sane according to the SpecialJoinInfos. Per bug #14105 from Vojtech Rylko. This was apparently induced by commit 8703059c6 (though now that I've seen it, I wonder whether there are related cases that could have failed before that); so back-patch to all active branches. Unfortunately, that patch also went into 9.0, so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
}
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* If our join condition contains any PlaceHolderVars that need to be
* evaluated above the lower OJ, then we can't commute with it.
*/
if (otherinfo->ojrelid != 0)
have_unsafe_phvs =
contain_placeholder_references_to(root,
(Node *) clause,
otherinfo->ojrelid);
else
have_unsafe_phvs = false;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
/*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* For a lower OJ in our LHS, if our join condition uses the lower
* join's RHS and is not strict for that rel, we must preserve the
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
* ordering of the two OJs, so add lower OJ's full syntactic relset to
* min_lefthand. (We must use its full syntactic relset, not just its
* min_lefthand + min_righthand. This is because there might be other
* OJs below this one that this one can commute with, but we cannot
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* commute with them if we don't with this one.) Also, if we have
* unsafe PHVs or the current join is a semijoin or antijoin, we must
* preserve ordering regardless of strictness.
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
*
* Note: I believe we have to insist on being strict for at least one
* rel in the lower OJ's min_righthand, not its whole syn_righthand.
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
*
* When we don't need to preserve ordering, check to see if outer join
* identity 3 applies, and if so, remove the lower OJ's ojrelid from
* our min_lefthand so that commutation is allowed.
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*/
Tighten up join ordering rules to account for recent more-careful analysis of the associativity of antijoins. Also improve optimizer/README discussion of outer join ordering rules.
2009-02-27 23:41:38 +01:00
if (bms_overlap(left_rels, otherinfo->syn_righthand))
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
{
Tighten up join ordering rules to account for recent more-careful analysis of the associativity of antijoins. Also improve optimizer/README discussion of outer join ordering rules.
2009-02-27 23:41:38 +01:00
if (bms_overlap(clause_relids, otherinfo->syn_righthand) &&
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
(have_unsafe_phvs ||
jointype == JOIN_SEMI || jointype == JOIN_ANTI ||
Tighten up join ordering rules to account for recent more-careful analysis of the associativity of antijoins. Also improve optimizer/README discussion of outer join ordering rules.
2009-02-27 23:41:38 +01:00
!bms_overlap(strict_relids, otherinfo->min_righthand)))
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Preserve ordering */
Tighten up join ordering rules to account for recent more-careful analysis of the associativity of antijoins. Also improve optimizer/README discussion of outer join ordering rules.
2009-02-27 23:41:38 +01:00
min_lefthand = bms_add_members(min_lefthand,
otherinfo->syn_lefthand);
min_lefthand = bms_add_members(min_lefthand,
otherinfo->syn_righthand);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
if (otherinfo->ojrelid != 0)
min_lefthand = bms_add_member(min_lefthand,
otherinfo->ojrelid);
}
else if (jointype == JOIN_LEFT &&
otherinfo->jointype == JOIN_LEFT &&
Further fixes in qual nullingrel adjustment for outer join commutation. One of the add_nulling_relids calls in deconstruct_distribute_oj_quals added an OJ relid to too few Vars, while the other added it to too many. We should consider the syntactic structure not min_left/righthand while deciding which Vars to decorate, and when considering pushing up a lower outer join pursuant to transforming the second form of OJ identity 3 to the first form, we only want to decorate Vars coming from its LHS. In a related bug, I realized that make_outerjoininfo was failing to check a very basic property that's needed to apply OJ identity 3: the syntactically-upper outer join clause can't refer to the lower join's LHS. This didn't break the join order restriction logic, but it led to setting bogus commute_xxx bits, possibly resulting in bogus nullingrel markings in modified quals. Richard Guo and Tom Lane Discussion: https://postgr.es/m/CAMbWs497CmBruMx1SOjepWEz+T5NWa4scqbdE9v7ZzSXqH_gQw@mail.gmail.com Discussion: https://postgr.es/m/CAEP4nAx9C5gXNBfEA0JBfz7B+5f1Bawt-RWQWyhev-wdps8BZA@mail.gmail.com
2023-02-10 19:31:00 +01:00
bms_overlap(strict_relids, otherinfo->min_righthand) &&
!bms_overlap(clause_relids, otherinfo->syn_lefthand))
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
{
/* Identity 3 applies, so remove the ordering restriction */
min_lefthand = bms_del_member(min_lefthand, otherinfo->ojrelid);
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
/* Record the (still tentative) commutability relationship */
commute_below_l =
bms_add_member(commute_below_l, otherinfo->ojrelid);
Tighten up join ordering rules to account for recent more-careful analysis of the associativity of antijoins. Also improve optimizer/README discussion of outer join ordering rules.
2009-02-27 23:41:38 +01:00
}
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
}
pgindent run for 8.2.
2006-10-04 02:30:14 +02:00
Current implementation of FOR UPDATE has no hope of working correctly for relations on the nullable side of an OUTER JOIN. For now I think we'd better refuse such queries.
2001-05-14 22:25:00 +02:00
/*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* For a lower OJ in our RHS, if our join condition does not use the
* lower join's RHS and the lower OJ's join condition is strict, we
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
* can interchange the ordering of the two OJs; otherwise we must add
Further fixes for degenerate outer join clauses. Further testing revealed that commit f69b4b9495269cc4 was still a few bricks shy of a load: minor tweaking of the previous test cases resulted in the same wrong-outer-join-order problem coming back. After study I concluded that my previous changes in make_outerjoininfo() were just accidentally masking the problem, and should be reverted in favor of forcing syntactic join order whenever an upper outer join's predicate doesn't mention a lower outer join's LHS. This still allows the chained-outer-joins style that is the normally optimizable case. I also tightened things up some more in join_is_legal(). It seems to me on review that what's really happening in the exception case where we ignore a mismatched special join is that we're allowing the proposed join to associate into the RHS of the outer join we're comparing it to. As such, we should *always* insist that the proposed join be a left join, which eliminates a bunch of rather dubious argumentation. The case where we weren't enforcing that was the one that was already known buggy anyway (it had a violatable Assert before the aforesaid commit) so it hardly deserves a lot of deference. Back-patch to all active branches, like the previous patch. The added regression test case failed in all branches back to 9.1, and I think it's only an unrelated change in costing calculations that kept 9.0 from choosing a broken plan.
2015-08-06 21:35:27 +02:00
* the lower OJ's full syntactic relset to min_righthand.
*
* Also, if our join condition does not use the lower join's LHS
* either, force the ordering to be preserved. Otherwise we can end
* up with SpecialJoinInfos with identical min_righthands, which can
* confuse join_is_legal (see discussion in backend/optimizer/README).
*
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* Also, we must preserve ordering anyway if we have unsafe PHVs, or
* if either this join or the lower OJ is a semijoin or antijoin.
Repair planner bug introduced in 8.2 by ability to rearrange outer joins: in cases where a sub-SELECT inserts a WHERE clause between two outer joins, that clause may prevent us from re-ordering the two outer joins. The code was considering only the joins' own ON-conditions in determining reordering safety, which is not good enough. Add a "delay_upper_joins" flag to OuterJoinInfo to flag that we have detected such a clause and higher-level outer joins shouldn't be permitted to commute with this one. (This might seem overly coarse, but given the current rules for OJ reordering, it's sufficient AFAICT.) The failure case is actually pretty narrow: it needs a WHERE clause within the RHS of a left join that checks the RHS of a lower left join, but is not strict for that RHS (else we'd have simplified the lower join to a plain join). Even then no failure will be manifest unless the planner chooses to rearrange the join order. Per bug report from Adam Terrey.
2007-05-23 01:23:58 +02:00
*
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* When we don't need to preserve ordering, check to see if outer join
* identity 3 applies, and if so, remove the lower OJ's ojrelid from
* our min_righthand so that commutation is allowed.
Current implementation of FOR UPDATE has no hope of working correctly for relations on the nullable side of an OUTER JOIN. For now I think we'd better refuse such queries.
2001-05-14 22:25:00 +02:00
*/
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
if (bms_overlap(right_rels, otherinfo->syn_righthand))
Current implementation of FOR UPDATE has no hope of working correctly for relations on the nullable side of an OUTER JOIN. For now I think we'd better refuse such queries.
2001-05-14 22:25:00 +02:00
{
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
if (bms_overlap(clause_relids, otherinfo->syn_righthand) ||
Further fixes for degenerate outer join clauses. Further testing revealed that commit f69b4b9495269cc4 was still a few bricks shy of a load: minor tweaking of the previous test cases resulted in the same wrong-outer-join-order problem coming back. After study I concluded that my previous changes in make_outerjoininfo() were just accidentally masking the problem, and should be reverted in favor of forcing syntactic join order whenever an upper outer join's predicate doesn't mention a lower outer join's LHS. This still allows the chained-outer-joins style that is the normally optimizable case. I also tightened things up some more in join_is_legal(). It seems to me on review that what's really happening in the exception case where we ignore a mismatched special join is that we're allowing the proposed join to associate into the RHS of the outer join we're comparing it to. As such, we should *always* insist that the proposed join be a left join, which eliminates a bunch of rather dubious argumentation. The case where we weren't enforcing that was the one that was already known buggy anyway (it had a violatable Assert before the aforesaid commit) so it hardly deserves a lot of deference. Back-patch to all active branches, like the previous patch. The added regression test case failed in all branches back to 9.1, and I think it's only an unrelated change in costing calculations that kept 9.0 from choosing a broken plan.
2015-08-06 21:35:27 +02:00
!bms_overlap(clause_relids, otherinfo->min_lefthand) ||
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
have_unsafe_phvs ||
Ooops ... make_outerjoininfo wasn't actually enforcing the join order restrictions specified for semijoins in optimizer/README, to wit that you can't reassociate outer joins into or out of the RHS of a semijoin. Per report from Heikki.
2009-05-07 22:13:09 +02:00
jointype == JOIN_SEMI ||
Prevent improper reordering of antijoins vs. outer joins. An outer join appearing within the RHS of an antijoin can't commute with the antijoin, but somehow I missed teaching make_outerjoininfo() about that. In Teodor Sigaev's recent trouble report, this manifests as a "could not find RelOptInfo for given relids" error within eqjoinsel(); but I think silently wrong query results are possible too, if the planner misorders the joins and doesn't happen to trigger any internal consistency checks. It's broken as far back as we had antijoins, so back-patch to all supported branches.
2015-04-25 22:44:27 +02:00
jointype == JOIN_ANTI ||
Fix another semijoin-ordering bug. We already knew that we couldn't reorder a semijoin into or out of the righthand side of another semijoin, but actually it doesn't work to reorder it into or out of the righthand side of a left or antijoin, either. Per bug #4906 from Mathieu Fenniak. This was sloppy thinking on my part. This identity does work: ( A left join B on (Pab) ) semijoin C on (Pac) == ( A semijoin C on (Pac) ) left join B on (Pab) but I failed to see that that doesn't mean this does: ( A left join B on (Pab) ) semijoin C on (Pbc) != A left join ( B semijoin C on (Pbc) ) on (Pab)
2009-07-21 04:02:44 +02:00
otherinfo->jointype == JOIN_SEMI ||
Tighten up join ordering rules to account for recent more-careful analysis of the associativity of antijoins. Also improve optimizer/README discussion of outer join ordering rules.
2009-02-27 23:41:38 +01:00
otherinfo->jointype == JOIN_ANTI ||
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
!otherinfo->lhs_strict)
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Preserve ordering */
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
min_righthand = bms_add_members(min_righthand,
otherinfo->syn_lefthand);
min_righthand = bms_add_members(min_righthand,
otherinfo->syn_righthand);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
if (otherinfo->ojrelid != 0)
min_righthand = bms_add_member(min_righthand,
otherinfo->ojrelid);
}
else if (jointype == JOIN_LEFT &&
otherinfo->jointype == JOIN_LEFT &&
otherinfo->lhs_strict)
{
/* Identity 3 applies, so remove the ordering restriction */
min_righthand = bms_del_member(min_righthand,
otherinfo->ojrelid);
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
/* Record the (still tentative) commutability relationship */
commute_below_r =
bms_add_member(commute_below_r, otherinfo->ojrelid);
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
}
Current implementation of FOR UPDATE has no hope of working correctly for relations on the nullable side of an OUTER JOIN. For now I think we'd better refuse such queries.
2001-05-14 22:25:00 +02:00
}
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Fix PlaceHolderVar mechanism's interaction with outer joins. The point of a PlaceHolderVar is to allow a non-strict expression to be evaluated below an outer join, after which its value bubbles up like a Var and can be forced to NULL when the outer join's semantics require that. However, there was a serious design oversight in that, namely that we didn't ensure that there was actually a correct place in the plan tree to evaluate the placeholder :-(. It may be necessary to delay evaluation of an outer join to ensure that a placeholder that should be evaluated below the join can be evaluated there. Per recent bug report from Kirill Simonov. Back-patch to 8.4 where the PlaceHolderVar mechanism was introduced.
2010-09-28 18:08:56 +02:00
/*
* Examine PlaceHolderVars. If a PHV is supposed to be evaluated within
Remove ph_may_need from PlaceHolderInfo, with attendant simplifications. The planner logic that attempted to make a preliminary estimate of the ph_needed levels for PlaceHolderVars seems to be completely broken by lateral references. Fortunately, the potential join order optimization that this code supported seems to be of relatively little value in practice; so let's just get rid of it rather than trying to fix it. Getting rid of this allows fairly substantial simplifications in placeholder.c, too, so planning in such cases should be a bit faster. Issue noted while pursuing bugs reported by Jeremy Evans and Antonin Houska, though this doesn't in itself fix either of their reported cases. What this does do is prevent an Assert crash in the kind of query illustrated by the added regression test. (I'm not sure that the plan for that query is stable enough across platforms to be usable as a regression test output ... but we'll soon find out from the buildfarm.) Back-patch to 9.3. The problem case can't arise without LATERAL, so no need to touch older branches.
2013-08-15 00:38:32 +02:00
* this join's nullable side, then ensure that min_righthand contains the
* full eval_at set of the PHV. This ensures that the PHV actually can be
* evaluated within the RHS. Note that this works only because we should
* already have determined the final eval_at level for any PHV
* syntactically within this join.
Fix PlaceHolderVar mechanism's interaction with outer joins. The point of a PlaceHolderVar is to allow a non-strict expression to be evaluated below an outer join, after which its value bubbles up like a Var and can be forced to NULL when the outer join's semantics require that. However, there was a serious design oversight in that, namely that we didn't ensure that there was actually a correct place in the plan tree to evaluate the placeholder :-(. It may be necessary to delay evaluation of an outer join to ensure that a placeholder that should be evaluated below the join can be evaluated there. Per recent bug report from Kirill Simonov. Back-patch to 8.4 where the PlaceHolderVar mechanism was introduced.
2010-09-28 18:08:56 +02:00
*/
foreach(l, root->placeholder_list)
{
PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l);
Relids ph_syn_level = phinfo->ph_var->phrels;
/* Ignore placeholder if it didn't syntactically come from RHS */
if (!bms_is_subset(ph_syn_level, right_rels))
continue;
/* Else, prevent join from being formed before we eval the PHV */
min_righthand = bms_add_members(min_righthand, phinfo->ph_eval_at);
}
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
/*
* If we found nothing to put in min_lefthand, punt and make it the full
* LHS, to avoid having an empty min_lefthand which will confuse later
* processing. (We don't try to be smart about such cases, just correct.)
Further fixes for degenerate outer join clauses. Further testing revealed that commit f69b4b9495269cc4 was still a few bricks shy of a load: minor tweaking of the previous test cases resulted in the same wrong-outer-join-order problem coming back. After study I concluded that my previous changes in make_outerjoininfo() were just accidentally masking the problem, and should be reverted in favor of forcing syntactic join order whenever an upper outer join's predicate doesn't mention a lower outer join's LHS. This still allows the chained-outer-joins style that is the normally optimizable case. I also tightened things up some more in join_is_legal(). It seems to me on review that what's really happening in the exception case where we ignore a mismatched special join is that we're allowing the proposed join to associate into the RHS of the outer join we're comparing it to. As such, we should *always* insist that the proposed join be a left join, which eliminates a bunch of rather dubious argumentation. The case where we weren't enforcing that was the one that was already known buggy anyway (it had a violatable Assert before the aforesaid commit) so it hardly deserves a lot of deference. Back-patch to all active branches, like the previous patch. The added regression test case failed in all branches back to 9.1, and I think it's only an unrelated change in costing calculations that kept 9.0 from choosing a broken plan.
2015-08-06 21:35:27 +02:00
* Likewise for min_righthand.
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
*/
if (bms_is_empty(min_lefthand))
min_lefthand = bms_copy(left_rels);
Further fixes for degenerate outer join clauses. Further testing revealed that commit f69b4b9495269cc4 was still a few bricks shy of a load: minor tweaking of the previous test cases resulted in the same wrong-outer-join-order problem coming back. After study I concluded that my previous changes in make_outerjoininfo() were just accidentally masking the problem, and should be reverted in favor of forcing syntactic join order whenever an upper outer join's predicate doesn't mention a lower outer join's LHS. This still allows the chained-outer-joins style that is the normally optimizable case. I also tightened things up some more in join_is_legal(). It seems to me on review that what's really happening in the exception case where we ignore a mismatched special join is that we're allowing the proposed join to associate into the RHS of the outer join we're comparing it to. As such, we should *always* insist that the proposed join be a left join, which eliminates a bunch of rather dubious argumentation. The case where we weren't enforcing that was the one that was already known buggy anyway (it had a violatable Assert before the aforesaid commit) so it hardly deserves a lot of deference. Back-patch to all active branches, like the previous patch. The added regression test case failed in all branches back to 9.1, and I think it's only an unrelated change in costing calculations that kept 9.0 from choosing a broken plan.
2015-08-06 21:35:27 +02:00
if (bms_is_empty(min_righthand))
min_righthand = bms_copy(right_rels);
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
/* Now they'd better be nonempty */
Assert(!bms_is_empty(min_lefthand));
Assert(!bms_is_empty(min_righthand));
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/* Shouldn't overlap either */
Rewrite make_outerjoininfo's construction of min_lefthand and min_righthand sets for outer joins, in the light of bug #3588 and additional thought and experimentation. The original methodology was fatally flawed for nests of more than two outer joins: it got the relationships between adjacent joins right, but didn't always come to the right conclusions about whether a join could be interchanged with one two or more levels below it. This was largely caused by a mistaken idea that we should use the min_lefthand + min_righthand sets of a sub-join as the minimum left or right input set of an upper join when we conclude that the sub-join can't commute with the upper one. If there's a still-lower join that the sub-join *can* commute with, this method led us to think that that one could commute with the topmost join; which it can't. Another problem (not directly connected to bug #3588) was that make_outerjoininfo's processing-order-dependent method for enforcing outer join identity #3 didn't work right: if we decided that join A could safely commute with lower join B, we dropped all information about sub-joins under B that join A could perhaps not safely commute with, because we removed B's entire min_righthand from A's. To fix, make an explicit computation of all inner join combinations that occur below an outer join, and add to that the full syntactic relsets of any lower outer joins that we determine it can't commute with. This method gives much more direct enforcement of the outer join rearrangement identities, and it turns out not to cost a lot of additional bookkeeping. Thanks to Richard Harris for the bug report and test case.
2007-08-31 03:44:06 +02:00
Assert(!bms_overlap(min_lefthand, min_righthand));
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
sjinfo->min_lefthand = min_lefthand;
sjinfo->min_righthand = min_righthand;
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
/*
* Now that we've identified the correct min_lefthand and min_righthand,
* any commute_below_l or commute_below_r relids that have not gotten
* added back into those sets (due to intervening outer joins) are indeed
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
* commutable with this one.
*
* First, delete any subsequently-added-back relids (this is easier than
* maintaining commute_below_l/r precisely through all the above).
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
*/
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
commute_below_l = bms_del_members(commute_below_l, min_lefthand);
commute_below_r = bms_del_members(commute_below_r, min_righthand);
/* Anything left? */
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
if (commute_below_l || commute_below_r)
{
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
/* Yup, so we must update the derived data in the SpecialJoinInfos */
sjinfo->commute_below_l = commute_below_l;
sjinfo->commute_below_r = commute_below_r;
foreach(l, root->join_info_list)
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
{
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
SpecialJoinInfo *otherinfo = (SpecialJoinInfo *) lfirst(l);
if (bms_is_member(otherinfo->ojrelid, commute_below_l))
otherinfo->commute_above_l =
bms_add_member(otherinfo->commute_above_l, ojrelid);
else if (bms_is_member(otherinfo->ojrelid, commute_below_r))
otherinfo->commute_above_r =
bms_add_member(otherinfo->commute_above_r, ojrelid);
Fix over-optimistic updating of info about commutable outer joins. make_outerjoininfo was set up to update SpecialJoinInfo's commute_below, commute_above_l, commute_above_r fields as soon as it found a pair of outer joins that look like they can commute. However, this decision could be negated later in the same loop due to finding an intermediate outer join that prevents commutation. That left us with commute_xxx fields that were contradictory to the join order restrictions expressed in min_lefthand/min_righthand. The latter fields would keep us from actually choosing a bad join order; but the inconsistent commute_xxx fields could bollix details such as the varnullingrels values created for intermediate join relation targetlists, ending in an assertion failure in setrefs.c. To fix, wait till the end of make_outerjoininfo where we have accurate values for min_lefthand/min_righthand, and then insert only relids not present in those sets into the commute_xxx fields. Per SQLSmith testing by Robins Tharakan. Note that while Robins bisected the failure to commit b448f1c8d, it's really the fault of 2489d76c4. The outerjoin_delayed logic removed in the later commit was keeping us from deciding that troublesome join pairs commute, at least in the specific example seen here. Discussion: https://postgr.es/m/CAEP4nAyAORgE8K_RHSmvWbE9UaChhjbEL1RrDU3neePwwRUB=A@mail.gmail.com
2023-02-05 20:25:10 +01:00
}
}
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
return sjinfo;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
Improve planner's cost estimation in the presence of semijoins. If we have a semijoin, say SELECT * FROM x WHERE x1 IN (SELECT y1 FROM y) and we're estimating the cost of a parameterized indexscan on x, the number of repetitions of the indexscan should not be taken as the size of y; it'll really only be the number of distinct values of y1, because the only valid plan with y on the outside of a nestloop would require y to be unique-ified before joining it to x. Most of the time this doesn't make that much difference, but sometimes it can lead to drastically underestimating the cost of the indexscan and hence choosing a bad plan, as pointed out by David Kubečka. Fixing this is a bit difficult because parameterized indexscans are costed out quite early in the planning process, before we have the information that would be needed to call estimate_num_groups() and thereby estimate the number of distinct values of the join column(s). However we can move the code that extracts a semijoin RHS's unique-ification columns, so that it's done in initsplan.c rather than on-the-fly in create_unique_path(). That shouldn't make any difference speed-wise and it's really a bit cleaner too. The other bit of information we need is the size of the semijoin RHS, which is easy if it's a single relation (we make those estimates before considering indexscan costs) but problematic if it's a join relation. The solution adopted here is just to use the product of the sizes of the join component rels. That will generally be an overestimate, but since estimate_num_groups() only uses this input as a clamp, an overestimate shouldn't hurt us too badly. In any case we don't allow this new logic to produce a value larger than we would have chosen before, so that at worst an overestimate leaves us no wiser than we were before.
2015-03-12 02:21:00 +01:00
/*
* compute_semijoin_info
* Fill semijoin-related fields of a new SpecialJoinInfo
*
* Note: this relies on only the jointype and syn_righthand fields of the
* SpecialJoinInfo; the rest may not be set yet.
*/
static void
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
compute_semijoin_info(PlannerInfo *root, SpecialJoinInfo *sjinfo, List *clause)
Improve planner's cost estimation in the presence of semijoins. If we have a semijoin, say SELECT * FROM x WHERE x1 IN (SELECT y1 FROM y) and we're estimating the cost of a parameterized indexscan on x, the number of repetitions of the indexscan should not be taken as the size of y; it'll really only be the number of distinct values of y1, because the only valid plan with y on the outside of a nestloop would require y to be unique-ified before joining it to x. Most of the time this doesn't make that much difference, but sometimes it can lead to drastically underestimating the cost of the indexscan and hence choosing a bad plan, as pointed out by David Kubečka. Fixing this is a bit difficult because parameterized indexscans are costed out quite early in the planning process, before we have the information that would be needed to call estimate_num_groups() and thereby estimate the number of distinct values of the join column(s). However we can move the code that extracts a semijoin RHS's unique-ification columns, so that it's done in initsplan.c rather than on-the-fly in create_unique_path(). That shouldn't make any difference speed-wise and it's really a bit cleaner too. The other bit of information we need is the size of the semijoin RHS, which is easy if it's a single relation (we make those estimates before considering indexscan costs) but problematic if it's a join relation. The solution adopted here is just to use the product of the sizes of the join component rels. That will generally be an overestimate, but since estimate_num_groups() only uses this input as a clamp, an overestimate shouldn't hurt us too badly. In any case we don't allow this new logic to produce a value larger than we would have chosen before, so that at worst an overestimate leaves us no wiser than we were before.
2015-03-12 02:21:00 +01:00
{
List *semi_operators;
List *semi_rhs_exprs;
bool all_btree;
bool all_hash;
ListCell *lc;
/* Initialize semijoin-related fields in case we can't unique-ify */
sjinfo->semi_can_btree = false;
sjinfo->semi_can_hash = false;
sjinfo->semi_operators = NIL;
sjinfo->semi_rhs_exprs = NIL;
/* Nothing more to do if it's not a semijoin */
if (sjinfo->jointype != JOIN_SEMI)
return;
/*
* Look to see whether the semijoin's join quals consist of AND'ed
* equality operators, with (only) RHS variables on only one side of each
* one. If so, we can figure out how to enforce uniqueness for the RHS.
*
* Note that the input clause list is the list of quals that are
* *syntactically* associated with the semijoin, which in practice means
* the synthesized comparison list for an IN or the WHERE of an EXISTS.
* Particularly in the latter case, it might contain clauses that aren't
* *semantically* associated with the join, but refer to just one side or
* the other. We can ignore such clauses here, as they will just drop
* down to be processed within one side or the other. (It is okay to
* consider only the syntactically-associated clauses here because for a
* semijoin, no higher-level quals could refer to the RHS, and so there
* can be no other quals that are semantically associated with this join.
* We do things this way because it is useful to have the set of potential
* unique-ification expressions before we can extract the list of quals
* that are actually semantically associated with the particular join.)
*
* Note that the semi_operators list consists of the joinqual operators
* themselves (but commuted if needed to put the RHS value on the right).
* These could be cross-type operators, in which case the operator
* actually needed for uniqueness is a related single-type operator. We
* assume here that that operator will be available from the btree or hash
* opclass when the time comes ... if not, create_unique_plan() will fail.
*/
semi_operators = NIL;
semi_rhs_exprs = NIL;
all_btree = true;
all_hash = enable_hashagg; /* don't consider hash if not enabled */
foreach(lc, clause)
{
OpExpr *op = (OpExpr *) lfirst(lc);
Oid opno;
Node *left_expr;
Node *right_expr;
Relids left_varnos;
Relids right_varnos;
Relids all_varnos;
Oid opinputtype;
/* Is it a binary opclause? */
if (!IsA(op, OpExpr) ||
list_length(op->args) != 2)
{
/* No, but does it reference both sides? */
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
all_varnos = pull_varnos(root, (Node *) op);
Improve planner's cost estimation in the presence of semijoins. If we have a semijoin, say SELECT * FROM x WHERE x1 IN (SELECT y1 FROM y) and we're estimating the cost of a parameterized indexscan on x, the number of repetitions of the indexscan should not be taken as the size of y; it'll really only be the number of distinct values of y1, because the only valid plan with y on the outside of a nestloop would require y to be unique-ified before joining it to x. Most of the time this doesn't make that much difference, but sometimes it can lead to drastically underestimating the cost of the indexscan and hence choosing a bad plan, as pointed out by David Kubečka. Fixing this is a bit difficult because parameterized indexscans are costed out quite early in the planning process, before we have the information that would be needed to call estimate_num_groups() and thereby estimate the number of distinct values of the join column(s). However we can move the code that extracts a semijoin RHS's unique-ification columns, so that it's done in initsplan.c rather than on-the-fly in create_unique_path(). That shouldn't make any difference speed-wise and it's really a bit cleaner too. The other bit of information we need is the size of the semijoin RHS, which is easy if it's a single relation (we make those estimates before considering indexscan costs) but problematic if it's a join relation. The solution adopted here is just to use the product of the sizes of the join component rels. That will generally be an overestimate, but since estimate_num_groups() only uses this input as a clamp, an overestimate shouldn't hurt us too badly. In any case we don't allow this new logic to produce a value larger than we would have chosen before, so that at worst an overestimate leaves us no wiser than we were before.
2015-03-12 02:21:00 +01:00
if (!bms_overlap(all_varnos, sjinfo->syn_righthand) ||
bms_is_subset(all_varnos, sjinfo->syn_righthand))
{
/*
* Clause refers to only one rel, so ignore it --- unless it
* contains volatile functions, in which case we'd better
* punt.
*/
if (contain_volatile_functions((Node *) op))
return;
continue;
}
/* Non-operator clause referencing both sides, must punt */
return;
}
/* Extract data from binary opclause */
opno = op->opno;
left_expr = linitial(op->args);
right_expr = lsecond(op->args);
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
left_varnos = pull_varnos(root, left_expr);
right_varnos = pull_varnos(root, right_expr);
Improve planner's cost estimation in the presence of semijoins. If we have a semijoin, say SELECT * FROM x WHERE x1 IN (SELECT y1 FROM y) and we're estimating the cost of a parameterized indexscan on x, the number of repetitions of the indexscan should not be taken as the size of y; it'll really only be the number of distinct values of y1, because the only valid plan with y on the outside of a nestloop would require y to be unique-ified before joining it to x. Most of the time this doesn't make that much difference, but sometimes it can lead to drastically underestimating the cost of the indexscan and hence choosing a bad plan, as pointed out by David Kubečka. Fixing this is a bit difficult because parameterized indexscans are costed out quite early in the planning process, before we have the information that would be needed to call estimate_num_groups() and thereby estimate the number of distinct values of the join column(s). However we can move the code that extracts a semijoin RHS's unique-ification columns, so that it's done in initsplan.c rather than on-the-fly in create_unique_path(). That shouldn't make any difference speed-wise and it's really a bit cleaner too. The other bit of information we need is the size of the semijoin RHS, which is easy if it's a single relation (we make those estimates before considering indexscan costs) but problematic if it's a join relation. The solution adopted here is just to use the product of the sizes of the join component rels. That will generally be an overestimate, but since estimate_num_groups() only uses this input as a clamp, an overestimate shouldn't hurt us too badly. In any case we don't allow this new logic to produce a value larger than we would have chosen before, so that at worst an overestimate leaves us no wiser than we were before.
2015-03-12 02:21:00 +01:00
all_varnos = bms_union(left_varnos, right_varnos);
opinputtype = exprType(left_expr);
/* Does it reference both sides? */
if (!bms_overlap(all_varnos, sjinfo->syn_righthand) ||
bms_is_subset(all_varnos, sjinfo->syn_righthand))
{
/*
* Clause refers to only one rel, so ignore it --- unless it
* contains volatile functions, in which case we'd better punt.
*/
if (contain_volatile_functions((Node *) op))
return;
continue;
}
/* check rel membership of arguments */
if (!bms_is_empty(right_varnos) &&
bms_is_subset(right_varnos, sjinfo->syn_righthand) &&
!bms_overlap(left_varnos, sjinfo->syn_righthand))
{
/* typical case, right_expr is RHS variable */
}
else if (!bms_is_empty(left_varnos) &&
bms_is_subset(left_varnos, sjinfo->syn_righthand) &&
!bms_overlap(right_varnos, sjinfo->syn_righthand))
{
/* flipped case, left_expr is RHS variable */
opno = get_commutator(opno);
if (!OidIsValid(opno))
return;
right_expr = left_expr;
}
else
{
/* mixed membership of args, punt */
return;
}
/* all operators must be btree equality or hash equality */
if (all_btree)
{
/* oprcanmerge is considered a hint... */
if (!op_mergejoinable(opno, opinputtype) ||
get_mergejoin_opfamilies(opno) == NIL)
all_btree = false;
}
if (all_hash)
{
/* ... but oprcanhash had better be correct */
if (!op_hashjoinable(opno, opinputtype))
all_hash = false;
}
if (!(all_btree || all_hash))
return;
/* so far so good, keep building lists */
semi_operators = lappend_oid(semi_operators, opno);
semi_rhs_exprs = lappend(semi_rhs_exprs, copyObject(right_expr));
}
/* Punt if we didn't find at least one column to unique-ify */
if (semi_rhs_exprs == NIL)
return;
/*
* The expressions we'd need to unique-ify mustn't be volatile.
*/
if (contain_volatile_functions((Node *) semi_rhs_exprs))
return;
/*
* If we get here, we can unique-ify the semijoin's RHS using at least one
* of sorting and hashing. Save the information about how to do that.
*/
sjinfo->semi_can_btree = all_btree;
sjinfo->semi_can_hash = all_hash;
sjinfo->semi_operators = semi_operators;
sjinfo->semi_rhs_exprs = semi_rhs_exprs;
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* deconstruct_distribute_oj_quals
* Adjust LEFT JOIN quals to be suitable for commuted-left-join cases,
* then push them into the joinqual lists and EquivalenceClass structures.
*
* This runs immediately after we've completed the deconstruct_distribute scan.
* jtitems contains all the JoinTreeItems (in depth-first order), and jtitem
* is one that has postponed oj_joinclauses to deal with.
*/
static void
deconstruct_distribute_oj_quals(PlannerInfo *root,
List *jtitems,
JoinTreeItem *jtitem)
{
SpecialJoinInfo *sjinfo = jtitem->sjinfo;
Relids qualscope,
ojscope,
nonnullable_rels;
/* Recompute syntactic and semantic scopes of this left join */
qualscope = bms_union(sjinfo->syn_lefthand, sjinfo->syn_righthand);
qualscope = bms_add_member(qualscope, sjinfo->ojrelid);
ojscope = bms_union(sjinfo->min_lefthand, sjinfo->min_righthand);
nonnullable_rels = sjinfo->syn_lefthand;
/*
* If this join can commute with any other ones per outer-join identity 3,
* and it is the one providing the join clause with flexible semantics,
* then we have to generate variants of the join clause with different
* nullingrels labeling. Otherwise, just push out the postponed clause
* as-is.
*/
Assert(sjinfo->lhs_strict); /* else we shouldn't be here */
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
if (sjinfo->commute_above_r || sjinfo->commute_below_l)
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
{
Relids joins_above;
Relids joins_below;
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
Relids incompatible_joins;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
Relids joins_so_far;
List *quals;
int save_last_rinfo_serial;
ListCell *lc;
/* Identify the outer joins this one commutes with */
joins_above = sjinfo->commute_above_r;
Fix some issues with improper placement of outer join clauses. After applying outer-join identity 3 in the forward direction, it was possible for the planner to mistakenly apply a qual clause from above the two outer joins at the now-lower join level. This can give the wrong answer, since a value that would get nulled by the now-upper join might not yet be null. To fix, when we perform such a transformation, consider that the now-lower join hasn't really completed the outer join it's nominally responsible for and thus its relid set should not include that OJ's relid (nor should its output Vars have that nullingrel bit set). Instead we add those bits when the now-upper join is performed. The existing rules for qual placement then suffice to prevent higher qual clauses from dropping below the now-upper join. There are a few complications from needing to consider transitive closures in case multiple pushdowns have happened, but all in all it's not a very complex patch. This is all new logic (from 2489d76c4) so no need to back-patch. The added test cases all have the same results as in v15. Tom Lane and Richard Guo Discussion: https://postgr.es/m/0b819232-4b50-f245-1c7d-c8c61bf41827@postgrespro.ru
2023-05-17 17:13:52 +02:00
joins_below = sjinfo->commute_below_l;
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* Generate qual variants with different sets of nullingrels bits.
*
* We only need bit-sets that correspond to the successively less
* deeply syntactically-nested subsets of this join and its
* commutators. That's true first because obviously only those forms
* of the Vars and PHVs could appear elsewhere in the query, and
* second because the outer join identities do not provide a way to
* re-order such joins in a way that would require different marking.
* (That is, while the current join may commute with several others,
* none of those others can commute with each other.) To visit the
* interesting joins in syntactic nesting order, we rely on the
* jtitems list to be ordered that way.
*
* We first strip out all the nullingrels bits corresponding to
* commutating joins below this one, and then successively put them
* back as we crawl up the join stack.
*/
quals = jtitem->oj_joinclauses;
if (!bms_is_empty(joins_below))
quals = (List *) remove_nulling_relids((Node *) quals,
joins_below,
NULL);
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
/*
* We'll need to mark the lower versions of the quals as not safe to
* apply above not-yet-processed joins of the stack. This prevents
* possibly applying a cloned qual at the wrong join level.
*/
incompatible_joins = bms_union(joins_below, joins_above);
incompatible_joins = bms_add_member(incompatible_joins,
sjinfo->ojrelid);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* Each time we produce RestrictInfo(s) from these quals, reset the
* last_rinfo_serial counter, so that the RestrictInfos for the "same"
* qual condition get identical serial numbers. (This relies on the
* fact that we're not changing the qual list in any way that'd affect
* the number of RestrictInfos built from it.) This'll allow us to
* detect duplicative qual usage later.
*/
save_last_rinfo_serial = root->last_rinfo_serial;
joins_so_far = NULL;
foreach(lc, jtitems)
{
JoinTreeItem *otherjtitem = (JoinTreeItem *) lfirst(lc);
SpecialJoinInfo *othersj = otherjtitem->sjinfo;
bool below_sjinfo = false;
bool above_sjinfo = false;
Relids this_qualscope;
Relids this_ojscope;
bool allow_equivalence,
has_clone,
is_clone;
if (othersj == NULL)
continue; /* not an outer-join item, ignore */
if (bms_is_member(othersj->ojrelid, joins_below))
{
/* othersj commutes with sjinfo from below left */
below_sjinfo = true;
}
else if (othersj == sjinfo)
{
/* found our join in syntactic order */
Assert(bms_equal(joins_so_far, joins_below));
}
else if (bms_is_member(othersj->ojrelid, joins_above))
{
/* othersj commutes with sjinfo from above */
above_sjinfo = true;
}
else
{
/* othersj is not relevant, ignore */
continue;
}
/* Reset serial counter for this version of the quals */
root->last_rinfo_serial = save_last_rinfo_serial;
/*
* When we are looking at joins above sjinfo, we are envisioning
* pushing sjinfo to above othersj, so add othersj's nulling bit
Further fixes in qual nullingrel adjustment for outer join commutation. One of the add_nulling_relids calls in deconstruct_distribute_oj_quals added an OJ relid to too few Vars, while the other added it to too many. We should consider the syntactic structure not min_left/righthand while deciding which Vars to decorate, and when considering pushing up a lower outer join pursuant to transforming the second form of OJ identity 3 to the first form, we only want to decorate Vars coming from its LHS. In a related bug, I realized that make_outerjoininfo was failing to check a very basic property that's needed to apply OJ identity 3: the syntactically-upper outer join clause can't refer to the lower join's LHS. This didn't break the join order restriction logic, but it led to setting bogus commute_xxx bits, possibly resulting in bogus nullingrel markings in modified quals. Richard Guo and Tom Lane Discussion: https://postgr.es/m/CAMbWs497CmBruMx1SOjepWEz+T5NWa4scqbdE9v7ZzSXqH_gQw@mail.gmail.com Discussion: https://postgr.es/m/CAEP4nAx9C5gXNBfEA0JBfz7B+5f1Bawt-RWQWyhev-wdps8BZA@mail.gmail.com
2023-02-10 19:31:00 +01:00
* before distributing the quals. We should add it to Vars coming
* from the current join's LHS: we want to transform the second
* form of OJ identity 3 to the first form, in which Vars of
* relation B will appear nulled by the syntactically-upper OJ
* within the Pbc clause, but those of relation C will not. (In
* the notation used by optimizer/README, we're converting a qual
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
* of the form Pbc to Pb*c.) Of course, we must also remove that
* bit from the incompatible_joins value, else we'll make a qual
* that can't be placed anywhere.
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
*/
if (above_sjinfo)
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
quals = (List *)
add_nulling_relids((Node *) quals,
Further fixes in qual nullingrel adjustment for outer join commutation. One of the add_nulling_relids calls in deconstruct_distribute_oj_quals added an OJ relid to too few Vars, while the other added it to too many. We should consider the syntactic structure not min_left/righthand while deciding which Vars to decorate, and when considering pushing up a lower outer join pursuant to transforming the second form of OJ identity 3 to the first form, we only want to decorate Vars coming from its LHS. In a related bug, I realized that make_outerjoininfo was failing to check a very basic property that's needed to apply OJ identity 3: the syntactically-upper outer join clause can't refer to the lower join's LHS. This didn't break the join order restriction logic, but it led to setting bogus commute_xxx bits, possibly resulting in bogus nullingrel markings in modified quals. Richard Guo and Tom Lane Discussion: https://postgr.es/m/CAMbWs497CmBruMx1SOjepWEz+T5NWa4scqbdE9v7ZzSXqH_gQw@mail.gmail.com Discussion: https://postgr.es/m/CAEP4nAx9C5gXNBfEA0JBfz7B+5f1Bawt-RWQWyhev-wdps8BZA@mail.gmail.com
2023-02-10 19:31:00 +01:00
sjinfo->syn_lefthand,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bms_make_singleton(othersj->ojrelid));
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
incompatible_joins = bms_del_member(incompatible_joins,
othersj->ojrelid);
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* Compute qualscope and ojscope for this join level */
this_qualscope = bms_union(qualscope, joins_so_far);
this_ojscope = bms_union(ojscope, joins_so_far);
if (above_sjinfo)
{
/* othersj is not yet in joins_so_far, but we need it */
this_qualscope = bms_add_member(this_qualscope,
othersj->ojrelid);
this_ojscope = bms_add_member(this_ojscope,
othersj->ojrelid);
/* sjinfo is in joins_so_far, and we don't want it */
this_ojscope = bms_del_member(this_ojscope,
sjinfo->ojrelid);
}
/*
* We generate EquivalenceClasses only from the first form of the
* quals, with the fewest nullingrels bits set. An EC made from
* this version of the quals can be useful below the outer-join
* nest, whereas versions with some nullingrels bits set would not
* be. We cannot generate ECs from more than one version, or
* we'll make nonsensical conclusions that Vars with nullingrels
* bits set are equal to their versions without. Fortunately,
* such ECs wouldn't be very useful anyway, because they'd equate
* values not observable outside the join nest. (See
* optimizer/README.)
*
* The first form of the quals is also the only one marked as
* has_clone rather than is_clone.
*/
allow_equivalence = (joins_so_far == NULL);
has_clone = allow_equivalence;
is_clone = !has_clone;
distribute_quals_to_rels(root, quals,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
otherjtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
sjinfo,
root->qual_security_level,
this_qualscope,
this_ojscope, nonnullable_rels,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
bms_copy(incompatible_joins),
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
allow_equivalence,
has_clone,
is_clone,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
NULL); /* no more postponement */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* Adjust qual nulling bits for next level up, if needed. We
* don't want to put sjinfo's own bit in at all, and if we're
Further fixes in qual nullingrel adjustment for outer join commutation. One of the add_nulling_relids calls in deconstruct_distribute_oj_quals added an OJ relid to too few Vars, while the other added it to too many. We should consider the syntactic structure not min_left/righthand while deciding which Vars to decorate, and when considering pushing up a lower outer join pursuant to transforming the second form of OJ identity 3 to the first form, we only want to decorate Vars coming from its LHS. In a related bug, I realized that make_outerjoininfo was failing to check a very basic property that's needed to apply OJ identity 3: the syntactically-upper outer join clause can't refer to the lower join's LHS. This didn't break the join order restriction logic, but it led to setting bogus commute_xxx bits, possibly resulting in bogus nullingrel markings in modified quals. Richard Guo and Tom Lane Discussion: https://postgr.es/m/CAMbWs497CmBruMx1SOjepWEz+T5NWa4scqbdE9v7ZzSXqH_gQw@mail.gmail.com Discussion: https://postgr.es/m/CAEP4nAx9C5gXNBfEA0JBfz7B+5f1Bawt-RWQWyhev-wdps8BZA@mail.gmail.com
2023-02-10 19:31:00 +01:00
* above sjinfo then we did it already. Here, we should mark all
* Vars coming from the lower join's RHS. (Again, we are
* converting a qual of the form Pbc to Pb*c, but now we are
* putting back bits that were there in the parser output and were
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
* temporarily stripped above.) Update incompatible_joins too.
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
*/
if (below_sjinfo)
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
{
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
quals = (List *)
add_nulling_relids((Node *) quals,
Further fixes in qual nullingrel adjustment for outer join commutation. One of the add_nulling_relids calls in deconstruct_distribute_oj_quals added an OJ relid to too few Vars, while the other added it to too many. We should consider the syntactic structure not min_left/righthand while deciding which Vars to decorate, and when considering pushing up a lower outer join pursuant to transforming the second form of OJ identity 3 to the first form, we only want to decorate Vars coming from its LHS. In a related bug, I realized that make_outerjoininfo was failing to check a very basic property that's needed to apply OJ identity 3: the syntactically-upper outer join clause can't refer to the lower join's LHS. This didn't break the join order restriction logic, but it led to setting bogus commute_xxx bits, possibly resulting in bogus nullingrel markings in modified quals. Richard Guo and Tom Lane Discussion: https://postgr.es/m/CAMbWs497CmBruMx1SOjepWEz+T5NWa4scqbdE9v7ZzSXqH_gQw@mail.gmail.com Discussion: https://postgr.es/m/CAEP4nAx9C5gXNBfEA0JBfz7B+5f1Bawt-RWQWyhev-wdps8BZA@mail.gmail.com
2023-02-10 19:31:00 +01:00
othersj->syn_righthand,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bms_make_singleton(othersj->ojrelid));
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
incompatible_joins = bms_del_member(incompatible_joins,
othersj->ojrelid);
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/* ... and track joins processed so far */
joins_so_far = bms_add_member(joins_so_far, othersj->ojrelid);
}
}
else
{
/* No commutation possible, just process the postponed clauses */
distribute_quals_to_rels(root, jtitem->oj_joinclauses,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
sjinfo,
root->qual_security_level,
qualscope,
ojscope, nonnullable_rels,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
NULL, /* incompatible_relids */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
true, /* allow_equivalence */
false, false, /* not clones */
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
NULL); /* no more postponement */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
}
}
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
/*****************************************************************************
*
* QUALIFICATIONS
*
*****************************************************************************/
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
/*
* distribute_quals_to_rels
* Convenience routine to apply distribute_qual_to_rels to each element
* of an AND'ed list of clauses.
*/
static void
distribute_quals_to_rels(PlannerInfo *root, List *clauses,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
SpecialJoinInfo *sjinfo,
Index security_level,
Relids qualscope,
Relids ojscope,
Relids outerjoin_nonnullable,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
Relids incompatible_relids,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bool allow_equivalence,
bool has_clone,
bool is_clone,
List **postponed_oj_qual_list)
{
ListCell *lc;
foreach(lc, clauses)
{
Node *clause = (Node *) lfirst(lc);
distribute_qual_to_rels(root, clause,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
sjinfo,
security_level,
qualscope,
ojscope,
outerjoin_nonnullable,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
incompatible_relids,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
allow_equivalence,
has_clone,
is_clone,
postponed_oj_qual_list);
}
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
* distribute_qual_to_rels
Simplify the planner's join clause management by storing join clauses of a relation in a flat 'joininfo' list. The former arrangement grouped the join clauses according to the set of unjoined relids used in each; however, profiling on test cases involving lots of joins proves that that data structure is a net loss. It takes more time to group the join clauses together than is saved by avoiding duplicate tests later. It doesn't help any that there are usually not more than one or two clauses per group ...
2005-06-09 06:19:00 +02:00
* Add clause information to either the baserestrictinfo or joininfo list
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
* (depending on whether the clause is a join) of each base relation
* mentioned in the clause. A RestrictInfo node is created and added to
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* the appropriate list for each rel. Alternatively, if the clause uses a
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* mergejoinable operator, enter its left- and right-side expressions into
* the query's EquivalenceClasses.
*
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* In some cases, quals will be added to parent jtitems' lateral_clauses
* or to postponed_oj_qual_list instead of being processed right away.
* These will be dealt with in later calls of deconstruct_distribute.
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
* 'clause': the qual clause to be distributed
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* 'jtitem': the JoinTreeItem for the containing jointree node
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* 'sjinfo': join's SpecialJoinInfo (NULL for an inner join or WHERE clause)
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
* 'security_level': security_level to assign to the qual
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* 'qualscope': set of base+OJ rels the qual's syntactic scope covers
* 'ojscope': NULL if not an outer-join qual, else the minimum set of base+OJ
* rels needed to form this join
Be smart about outer-join qualifications that mention only one side of the join, per recent discussion on pgsql-sql. Not clear that this will come up often in real queries, but it's not any more expensive to do it right, so we may as well do it right.
2003-03-03 00:46:34 +01:00
* 'outerjoin_nonnullable': NULL if not an outer-join qual, else the set of
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* base+OJ rels appearing on the outer (nonnullable) side of the join
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* (for FULL JOIN this includes both sides of the join, and must in fact
* equal qualscope)
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
* 'incompatible_relids': the set of outer-join relid(s) that must not be
* computed below this qual. We only bother to compute this for
* "clone" quals, otherwise it can be left NULL.
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* 'allow_equivalence': true if it's okay to convert clause into an
* EquivalenceClass
* 'has_clone': has_clone property to assign to the qual
* 'is_clone': is_clone property to assign to the qual
* 'postponed_oj_qual_list': if not NULL, non-degenerate outer join clauses
* should be added to this list instead of being processed (list entries
* are just the bare clauses)
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
*
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
* 'qualscope' identifies what level of JOIN the qual came from syntactically.
* 'ojscope' is needed if we decide to force the qual up to the outer-join
* level, which will be ojscope not necessarily qualscope.
Simplify overcomplicated (and overly restrictive) test to see whether an IS NULL condition is rendered redundant by detection of an antijoin. If we know that a join is an antijoin, then *any* Var coming out of its righthand side must be NULL, not only the joining column(s). Also, it's still gonna be null after being passed up through higher joins, whether they're outer joins or not. I was misled by a faulty analogy to reduce_outer_joins() in the original coding. But consider select * from a left join b on a.x = b.y where b.y is null and b.z is null; The first IS NULL condition justifies deciding that the join is an antijoin (if the = is strict) and then the second one is just plain redundant.
2009-02-20 01:01:03 +01:00
*
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
* At the time this is called, root->join_info_list must contain entries for
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
* at least those special joins that are syntactically below this qual.
* (We now need that only for detection of redundant IS NULL quals.)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
static void
Remove planner's private fields from Query struct, and put them into a new PlannerInfo struct, which is passed around instead of the bare Query in all the planning code. This commit is essentially just a code-beautification exercise, but it does open the door to making larger changes to the planner data structures without having to muck with the widely-known Query struct.
2005-06-06 00:32:58 +02:00
distribute_qual_to_rels(PlannerInfo *root, Node *clause,
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *jtitem,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
SpecialJoinInfo *sjinfo,
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
Index security_level,
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Relids qualscope,
Relids ojscope,
Fix planning of non-strict equivalence clauses above outer joins. If a potential equivalence clause references a variable from the nullable side of an outer join, the planner needs to take care that derived clauses are not pushed to below the outer join; else they may use the wrong value for the variable. (The problem arises only with non-strict clauses, since if an upper clause can be proven strict then the outer join will get simplified to a plain join.) The planner attempted to prevent this type of error by checking that potential equivalence clauses aren't outerjoin-delayed as a whole, but actually we have to check each side separately, since the two sides of the clause will get moved around separately if it's treated as an equivalence. Bugs of this type can be demonstrated as far back as 7.4, even though releases before 8.3 had only a very ad-hoc notion of equivalence clauses. In addition, we neglected to account for the possibility that such clauses might have nonempty nullable_relids even when not outerjoin-delayed; so the equivalence-class machinery lacked logic to compute correct nullable_relids values for clauses it constructs. This oversight was harmless before 9.2 because we were only using RestrictInfo.nullable_relids for OR clauses; but as of 9.2 it could result in pushing constructed equivalence clauses to incorrect places. (This accounts for bug #7604 from Bill MacArthur.) Fix the first problem by adding a new test check_equivalence_delay() in distribute_qual_to_rels, and fix the second one by adding code in equivclass.c and called functions to set correct nullable_relids for generated clauses. Although I believe the second part of this is not currently necessary before 9.2, I chose to back-patch it anyway, partly to keep the logic similar across branches and partly because it seems possible we might find other reasons why we need valid values of nullable_relids in the older branches. Add regression tests illustrating these problems. In 9.0 and up, also add test cases checking that we can push constants through outer joins, since we've broken that optimization before and I nearly broke it again with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
Relids outerjoin_nonnullable,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
Relids incompatible_relids,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
bool allow_equivalence,
bool has_clone,
bool is_clone,
List **postponed_oj_qual_list)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Fix bushy plans. Cleanup.
1999-02-18 01:49:48 +01:00
Relids relids;
Adjust the definition of is_pushed_down so that it's always true for INNER JOIN quals, just like WHERE quals, even if they reference every one of the join's relations. Now that we can reorder outer and inner joins, it's possible for such a qual to end up being assigned to an outer join plan node, and we mustn't have it treated as a join qual rather than a filter qual for the node. (If it were, the join could produce null-extended rows that it shouldn't.) Per bug report from Pelle Johansson.
2007-02-16 21:57:19 +01:00
bool is_pushed_down;
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
bool pseudoconstant = false;
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
bool maybe_equivalence;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
bool maybe_outer_join;
Rewrite OR indexscan processing to be more flexible. We can now for the first time generate an OR indexscan for a two-column index when the WHERE condition is like 'col1 = foo AND (col2 = bar OR col2 = baz)' --- before, the OR had to be on the first column of the index or we'd not notice the possibility of using it. Some progress towards extracting OR indexscans from subclauses of an OR that references multiple relations, too, although this code is #ifdef'd out because it needs more work.
2004-01-04 01:07:32 +01:00
RestrictInfo *restrictinfo;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
Clean up messy clause-selectivity code in clausesel.c; repair bug identified by Hiroshi (incorrect cost attributed to OR clauses after multiple passes through set_rest_selec()). I think the code was trying to allow selectivities of OR subclauses to be passed in from outside, but noplace was actually passing any useful data, and set_rest_selec() was passing wrong data. Restructure representation of "indexqual" in IndexPath nodes so that it is the same as for indxqual in completed IndexScan nodes: namely, a toplevel list with an entry for each pass of the index scan, having sublists that are implicitly-ANDed index qual conditions for that pass. You don't want to know what the old representation was :-( Improve documentation of OR-clause indexscan functions. Remove useless 'notclause' field from RestrictInfo nodes. (This might force an initdb for anyone who has stored rules containing RestrictInfos, but I do not think that RestrictInfo ever appears in completed plans.)
1999-07-25 01:21:14 +02:00
/*
Merge restrictlist_selectivity into clauselist_selectivity by teaching the latter to accept either RestrictInfo nodes or bare clause expressions; and cache the selectivity result in the RestrictInfo node when possible. This extends the caching behavior of approx_selectivity to many more contexts, and should reduce duplicate selectivity calculations.
2004-01-04 04:51:52 +01:00
* Retrieve all relids mentioned within the clause.
Clean up messy clause-selectivity code in clausesel.c; repair bug identified by Hiroshi (incorrect cost attributed to OR clauses after multiple passes through set_rest_selec()). I think the code was trying to allow selectivities of OR subclauses to be passed in from outside, but noplace was actually passing any useful data, and set_rest_selec() was passing wrong data. Restructure representation of "indexqual" in IndexPath nodes so that it is the same as for indxqual in completed IndexScan nodes: namely, a toplevel list with an entry for each pass of the index scan, having sublists that are implicitly-ANDed index qual conditions for that pass. You don't want to know what the old representation was :-( Improve documentation of OR-clause indexscan functions. Remove useless 'notclause' field from RestrictInfo nodes. (This might force an initdb for anyone who has stored rules containing RestrictInfos, but I do not think that RestrictInfo ever appears in completed plans.)
1999-07-25 01:21:14 +02:00
*/
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
relids = pull_varnos(root, clause);
Clean up messy clause-selectivity code in clausesel.c; repair bug identified by Hiroshi (incorrect cost attributed to OR clauses after multiple passes through set_rest_selec()). I think the code was trying to allow selectivities of OR subclauses to be passed in from outside, but noplace was actually passing any useful data, and set_rest_selec() was passing wrong data. Restructure representation of "indexqual" in IndexPath nodes so that it is the same as for indxqual in completed IndexScan nodes: namely, a toplevel list with an entry for each pass of the index scan, having sublists that are implicitly-ANDed index qual conditions for that pass. You don't want to know what the old representation was :-( Improve documentation of OR-clause indexscan functions. Remove useless 'notclause' field from RestrictInfo nodes. (This might force an initdb for anyone who has stored rules containing RestrictInfos, but I do not think that RestrictInfo ever appears in completed plans.)
1999-07-25 01:21:14 +02:00
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
/*
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
* In ordinary SQL, a WHERE or JOIN/ON clause can't reference any rels
* that aren't within its syntactic scope; however, if we pulled up a
* LATERAL subquery then we might find such references in quals that have
* been pulled up. We need to treat such quals as belonging to the join
* level that includes every rel they reference. Although we could make
* pull_up_subqueries() place such quals correctly to begin with, it's
* easier to handle it here. When we find a clause that contains Vars
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
* outside its syntactic scope, locate the nearest parent join level that
* includes all the required rels and add the clause to that level's
* lateral_clauses list. We'll process it when we reach that join level.
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
*/
if (!bms_is_subset(relids, qualscope))
{
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
JoinTreeItem *pitem;
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
Assert(root->hasLateralRTEs); /* shouldn't happen otherwise */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
Assert(sjinfo == NULL); /* mustn't postpone past outer join */
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
for (pitem = jtitem->jti_parent; pitem; pitem = pitem->jti_parent)
{
if (bms_is_subset(relids, pitem->qualscope))
{
pitem->lateral_clauses = lappend(pitem->lateral_clauses,
clause);
return;
}
/*
* We should not be postponing any quals past an outer join. If
* this Assert fires, pull_up_subqueries() messed up.
*/
Assert(pitem->sjinfo == NULL);
}
elog(ERROR, "failed to postpone qual containing lateral reference");
Fix qual-clause-misplacement issues with pulled-up LATERAL subqueries. In an example such as SELECT * FROM i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true; it is safe to pull up the LATERAL subquery into its parent, but we must then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN. The previous coding in deconstruct_recurse mistakenly labeled the clause as "is_pushed_down", resulting in wrong semantics if the clause were applied at the join node, as per an example submitted awhile ago by Jeremy Evans. To fix, postpone processing of such clauses until we return back up to the appropriate recursion depth in deconstruct_recurse. In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery; we previously missed the possibility that the LATERAL subquery might itself contain an outer join that makes lateral references in lower quals unsafe. A regression test case equivalent to Jeremy's example was already in my commit of yesterday, but was giving the wrong results because of this bug. This patch fixes the expected output for that, and also adds a test case for the second problem.
2013-08-19 19:19:25 +02:00
}
/*
* If it's an outer-join clause, also check that relids is a subset of
* ojscope. (This should not fail if the syntactic scope check passed.)
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
*/
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
if (ojscope && !bms_is_subset(relids, ojscope))
Wording cleanup for error messages. Also change can't -> cannot. Standard English uses "may", "can", and "might" in different ways: may - permission, "You may borrow my rake." can - ability, "I can lift that log." might - possibility, "It might rain today." Unfortunately, in conversational English, their use is often mixed, as in, "You may use this variable to do X", when in fact, "can" is a better choice. Similarly, "It may crash" is better stated, "It might crash".
2007-02-01 20:10:30 +01:00
elog(ERROR, "JOIN qualification cannot refer to other relations");
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
/*
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
* If the clause is variable-free, our normal heuristic for pushing it
* down to just the mentioned rels doesn't work, because there are none.
*
* If the clause is an outer-join clause, we must force it to the OJ's
* semantic level to preserve semantics.
*
* Otherwise, when the clause contains volatile functions, we force it to
* be evaluated at its original syntactic level. This preserves the
* expected semantics.
*
* When the clause contains no volatile functions either, it is actually a
* pseudoconstant clause that will not change value during any one
* execution of the plan, and hence can be used as a one-time qual in a
* gating Result plan node. We put such a clause into the regular
* RestrictInfo lists for the moment, but eventually createplan.c will
* pull it out and make a gating Result node immediately above whatever
* plan node the pseudoconstant clause is assigned to. It's usually best
Fix some issues with wrong placement of pseudo-constant quals. initsplan.c figured that it could push Var-free qual clauses to the top of the current JoinDomain, which is okay in the abstract. But if the current domain is inside some outer join, and we later commute an inside-the-domain outer join with one outside it, we end up placing the pushed-up qual clause incorrectly. In distribute_qual_to_rels, avoid this by using the syntactic scope of the qual clause; with the exception that if we're in the top-level join domain we can still use the full query relid set, ensuring the resulting gating Result node goes to the top of the plan. (This is approximately as smart as the pre-v16 code was. Perhaps we can do better later, but it's not clear that such cases are worth a lot of sweat.) In process_implied_equality, we don't have a clear notion of syntactic scope, but we do have the results of SpecialJoinInfo construction. Thumb through those and remove any lower outer joins that might get commuted to above the join domain. Again, we can make an exception for the top-level join domain. It'd be possible to work harder here (for example, by keeping outer joins that aren't shown as potentially commutable), but I'm going to stop here for the moment. This issue has convinced me that the current representation of join domains probably needs further refinement, so I'm disinclined to write inessential dependent logic just yet. In passing, tighten the qualscope passed to process_implied_equality by generate_base_implied_equalities_no_const; there's no need for it to be larger than the rel we are currently considering. Tom Lane and Richard Guo, per report from Tender Wang. Discussion: https://postgr.es/m/CAHewXNk9eJ35ru5xATWioTV4+xZPHptjy9etdcNPjUfY9RQ+uQ@mail.gmail.com
2023-02-22 18:39:07 +01:00
* to put a gating node as high in the plan tree as possible.
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
*/
Replace planner's representation of relation sets, per pghackers discussion. Instead of Lists of integers, we now store variable-length bitmap sets. This should be faster as well as less error-prone.
2003-02-08 21:20:55 +01:00
if (bms_is_empty(relids))
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
{
if (ojscope)
{
/* clause is attached to outer join, eval it there */
Fix some planner issues found while investigating Kevin Grittner's report of poorer planning in 8.3 than 8.2: 1. After pushing a constant across an outer join --- ie, given "a LEFT JOIN b ON (a.x = b.y) WHERE a.x = 42", we can deduce that b.y is sort of equal to 42, in the sense that we needn't fetch any b rows where it isn't 42 --- loop to see if any additional deductions can be made. Previous releases did that by recursing, but I had mistakenly thought that this was no longer necessary given the EquivalenceClass machinery. 2. Allow pushing constants across outer join conditions even if the condition is outerjoin_delayed due to a lower outer join. This is safe as long as the condition is strict and we re-test it at the upper join. 3. Keep the outer-join clause even if we successfully push a constant across it. This is *necessary* in the outerjoin_delayed case, but even in the simple case, it seems better to do this to ensure that the join search order heuristics will consider the join as reasonable to make. Mark such a clause as having selectivity 1.0, though, since it's not going to eliminate very many rows after application of the constant condition. 4. Tweak have_relevant_eclass_joinclause to report that two relations are joinable when they have vars that are equated to the same constant. We won't actually generate any joinclause from such an EquivalenceClass, but again it seems that in such a case it's a good idea to consider the join as worth costing out. 5. Fix a bug in select_mergejoin_clauses that was exposed by these changes: we have to reject candidate mergejoin clauses if either side was equated to a constant, because we can't construct a canonical pathkey list for such a clause. This is an implementation restriction that might be worth fixing someday, but it doesn't seem critical to get it done for 8.3.
2008-01-09 21:42:29 +01:00
relids = bms_copy(ojscope);
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
/* mustn't use as gating qual, so don't mark pseudoconstant */
}
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
else if (contain_volatile_functions(clause))
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
{
/* eval at original syntactic level */
Fix some planner issues found while investigating Kevin Grittner's report of poorer planning in 8.3 than 8.2: 1. After pushing a constant across an outer join --- ie, given "a LEFT JOIN b ON (a.x = b.y) WHERE a.x = 42", we can deduce that b.y is sort of equal to 42, in the sense that we needn't fetch any b rows where it isn't 42 --- loop to see if any additional deductions can be made. Previous releases did that by recursing, but I had mistakenly thought that this was no longer necessary given the EquivalenceClass machinery. 2. Allow pushing constants across outer join conditions even if the condition is outerjoin_delayed due to a lower outer join. This is safe as long as the condition is strict and we re-test it at the upper join. 3. Keep the outer-join clause even if we successfully push a constant across it. This is *necessary* in the outerjoin_delayed case, but even in the simple case, it seems better to do this to ensure that the join search order heuristics will consider the join as reasonable to make. Mark such a clause as having selectivity 1.0, though, since it's not going to eliminate very many rows after application of the constant condition. 4. Tweak have_relevant_eclass_joinclause to report that two relations are joinable when they have vars that are equated to the same constant. We won't actually generate any joinclause from such an EquivalenceClass, but again it seems that in such a case it's a good idea to consider the join as worth costing out. 5. Fix a bug in select_mergejoin_clauses that was exposed by these changes: we have to reject candidate mergejoin clauses if either side was equated to a constant, because we can't construct a canonical pathkey list for such a clause. This is an implementation restriction that might be worth fixing someday, but it doesn't seem critical to get it done for 8.3.
2008-01-09 21:42:29 +01:00
relids = bms_copy(qualscope);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* again, can't mark pseudoconstant */
}
else
{
Fix some issues with wrong placement of pseudo-constant quals. initsplan.c figured that it could push Var-free qual clauses to the top of the current JoinDomain, which is okay in the abstract. But if the current domain is inside some outer join, and we later commute an inside-the-domain outer join with one outside it, we end up placing the pushed-up qual clause incorrectly. In distribute_qual_to_rels, avoid this by using the syntactic scope of the qual clause; with the exception that if we're in the top-level join domain we can still use the full query relid set, ensuring the resulting gating Result node goes to the top of the plan. (This is approximately as smart as the pre-v16 code was. Perhaps we can do better later, but it's not clear that such cases are worth a lot of sweat.) In process_implied_equality, we don't have a clear notion of syntactic scope, but we do have the results of SpecialJoinInfo construction. Thumb through those and remove any lower outer joins that might get commuted to above the join domain. Again, we can make an exception for the top-level join domain. It'd be possible to work harder here (for example, by keeping outer joins that aren't shown as potentially commutable), but I'm going to stop here for the moment. This issue has convinced me that the current representation of join domains probably needs further refinement, so I'm disinclined to write inessential dependent logic just yet. In passing, tighten the qualscope passed to process_implied_equality by generate_base_implied_equalities_no_const; there's no need for it to be larger than the rel we are currently considering. Tom Lane and Richard Guo, per report from Tender Wang. Discussion: https://postgr.es/m/CAHewXNk9eJ35ru5xATWioTV4+xZPHptjy9etdcNPjUfY9RQ+uQ@mail.gmail.com
2023-02-22 18:39:07 +01:00
/*
* If we are in the top-level join domain, we can push the qual to
* the top of the plan tree. Otherwise, be conservative and eval
* it at original syntactic level. (Ideally we'd push it to the
* top of the current join domain in all cases, but that causes
* problems if we later rearrange outer-join evaluation order.
* Pseudoconstant quals below the top level are a pretty odd case,
* so it's not clear that it's worth working hard on.)
*/
if (jtitem->jdomain == (JoinDomain *) linitial(root->join_domains))
relids = bms_copy(jtitem->jdomain->jd_relids);
else
relids = bms_copy(qualscope);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* mark as gating qual */
pseudoconstant = true;
/* tell createplan.c to check for gating quals */
root->hasPseudoConstantQuals = true;
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
}
}
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
Adjust the definition of is_pushed_down so that it's always true for INNER JOIN quals, just like WHERE quals, even if they reference every one of the join's relations. Now that we can reorder outer and inner joins, it's possible for such a qual to end up being assigned to an outer join plan node, and we mustn't have it treated as a join qual rather than a filter qual for the node. (If it were, the join could produce null-extended rows that it shouldn't.) Per bug report from Pelle Johansson.
2007-02-16 21:57:19 +01:00
/*----------
Be smart about outer-join qualifications that mention only one side of the join, per recent discussion on pgsql-sql. Not clear that this will come up often in real queries, but it's not any more expensive to do it right, so we may as well do it right.
2003-03-03 00:46:34 +01:00
* Check to see if clause application must be delayed by outer-join
* considerations.
Adjust the definition of is_pushed_down so that it's always true for INNER JOIN quals, just like WHERE quals, even if they reference every one of the join's relations. Now that we can reorder outer and inner joins, it's possible for such a qual to end up being assigned to an outer join plan node, and we mustn't have it treated as a join qual rather than a filter qual for the node. (If it were, the join could produce null-extended rows that it shouldn't.) Per bug report from Pelle Johansson.
2007-02-16 21:57:19 +01:00
*
* A word about is_pushed_down: we mark the qual as "pushed down" if
* it is (potentially) applicable at a level different from its original
* syntactic level. This flag is used to distinguish OUTER JOIN ON quals
* from other quals pushed down to the same joinrel. The rules are:
* WHERE quals and INNER JOIN quals: is_pushed_down = true.
* Non-degenerate OUTER JOIN quals: is_pushed_down = false.
* Degenerate OUTER JOIN quals: is_pushed_down = true.
* A "degenerate" OUTER JOIN qual is one that doesn't mention the
* non-nullable side, and hence can be pushed down into the nullable side
* without changing the join result. It is correct to treat it as a
* regular filter condition at the level where it is evaluated.
*
* Note: it is not immediately obvious that a simple boolean is enough
* for this: if for some reason we were to attach a degenerate qual to
* its original join level, it would need to be treated as an outer join
* qual there. However, this cannot happen, because all the rels the
* clause mentions must be in the outer join's min_righthand, therefore
* the join it needs must be formed before the outer join; and we always
* attach quals to the lowest level where they can be evaluated. But
* if we were ever to re-introduce a mechanism for delaying evaluation
* of "expensive" quals, this area would need work.
Change more places to be less trusting of RestrictInfo.is_pushed_down. On further reflection, commit e5d83995e didn't go far enough: pretty much everywhere in the planner that examines a clause's is_pushed_down flag ought to be changed to use the more complicated behavior where we also check the clause's required_relids. Otherwise we could make incorrect decisions about whether, say, a clause is safe to use as a hash clause. Some (many?) of these places are safe as-is, either because they are never reached while considering a parameterized path, or because there are additional checks that would reject a pushed-down clause anyway. However, it seems smarter to just code them all the same way rather than rely on easily-broken reasoning of that sort. In support of that, invent a new macro RINFO_IS_PUSHED_DOWN that should be used in place of direct tests on the is_pushed_down flag. Like the previous patch, back-patch to all supported branches. Discussion: https://postgr.es/m/f8128b11-c5bf-3539-48cd-234178b2314d@proxel.se
2018-04-20 21:19:16 +02:00
*
* Note: generally, use of is_pushed_down has to go through the macro
* RINFO_IS_PUSHED_DOWN, because that flag alone is not always sufficient
* to tell whether a clause must be treated as pushed-down in context.
* This seems like another reason why it should perhaps be rethought.
Adjust the definition of is_pushed_down so that it's always true for INNER JOIN quals, just like WHERE quals, even if they reference every one of the join's relations. Now that we can reorder outer and inner joins, it's possible for such a qual to end up being assigned to an outer join plan node, and we mustn't have it treated as a join qual rather than a filter qual for the node. (If it were, the join could produce null-extended rows that it shouldn't.) Per bug report from Pelle Johansson.
2007-02-16 21:57:19 +01:00
*----------
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
*/
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
if (bms_overlap(relids, outerjoin_nonnullable))
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
{
Be smart about outer-join qualifications that mention only one side of the join, per recent discussion on pgsql-sql. Not clear that this will come up often in real queries, but it's not any more expensive to do it right, so we may as well do it right.
2003-03-03 00:46:34 +01:00
/*
* The qual is attached to an outer join and mentions (some of the)
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* rels on the nonnullable side, so it's not degenerate. If the
* caller wants to postpone handling such clauses, just add it to
* postponed_oj_qual_list and return. (The work we've done up to here
* will have to be redone later, but there's not much of it.)
*/
if (postponed_oj_qual_list != NULL)
{
*postponed_oj_qual_list = lappend(*postponed_oj_qual_list, clause);
return;
}
/*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* We can't use such a clause to deduce equivalence (the left and
* right sides might be unequal above the join because one of them has
* gone to NULL) ... but we might be able to use it for more limited
Fix some planner issues found while investigating Kevin Grittner's report of poorer planning in 8.3 than 8.2: 1. After pushing a constant across an outer join --- ie, given "a LEFT JOIN b ON (a.x = b.y) WHERE a.x = 42", we can deduce that b.y is sort of equal to 42, in the sense that we needn't fetch any b rows where it isn't 42 --- loop to see if any additional deductions can be made. Previous releases did that by recursing, but I had mistakenly thought that this was no longer necessary given the EquivalenceClass machinery. 2. Allow pushing constants across outer join conditions even if the condition is outerjoin_delayed due to a lower outer join. This is safe as long as the condition is strict and we re-test it at the upper join. 3. Keep the outer-join clause even if we successfully push a constant across it. This is *necessary* in the outerjoin_delayed case, but even in the simple case, it seems better to do this to ensure that the join search order heuristics will consider the join as reasonable to make. Mark such a clause as having selectivity 1.0, though, since it's not going to eliminate very many rows after application of the constant condition. 4. Tweak have_relevant_eclass_joinclause to report that two relations are joinable when they have vars that are equated to the same constant. We won't actually generate any joinclause from such an EquivalenceClass, but again it seems that in such a case it's a good idea to consider the join as worth costing out. 5. Fix a bug in select_mergejoin_clauses that was exposed by these changes: we have to reject candidate mergejoin clauses if either side was equated to a constant, because we can't construct a canonical pathkey list for such a clause. This is an implementation restriction that might be worth fixing someday, but it doesn't seem critical to get it done for 8.3.
2008-01-09 21:42:29 +01:00
* deductions, if it is mergejoinable. So consider adding it to the
* lists of set-aside outer-join clauses.
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
*/
Fix some planner issues found while investigating Kevin Grittner's report of poorer planning in 8.3 than 8.2: 1. After pushing a constant across an outer join --- ie, given "a LEFT JOIN b ON (a.x = b.y) WHERE a.x = 42", we can deduce that b.y is sort of equal to 42, in the sense that we needn't fetch any b rows where it isn't 42 --- loop to see if any additional deductions can be made. Previous releases did that by recursing, but I had mistakenly thought that this was no longer necessary given the EquivalenceClass machinery. 2. Allow pushing constants across outer join conditions even if the condition is outerjoin_delayed due to a lower outer join. This is safe as long as the condition is strict and we re-test it at the upper join. 3. Keep the outer-join clause even if we successfully push a constant across it. This is *necessary* in the outerjoin_delayed case, but even in the simple case, it seems better to do this to ensure that the join search order heuristics will consider the join as reasonable to make. Mark such a clause as having selectivity 1.0, though, since it's not going to eliminate very many rows after application of the constant condition. 4. Tweak have_relevant_eclass_joinclause to report that two relations are joinable when they have vars that are equated to the same constant. We won't actually generate any joinclause from such an EquivalenceClass, but again it seems that in such a case it's a good idea to consider the join as worth costing out. 5. Fix a bug in select_mergejoin_clauses that was exposed by these changes: we have to reject candidate mergejoin clauses if either side was equated to a constant, because we can't construct a canonical pathkey list for such a clause. This is an implementation restriction that might be worth fixing someday, but it doesn't seem critical to get it done for 8.3.
2008-01-09 21:42:29 +01:00
is_pushed_down = false;
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
maybe_equivalence = false;
Fix some planner issues found while investigating Kevin Grittner's report of poorer planning in 8.3 than 8.2: 1. After pushing a constant across an outer join --- ie, given "a LEFT JOIN b ON (a.x = b.y) WHERE a.x = 42", we can deduce that b.y is sort of equal to 42, in the sense that we needn't fetch any b rows where it isn't 42 --- loop to see if any additional deductions can be made. Previous releases did that by recursing, but I had mistakenly thought that this was no longer necessary given the EquivalenceClass machinery. 2. Allow pushing constants across outer join conditions even if the condition is outerjoin_delayed due to a lower outer join. This is safe as long as the condition is strict and we re-test it at the upper join. 3. Keep the outer-join clause even if we successfully push a constant across it. This is *necessary* in the outerjoin_delayed case, but even in the simple case, it seems better to do this to ensure that the join search order heuristics will consider the join as reasonable to make. Mark such a clause as having selectivity 1.0, though, since it's not going to eliminate very many rows after application of the constant condition. 4. Tweak have_relevant_eclass_joinclause to report that two relations are joinable when they have vars that are equated to the same constant. We won't actually generate any joinclause from such an EquivalenceClass, but again it seems that in such a case it's a good idea to consider the join as worth costing out. 5. Fix a bug in select_mergejoin_clauses that was exposed by these changes: we have to reject candidate mergejoin clauses if either side was equated to a constant, because we can't construct a canonical pathkey list for such a clause. This is an implementation restriction that might be worth fixing someday, but it doesn't seem critical to get it done for 8.3.
2008-01-09 21:42:29 +01:00
maybe_outer_join = true;
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/*
* Now force the qual to be evaluated exactly at the level of joining
* corresponding to the outer join. We cannot let it get pushed down
* into the nonnullable side, since then we'd produce no output rows,
* rather than the intended single null-extended row, for any
* nonnullable-side rows failing the qual.
Be smart about outer-join qualifications that mention only one side of the join, per recent discussion on pgsql-sql. Not clear that this will come up often in real queries, but it's not any more expensive to do it right, so we may as well do it right.
2003-03-03 00:46:34 +01:00
*/
Teach planner how to rearrange join order for some classes of OUTER JOIN. Per my recent proposal. I ended up basing the implementation on the existing mechanism for enforcing valid join orders of IN joins --- the rules for valid outer-join orders are somewhat similar.
2005-12-20 03:30:36 +01:00
Assert(ojscope);
relids = ojscope;
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
Assert(!pseudoconstant);
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
else
{
Adjust the definition of is_pushed_down so that it's always true for INNER JOIN quals, just like WHERE quals, even if they reference every one of the join's relations. Now that we can reorder outer and inner joins, it's possible for such a qual to end up being assigned to an outer join plan node, and we mustn't have it treated as a join qual rather than a filter qual for the node. (If it were, the join could produce null-extended rows that it shouldn't.) Per bug report from Pelle Johansson.
2007-02-16 21:57:19 +01:00
/*
* Normal qual clause or degenerate outer-join clause. Either way, we
* can mark it as pushed-down.
*/
is_pushed_down = true;
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
/*
* It's possible that this is an IS NULL clause that's redundant with
* a lower antijoin; if so we can just discard it. We need not test
* in any of the other cases, because this will only be possible for
* pushed-down clauses.
*/
if (check_redundant_nullability_qual(root, clause))
return;
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* Feed qual to the equivalence machinery, if allowed by caller */
maybe_equivalence = allow_equivalence;
pgindent run.
2003-08-04 02:43:34 +02:00
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/*
* Since it doesn't mention the LHS, it's certainly not useful as a
* set-aside OJ clause, even if it's in an OJ.
*/
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
maybe_outer_join = false;
First cut at full support for OUTER JOINs. There are still a few loose ends to clean up (see my message of same date to pghackers), but mostly it works. INITDB REQUIRED!
2000-09-12 23:07:18 +02:00
}
Adjust the definition of RestrictInfo's left_relids and right_relids fields: now they are valid whenever the clause is a binary opclause, not only when it is a potential join clause (there is a new boolean field canjoin to signal the latter condition). This lets us avoid recomputing the relid sets over and over while examining indexes. Still more work to do to make this as useful as it could be, because there are places that could use the info but don't have access to the RestrictInfo node.
2003-12-31 00:53:15 +01:00
/*
Rewrite OR indexscan processing to be more flexible. We can now for the first time generate an OR indexscan for a two-column index when the WHERE condition is like 'col1 = foo AND (col2 = bar OR col2 = baz)' --- before, the OR had to be on the first column of the index or we'd not notice the possibility of using it. Some progress towards extracting OR indexscans from subclauses of an OR that references multiple relations, too, although this code is #ifdef'd out because it needs more work.
2004-01-04 01:07:32 +01:00
* Build the RestrictInfo node itself.
Adjust the definition of RestrictInfo's left_relids and right_relids fields: now they are valid whenever the clause is a binary opclause, not only when it is a potential join clause (there is a new boolean field canjoin to signal the latter condition). This lets us avoid recomputing the relid sets over and over while examining indexes. Still more work to do to make this as useful as it could be, because there are places that could use the info but don't have access to the RestrictInfo node.
2003-12-31 00:53:15 +01:00
*/
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
restrictinfo = make_restrictinfo(root,
(Expr *) clause,
Add the ability to extract OR indexscan conditions from OR-of-AND join conditions in which each OR subclause includes a constraint on the same relation. This implements the other useful side-effect of conversion to CNF format, without its unpleasant side-effects. As per pghackers discussion of a few weeks ago.
2004-01-05 06:07:36 +01:00
is_pushed_down,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
has_clone,
is_clone,
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
pseudoconstant,
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
security_level,
Fix planner to restore its previous level of intelligence about pushing constants through full joins, as in select * from tenk1 a full join tenk1 b using (unique1) where unique1 = 42; which should generate a fairly cheap plan where we apply the constraint unique1 = 42 in each relation scan. This had been broken by my patch of 2008-06-27, which is now reverted in favor of a more invasive but hopefully less incorrect approach. That patch was meant to prevent incorrect extraction of OR'd indexclauses from OR conditions above an outer join. To do that correctly we need more information than the outerjoin_delay flag can provide, so add a nullable_relids field to RestrictInfo that records exactly which relations are nulled by outer joins that are underneath a particular qual clause. A side benefit is that we can make the test in create_or_index_quals more specific: it is now smart enough to extract an OR'd indexclause into the outer side of an outer join, even though it must not do so in the inner side. The old coding couldn't distinguish these cases so it could not do either.
2009-04-16 22:42:16 +02:00
relids,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
incompatible_relids,
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
outerjoin_nonnullable);
Adjust the definition of RestrictInfo's left_relids and right_relids fields: now they are valid whenever the clause is a binary opclause, not only when it is a potential join clause (there is a new boolean field canjoin to signal the latter condition). This lets us avoid recomputing the relid sets over and over while examining indexes. Still more work to do to make this as useful as it could be, because there are places that could use the info but don't have access to the RestrictInfo node.
2003-12-31 00:53:15 +01:00
Rewrite OR indexscan processing to be more flexible. We can now for the first time generate an OR indexscan for a two-column index when the WHERE condition is like 'col1 = foo AND (col2 = bar OR col2 = baz)' --- before, the OR had to be on the first column of the index or we'd not notice the possibility of using it. Some progress towards extracting OR indexscans from subclauses of an OR that references multiple relations, too, although this code is #ifdef'd out because it needs more work.
2004-01-04 01:07:32 +01:00
/*
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
* If it's a join clause, add vars used in the clause to targetlists of
* their relations, so that they will be emitted by the plan nodes that
* scan those relations (else they won't be available at the join node!).
*
* Normally we mark the vars as needed at the join identified by "relids".
* However, if this is a clone clause then ignore the outer-join relids in
* that set. Otherwise, vars appearing in a cloned clause would end up
* marked as having to propagate to the highest one of the commuting
* joins, which would often be an overestimate. For such clauses, correct
* var propagation is ensured by making ojscope include input rels from
* both sides of the join.
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
*
* Note: if the clause gets absorbed into an EquivalenceClass then this
* may be unnecessary, but for now we have to do it to cover the case
* where the EC becomes ec_broken and we end up reinserting the original
* clauses into the plan.
Rewrite OR indexscan processing to be more flexible. We can now for the first time generate an OR indexscan for a two-column index when the WHERE condition is like 'col1 = foo AND (col2 = bar OR col2 = baz)' --- before, the OR had to be on the first column of the index or we'd not notice the possibility of using it. Some progress towards extracting OR indexscans from subclauses of an OR that references multiple relations, too, although this code is #ifdef'd out because it needs more work.
2004-01-04 01:07:32 +01:00
*/
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
if (bms_membership(relids) == BMS_MULTIPLE)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Avoid listing ungrouped Vars in the targetlist of Agg-underneath-Window. Regular aggregate functions in combination with, or within the arguments of, window functions are OK per spec; they have the semantics that the aggregate output rows are computed and then we run the window functions over that row set. (Thus, this combination is not really useful unless there's a GROUP BY so that more than one aggregate output row is possible.) The case without GROUP BY could fail, as recently reported by Jeff Davis, because sloppy construction of the Agg node's targetlist resulted in extra references to possibly-ungrouped Vars appearing outside the aggregate function calls themselves. See the added regression test case for an example. Fixing this requires modifying the API of flatten_tlist and its underlying function pull_var_clause. I chose to make pull_var_clause's API for aggregates identical to what it was already doing for placeholders, since the useful behaviors turn out to be the same (error, report node as-is, or recurse into it). I also tightened the error checking in this area a bit: if it was ever valid to see an uplevel Var, Aggref, or PlaceHolderVar here, that was a long time ago, so complain instead of ignoring them. Backpatch into 9.1. The failure exists in 8.4 and 9.0 as well, but seeing that it only occurs in a basically-useless corner case, it doesn't seem worth the risks of changing a function API in a minor release. There might be third-party code using pull_var_clause.
2011-07-13 00:23:55 +02:00
List *vars = pull_var_clause(clause,
Refactor pull_var_clause's API to make it less tedious to extend. In commit 1d97c19a0f748e94 and later c1d9579dd8bf3c92, we extended pull_var_clause's API by adding enum-type arguments. That's sort of a pain to maintain, though, because it means every time we add a new behavior we must touch every last one of the call sites, even if there's a reasonable default behavior that most of them could use. Let's switch over to using a bitmask of flags, instead; that seems more maintainable and might save a nanosecond or two as well. This commit changes no behavior in itself, though I'm going to follow it up with one that does add a new behavior. In passing, remove flatten_tlist(), which has not been used since 9.1 and would otherwise need the same API changes. Removing these enums means that optimizer/tlist.h no longer needs to depend on optimizer/var.h. Changing that caused a number of C files to need addition of #include "optimizer/var.h" (probably we can thank old runs of pgrminclude for that); but on balance it seems like a good change anyway.
2016-03-10 21:52:58 +01:00
PVC_RECURSE_AGGREGATES |
Give pull_var_clause() reject/recurse/return behavior for WindowFuncs too. All along, this function should have treated WindowFuncs in a manner similar to Aggrefs, ie with an option whether or not to recurse into them. By not considering the case, it was always recursing, which is OK for most callers (although I suspect that the case in prepare_sort_from_pathkeys might represent a bug). But now we need return-without-recursing behavior as well. There are also more than a few callers that should never see a WindowFunc, and now we'll get some error checking on that.
2016-03-10 22:23:40 +01:00
PVC_RECURSE_WINDOWFUNCS |
Avoid listing ungrouped Vars in the targetlist of Agg-underneath-Window. Regular aggregate functions in combination with, or within the arguments of, window functions are OK per spec; they have the semantics that the aggregate output rows are computed and then we run the window functions over that row set. (Thus, this combination is not really useful unless there's a GROUP BY so that more than one aggregate output row is possible.) The case without GROUP BY could fail, as recently reported by Jeff Davis, because sloppy construction of the Agg node's targetlist resulted in extra references to possibly-ungrouped Vars appearing outside the aggregate function calls themselves. See the added regression test case for an example. Fixing this requires modifying the API of flatten_tlist and its underlying function pull_var_clause. I chose to make pull_var_clause's API for aggregates identical to what it was already doing for placeholders, since the useful behaviors turn out to be the same (error, report node as-is, or recurse into it). I also tightened the error checking in this area a bit: if it was ever valid to see an uplevel Var, Aggref, or PlaceHolderVar here, that was a long time ago, so complain instead of ignoring them. Backpatch into 9.1. The failure exists in 8.4 and 9.0 as well, but seeing that it only occurs in a basically-useless corner case, it doesn't seem worth the risks of changing a function API in a minor release. There might be third-party code using pull_var_clause.
2011-07-13 00:23:55 +02:00
PVC_INCLUDE_PLACEHOLDERS);
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
Relids where_needed;
Replace planner's representation of relation sets, per pghackers discussion. Instead of Lists of integers, we now store variable-length bitmap sets. This should be faster as well as less error-prone.
2003-02-08 21:20:55 +01:00
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
if (is_clone)
where_needed = bms_intersect(relids, root->all_baserels);
else
where_needed = relids;
add_vars_to_targetlist(root, vars, where_needed);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
list_free(vars);
Clean up handling of variable-free qual clauses. System now does the right thing with variable-free clauses that contain noncachable functions, such as 'WHERE random() < 0.5' --- these are evaluated once per potential output tuple. Expressions that contain only Params are now candidates to be indexscan quals --- for example, 'var = ($1 + 1)' can now be indexed. Cope with RelabelType nodes atop potential indexscan variables --- this oversight prevents 7.0.* from recognizing some potentially indexscanable situations.
2000-08-13 04:50:35 +02:00
}
New cost model for planning, incorporating a penalty for random page accesses versus sequential accesses, a (very crude) estimate of the effects of caching on random page accesses, and cost to evaluate WHERE- clause expressions. Export critical parameters for this model as SET variables. Also, create SET variables for the planner's enable flags (enable_seqscan, enable_indexscan, etc) so that these can be controlled more conveniently than via PGOPTIONS. Planner now estimates both startup cost (cost before retrieving first tuple) and total cost of each path, so it can optimize queries with LIMIT on a reasonable basis by interpolating between these costs. Same facility is a win for EXISTS(...) subqueries and some other cases. Redesign pathkey representation to achieve a major speedup in planning (I saw as much as 5X on a 10-way join); also minor changes in planner to reduce memory consumption by recycling discarded Path nodes and not constructing unnecessary lists. Minor cleanups to display more-plausible costs in some cases in EXPLAIN output. Initdb forced by change in interface to index cost estimation functions.
2000-02-15 21:49:31 +01:00
/*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* We check "mergejoinability" of every clause, not only join clauses,
* because we want to know about equivalences between vars of the same
* relation, or between vars and consts.
*/
check_mergejoinable(restrictinfo);
/*
* If it is a true equivalence clause, send it to the EquivalenceClass
* machinery. We do *not* attach it directly to any restriction or join
* lists. The EC code will propagate it to the appropriate places later.
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
*
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
* If the clause has a mergejoinable operator, yet isn't an equivalence
* because it is an outer-join clause, the EC code may still be able to do
* something with it. We add it to appropriate lists for further
* consideration later. Specifically:
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* If it is a left or right outer-join qualification that relates the two
* sides of the outer join (no funny business like leftvar1 = leftvar2 +
* rightvar), we add it to root->left_join_clauses or
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
* root->right_join_clauses according to which side the nonnullable
* variable appears on.
*
* If it is a full outer-join qualification, we add it to
* root->full_join_clauses. (Ideally we'd discard cases that aren't
* leftvar = rightvar, as we do for left/right joins, but this routine
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* doesn't have the info needed to do that; and the current usage of the
* full_join_clauses list doesn't require that, so it's not currently
* worth complicating this routine's API to make it possible.)
*
* If none of the above hold, pass it off to
* distribute_restrictinfo_to_rels().
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
*
* In all cases, it's important to initialize the left_ec and right_ec
* fields of a mergejoinable clause, so that all possibly mergejoinable
* expressions have representations in EquivalenceClasses. If
* process_equivalence is successful, it will take care of that;
* otherwise, we have to call initialize_mergeclause_eclasses to do it.
New cost model for planning, incorporating a penalty for random page accesses versus sequential accesses, a (very crude) estimate of the effects of caching on random page accesses, and cost to evaluate WHERE- clause expressions. Export critical parameters for this model as SET variables. Also, create SET variables for the planner's enable flags (enable_seqscan, enable_indexscan, etc) so that these can be controlled more conveniently than via PGOPTIONS. Planner now estimates both startup cost (cost before retrieving first tuple) and total cost of each path, so it can optimize queries with LIMIT on a reasonable basis by interpolating between these costs. Same facility is a win for EXISTS(...) subqueries and some other cases. Redesign pathkey representation to achieve a major speedup in planning (I saw as much as 5X on a 10-way join); also minor changes in planner to reduce memory consumption by recycling discarded Path nodes and not constructing unnecessary lists. Minor cleanups to display more-plausible costs in some cases in EXPLAIN output. Initdb forced by change in interface to index cost estimation functions.
2000-02-15 21:49:31 +01:00
*/
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
if (restrictinfo->mergeopfamilies)
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
{
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
if (maybe_equivalence)
{
Rethink treatment of "postponed" quals in deconstruct_jointree(). After pulling up LATERAL subqueries, we may have qual clauses that refer to relations outside their syntactic scope. Before doing any such pullup, prepjointree.c checks to make sure that it wouldn't create a semantically-invalid situation; but we leave it to deconstruct_jointree() to actually move these quals up the join tree to a place where they can be evaluated. In commit 2489d76c4, I (tgl) refactored deconstruct_jointree() in a way that caused assertion failures while moving such quals, because the new logic failed to distinguish "this jointree node is a parent of the source one" from "this jointree node is processed after the source one in depth-first order". Fix this, and at the same time reduce the overhead a bit, by getting rid of the common PostponedQual list and instead making each JoinTreeItem contain a list of quals that needed to be postponed to its level. We can help distribute_qual_to_rels find the appropriate JoinTreeItem efficiently by adding parent-item links to the JoinTreeItem data structure. This ends up being the same number of relid subset checks as the original (pre-bug) logic, but less list manipulation is required during multi-level postponements. Richard Guo and Tom Lane, per bug #17768 from Robins Tharakan. Discussion: https://postgr.es/m/17768-5ac8730ece54478f@postgresql.org
2023-02-04 18:45:53 +01:00
if (process_equivalence(root, &restrictinfo, jtitem->jdomain))
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
return;
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
/* EC rejected it, so set left_ec/right_ec the hard way ... */
Reduce "X = X" to "X IS NOT NULL", if it's easy to do so. If the operator is a strict btree equality operator, and X isn't volatile, then the clause must yield true for any non-null value of X, or null if X is null. At top level of a WHERE clause, we can ignore the distinction between false and null results, so it's valid to simplify the clause to "X IS NOT NULL". This is a useful improvement mainly because we'll get a far better selectivity estimate in most cases. Because such cases seldom arise in well-written queries, it is unappetizing to expend a lot of planner cycles looking for them ... but it turns out that there's a place we can shoehorn this in practically for free, because equivclass.c already has to detect and reject candidate equivalences of the form X = X. That doesn't catch every place that it would be valid to simplify to X IS NOT NULL, but it catches the typical case. Working harder doesn't seem justified. Patch by me, reviewed by Petr Jelinek Discussion: https://postgr.es/m/CAMjNa7cC4X9YR-vAJS-jSYCajhRDvJQnN7m2sLH1wLh-_Z2bsw@mail.gmail.com
2017-10-08 18:23:32 +02:00
if (restrictinfo->mergeopfamilies) /* EC might have changed this */
initialize_mergeclause_eclasses(root, restrictinfo);
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
/* ... and fall through to distribute_restrictinfo_to_rels */
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
}
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
else if (maybe_outer_join && restrictinfo->can_join)
{
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
/* we need to set up left_ec/right_ec the hard way */
initialize_mergeclause_eclasses(root, restrictinfo);
/* now see if it should go to any outer-join lists */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
Assert(sjinfo != NULL);
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
if (bms_is_subset(restrictinfo->left_relids,
outerjoin_nonnullable) &&
!bms_overlap(restrictinfo->right_relids,
outerjoin_nonnullable))
{
/* we have outervar = innervar */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
OuterJoinClauseInfo *ojcinfo = makeNode(OuterJoinClauseInfo);
ojcinfo->rinfo = restrictinfo;
ojcinfo->sjinfo = sjinfo;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
root->left_join_clauses = lappend(root->left_join_clauses,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
ojcinfo);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
return;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
}
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
if (bms_is_subset(restrictinfo->right_relids,
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
outerjoin_nonnullable) &&
!bms_overlap(restrictinfo->left_relids,
outerjoin_nonnullable))
{
/* we have innervar = outervar */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
OuterJoinClauseInfo *ojcinfo = makeNode(OuterJoinClauseInfo);
ojcinfo->rinfo = restrictinfo;
ojcinfo->sjinfo = sjinfo;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
root->right_join_clauses = lappend(root->right_join_clauses,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
ojcinfo);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
return;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
}
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
if (sjinfo->jointype == JOIN_FULL)
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
{
/* FULL JOIN (above tests cannot match in this case) */
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
OuterJoinClauseInfo *ojcinfo = makeNode(OuterJoinClauseInfo);
ojcinfo->rinfo = restrictinfo;
ojcinfo->sjinfo = sjinfo;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
root->full_join_clauses = lappend(root->full_join_clauses,
Make Vars be outer-join-aware. Traditionally we used the same Var struct to represent the value of a table column everywhere in parse and plan trees. This choice predates our support for SQL outer joins, and it's really a pretty bad idea with outer joins, because the Var's value can depend on where it is in the tree: it might go to NULL above an outer join. So expression nodes that are equal() per equalfuncs.c might not represent the same value, which is a huge correctness hazard for the planner. To improve this, decorate Var nodes with a bitmapset showing which outer joins (identified by RTE indexes) may have nulled them at the point in the parse tree where the Var appears. This allows us to trust that equal() Vars represent the same value. A certain amount of klugery is still needed to cope with cases where we re-order two outer joins, but it's possible to make it work without sacrificing that core principle. PlaceHolderVars receive similar decoration for the same reason. In the planner, we include these outer join bitmapsets into the relids that an expression is considered to depend on, and in consequence also add outer-join relids to the relids of join RelOptInfos. This allows us to correctly perceive whether an expression can be calculated above or below a particular outer join. This change affects FDWs that want to plan foreign joins. They *must* follow suit when labeling foreign joins in order to match with the core planner, but for many purposes (if postgres_fdw is any guide) they'd prefer to consider only base relations within the join. To support both requirements, redefine ForeignScan.fs_relids as base+OJ relids, and add a new field fs_base_relids that's set up by the core planner. Large though it is, this commit just does the minimum necessary to install the new mechanisms and get check-world passing again. Follow-up patches will perform some cleanup. (The README additions and comments mention some stuff that will appear in the follow-up.) Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:16:20 +01:00
ojcinfo);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
return;
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
}
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
/* nope, so fall through to distribute_restrictinfo_to_rels */
}
else
{
/* we still need to set up left_ec/right_ec */
initialize_mergeclause_eclasses(root, restrictinfo);
Teach planner about some cases where a restriction clause can be propagated inside an outer join. In particular, given LEFT JOIN ON (A = B) WHERE A = constant, we cannot conclude that B = constant at the top level (B might be null instead), but we can nonetheless put a restriction B = constant into the quals for B's relation, since no inner-side rows not meeting that condition can contribute to the final result. Similarly, given FULL JOIN USING (J) WHERE J = constant, we can't directly conclude that either input J variable = constant, but it's OK to push such quals into each input rel. Per recent gripe from Kim Bisgaard. Along the way, remove 'valid_everywhere' flag from RestrictInfo, as on closer analysis it was not being used for anything, and was defined backwards anyway.
2005-07-03 01:00:42 +02:00
}
}
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/* No EC special case applies, so push it into the clause lists */
distribute_restrictinfo_to_rels(root, restrictinfo);
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
/*
* check_redundant_nullability_qual
* Check to see if the qual is an IS NULL qual that is redundant with
* a lower JOIN_ANTI join.
*
* We want to suppress redundant IS NULL quals, not so much to save cycles
* as to avoid generating bogus selectivity estimates for them. So if
* redundancy is detected here, distribute_qual_to_rels() just throws away
* the qual.
*/
static bool
check_redundant_nullability_qual(PlannerInfo *root, Node *clause)
{
Var *forced_null_var;
ListCell *lc;
/* Check for IS NULL, and identify the Var forced to NULL */
forced_null_var = find_forced_null_var(clause);
if (forced_null_var == NULL)
return false;
/*
Simplify overcomplicated (and overly restrictive) test to see whether an IS NULL condition is rendered redundant by detection of an antijoin. If we know that a join is an antijoin, then *any* Var coming out of its righthand side must be NULL, not only the joining column(s). Also, it's still gonna be null after being passed up through higher joins, whether they're outer joins or not. I was misled by a faulty analogy to reduce_outer_joins() in the original coding. But consider select * from a left join b on a.x = b.y where b.y is null and b.z is null; The first IS NULL condition justifies deciding that the join is an antijoin (if the = is strict) and then the second one is just plain redundant.
2009-02-20 01:01:03 +01:00
* If the Var comes from the nullable side of a lower antijoin, the IS
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
* NULL condition is necessarily true. If it's not nulled by anything,
* there is no point in searching the join_info_list. Otherwise, we need
* to find out whether the nulling rel is an antijoin.
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
*/
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
if (forced_null_var->varnullingrels == NULL)
return false;
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
foreach(lc, root->join_info_list)
{
SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc);
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
/*
* This test will not succeed if sjinfo->ojrelid is zero, which is
* possible for an antijoin that was converted from a semijoin; but in
* such a case the Var couldn't have come from its nullable side.
*/
if (sjinfo->jointype == JOIN_ANTI && sjinfo->ojrelid != 0 &&
bms_is_member(sjinfo->ojrelid, forced_null_var->varnullingrels))
Simplify overcomplicated (and overly restrictive) test to see whether an IS NULL condition is rendered redundant by detection of an antijoin. If we know that a join is an antijoin, then *any* Var coming out of its righthand side must be NULL, not only the joining column(s). Also, it's still gonna be null after being passed up through higher joins, whether they're outer joins or not. I was misled by a faulty analogy to reduce_outer_joins() in the original coding. But consider select * from a left join b on a.x = b.y where b.y is null and b.z is null; The first IS NULL condition justifies deciding that the join is an antijoin (if the = is strict) and then the second one is just plain redundant.
2009-02-20 01:01:03 +01:00
return true;
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
}
Simplify overcomplicated (and overly restrictive) test to see whether an IS NULL condition is rendered redundant by detection of an antijoin. If we know that a join is an antijoin, then *any* Var coming out of its righthand side must be NULL, not only the joining column(s). Also, it's still gonna be null after being passed up through higher joins, whether they're outer joins or not. I was misled by a faulty analogy to reduce_outer_joins() in the original coding. But consider select * from a left join b on a.x = b.y where b.y is null and b.z is null; The first IS NULL condition justifies deciding that the join is an antijoin (if the = is strict) and then the second one is just plain redundant.
2009-02-20 01:01:03 +01:00
return false;
Implement SEMI and ANTI joins in the planner and executor. (Semijoins replace the old JOIN_IN code, but antijoins are new functionality.) Teach the planner to convert appropriate EXISTS and NOT EXISTS subqueries into semi and anti joins respectively. Also, LEFT JOINs with suitable upper-level IS NULL filters are recognized as being anti joins. Unify the InClauseInfo and OuterJoinInfo infrastructure into "SpecialJoinInfo". With that change, it becomes possible to associate a SpecialJoinInfo with every join attempt, which permits some cleanup of join selectivity estimation. That needs to be taken much further than this patch does, but the next step is to change the API for oprjoin selectivity functions, which seems like material for a separate patch. So for the moment the output size estimates for semi and especially anti joins are quite bogus.
2008-08-14 20:48:00 +02:00
}
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/*
* distribute_restrictinfo_to_rels
* Push a completed RestrictInfo into the proper restriction or join
* clause list(s).
*
* This is the last step of distribute_qual_to_rels() for ordinary qual
* clauses. Clauses that are interesting for equivalence-class processing
* are diverted to the EC machinery, but may ultimately get fed back here.
*/
void
distribute_restrictinfo_to_rels(PlannerInfo *root,
RestrictInfo *restrictinfo)
{
Relids relids = restrictinfo->required_relids;
RelOptInfo *rel;
switch (bms_membership(relids))
Deduce equality constraints that are implied by transitivity of mergejoinable qual clauses, and add them to the query quals. For example, WHERE a = b AND b = c will cause us to add AND a = c. This is necessary to ensure that it's safe to use these variables as interchangeable sort keys, which is something 7.0 knows how to do. Should provide a useful improvement in planning ability, too.
2000-07-24 05:11:01 +02:00
{
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
case BMS_SINGLETON:
/*
* There is only one relation participating in the clause, so it
* is a restriction clause for that relation.
*/
rel = find_base_rel(root, bms_singleton_member(relids));
/* Add clause to rel's restriction list */
rel->baserestrictinfo = lappend(rel->baserestrictinfo,
restrictinfo);
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
/* Update security level info */
rel->baserestrict_min_security = Min(rel->baserestrict_min_security,
restrictinfo->security_level);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
break;
case BMS_MULTIPLE:
/*
* The clause is a join clause, since there is more than one rel
* in its relid set.
*/
/*
* Check for hashjoinable operators. (We don't bother setting the
Support RIGHT and FULL OUTER JOIN in hash joins. This is advantageous first because it allows us to hash the smaller table regardless of the outer-join type, and second because hash join can be more flexible than merge join in dealing with arbitrary join quals in a FULL join. For merge join all the join quals have to be mergejoinable, but hash join will work so long as there's at least one hashjoinable qual --- the others can be any condition. (This is true essentially because we don't keep per-inner-tuple match flags in merge join, while hash join can do so.) To do this, we need a has-it-been-matched flag for each tuple in the hashtable, not just one for the current outer tuple. The key idea that makes this practical is that we can store the match flag in the tuple's infomask, since there are lots of bits there that are of no interest for a MinimalTuple. So we aren't increasing the size of the hashtable at all for the feature. To write this without turning the hash code into even more of a pile of spaghetti than it already was, I rewrote ExecHashJoin in a state-machine style, similar to ExecMergeJoin. Other than that decision, it was pretty straightforward.
2010-12-31 02:24:55 +01:00
* hashjoin info except in true join clauses.)
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
*/
Support RIGHT and FULL OUTER JOIN in hash joins. This is advantageous first because it allows us to hash the smaller table regardless of the outer-join type, and second because hash join can be more flexible than merge join in dealing with arbitrary join quals in a FULL join. For merge join all the join quals have to be mergejoinable, but hash join will work so long as there's at least one hashjoinable qual --- the others can be any condition. (This is true essentially because we don't keep per-inner-tuple match flags in merge join, while hash join can do so.) To do this, we need a has-it-been-matched flag for each tuple in the hashtable, not just one for the current outer tuple. The key idea that makes this practical is that we can store the match flag in the tuple's infomask, since there are lots of bits there that are of no interest for a MinimalTuple. So we aren't increasing the size of the hashtable at all for the feature. To write this without turning the hash code into even more of a pile of spaghetti than it already was, I rewrote ExecHashJoin in a state-machine style, similar to ExecMergeJoin. Other than that decision, it was pretty straightforward.
2010-12-31 02:24:55 +01:00
check_hashjoinable(restrictinfo);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
/*
* Likewise, check if the clause is suitable to be used with a
Change the name of the Result Cache node to Memoize "Result Cache" was never a great name for this node, but nobody managed to come up with another name that anyone liked enough. That was until David Johnston mentioned "Node Memoization", which Tom Lane revised to just "Memoize". People seem to like "Memoize", so let's do the rename. Reviewed-by: Justin Pryzby Discussion: https://postgr.es/m/20210708165145.GG1176@momjian.us Backpatch-through: 14, where Result Cache was introduced
2021-07-14 02:43:58 +02:00
* Memoize node to cache inner tuples during a parameterized
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
* nested loop.
*/
Change the name of the Result Cache node to Memoize "Result Cache" was never a great name for this node, but nobody managed to come up with another name that anyone liked enough. That was until David Johnston mentioned "Node Memoization", which Tom Lane revised to just "Memoize". People seem to like "Memoize", so let's do the rename. Reviewed-by: Justin Pryzby Discussion: https://postgr.es/m/20210708165145.GG1176@momjian.us Backpatch-through: 14, where Result Cache was introduced
2021-07-14 02:43:58 +02:00
check_memoizable(restrictinfo);
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/*
* Add clause to the join lists of all the relevant relations.
*/
add_join_clause_to_rels(root, restrictinfo, relids);
break;
default:
/*
* clause references no rels, and therefore we have no place to
* attach it. Shouldn't get here if callers are working properly.
*/
elog(ERROR, "cannot cope with variable-free clause");
break;
Deduce equality constraints that are implied by transitivity of mergejoinable qual clauses, and add them to the query quals. For example, WHERE a = b AND b = c will cause us to add AND a = c. This is necessary to ensure that it's safe to use these variables as interchangeable sort keys, which is something 7.0 knows how to do. Should provide a useful improvement in planning ability, too.
2000-07-24 05:11:01 +02:00
}
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
}
pgindent run. Make it all clean.
2001-03-22 05:01:46 +01:00
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/*
* process_implied_equality
* Create a restrictinfo item that says "item1 op item2", and push it
* into the appropriate lists. (In practice opno is always a btree
* equality operator.)
*
* "qualscope" is the nominal syntactic level to impute to the restrictinfo.
* This must contain at least all the rels used in the expressions, but it
* is used only to set the qual application level when both exprs are
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
* variable-free. (Hence, it should usually match the join domain in which
* the clause applies.) Otherwise the qual is applied at the lowest join
* level that provides all its variables.
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
*
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
* "security_level" is the security level to assign to the new restrictinfo.
*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* "both_const" indicates whether both items are known pseudo-constant;
* in this case it is worth applying eval_const_expressions() in case we
* can produce constant TRUE or constant FALSE. (Otherwise it's not,
* because the expressions went through eval_const_expressions already.)
*
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
* Returns the generated RestrictInfo, if any. The result will be NULL
* if both_const is true and we successfully reduced the clause to
* constant TRUE.
*
Fix planning of non-strict equivalence clauses above outer joins. If a potential equivalence clause references a variable from the nullable side of an outer join, the planner needs to take care that derived clauses are not pushed to below the outer join; else they may use the wrong value for the variable. (The problem arises only with non-strict clauses, since if an upper clause can be proven strict then the outer join will get simplified to a plain join.) The planner attempted to prevent this type of error by checking that potential equivalence clauses aren't outerjoin-delayed as a whole, but actually we have to check each side separately, since the two sides of the clause will get moved around separately if it's treated as an equivalence. Bugs of this type can be demonstrated as far back as 7.4, even though releases before 8.3 had only a very ad-hoc notion of equivalence clauses. In addition, we neglected to account for the possibility that such clauses might have nonempty nullable_relids even when not outerjoin-delayed; so the equivalence-class machinery lacked logic to compute correct nullable_relids values for clauses it constructs. This oversight was harmless before 9.2 because we were only using RestrictInfo.nullable_relids for OR clauses; but as of 9.2 it could result in pushing constructed equivalence clauses to incorrect places. (This accounts for bug #7604 from Bill MacArthur.) Fix the first problem by adding a new test check_equivalence_delay() in distribute_qual_to_rels, and fix the second one by adding code in equivclass.c and called functions to set correct nullable_relids for generated clauses. Although I believe the second part of this is not currently necessary before 9.2, I chose to back-patch it anyway, partly to keep the logic similar across branches and partly because it seems possible we might find other reasons why we need valid values of nullable_relids in the older branches. Add regression tests illustrating these problems. In 9.0 and up, also add test cases checking that we can push constants through outer joins, since we've broken that optimization before and I nearly broke it again with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
* Note: this function will copy item1 and item2, but it is caller's
* responsibility to make sure that the Relids parameters are fresh copies
* not shared with other uses.
*
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
* Note: we do not do initialize_mergeclause_eclasses() here. It is
* caller's responsibility that left_ec/right_ec be set as necessary.
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
*/
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
RestrictInfo *
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
process_implied_equality(PlannerInfo *root,
Oid opno,
Revise collation derivation method and expression-tree representation. All expression nodes now have an explicit output-collation field, unless they are known to only return a noncollatable data type (such as boolean or record). Also, nodes that can invoke collation-aware functions store a separate field that is the collation value to pass to the function. This avoids confusion that arises when a function has collatable inputs and noncollatable output type, or vice versa. Also, replace the parser's on-the-fly collation assignment method with a post-pass over the completed expression tree. This allows us to use a more complex (and hopefully more nearly spec-compliant) assignment rule without paying for it in extra storage in every expression node. Fix assorted bugs in the planner's handling of collations by making collation one of the defining properties of an EquivalenceClass and by converting CollateExprs into discardable RelabelType nodes during expression preprocessing.
2011-03-20 01:29:08 +01:00
Oid collation,
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
Expr *item1,
Expr *item2,
Relids qualscope,
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
Index security_level,
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
bool both_const)
{
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
RestrictInfo *restrictinfo;
Node *clause;
Relids relids;
bool pseudoconstant = false;
Deduce equality constraints that are implied by transitivity of mergejoinable qual clauses, and add them to the query quals. For example, WHERE a = b AND b = c will cause us to add AND a = c. This is necessary to ensure that it's safe to use these variables as interchangeable sort keys, which is something 7.0 knows how to do. Should provide a useful improvement in planning ability, too.
2000-07-24 05:11:01 +02:00
process_implied_equality must copy the substructure of the clauses it is generating, to avoid problems when subselects are involved. Per report from Damon Hart.
2004-02-27 22:42:00 +01:00
/*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* Build the new clause. Copy to ensure it shares no substructure with
* original (this is necessary in case there are subselects in there...)
process_implied_equality must copy the substructure of the clauses it is generating, to avoid problems when subselects are involved. Per report from Damon Hart.
2004-02-27 22:42:00 +01:00
*/
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
clause = (Node *) make_opclause(opno,
BOOLOID, /* opresulttype */
false, /* opretset */
copyObject(item1),
copyObject(item2),
InvalidOid,
collation);
Deduce equality constraints that are implied by transitivity of mergejoinable qual clauses, and add them to the query quals. For example, WHERE a = b AND b = c will cause us to add AND a = c. This is necessary to ensure that it's safe to use these variables as interchangeable sort keys, which is something 7.0 knows how to do. Should provide a useful improvement in planning ability, too.
2000-07-24 05:11:01 +02:00
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/* If both constant, try to reduce to a boolean constant. */
if (both_const)
{
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
clause = eval_const_expressions(root, clause);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
/* If we produced const TRUE, just drop the clause */
if (clause && IsA(clause, Const))
{
Const *cclause = (Const *) clause;
Assert(cclause->consttype == BOOLOID);
if (!cclause->constisnull && DatumGetBool(cclause->constvalue))
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
return NULL;
}
}
/*
* The rest of this is a very cut-down version of distribute_qual_to_rels.
* We can skip most of the work therein, but there are a couple of special
* cases we still have to handle.
*
* Retrieve all relids mentioned within the possibly-simplified clause.
*/
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
relids = pull_varnos(root, clause);
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
Assert(bms_is_subset(relids, qualscope));
/*
* If the clause is variable-free, our normal heuristic for pushing it
* down to just the mentioned rels doesn't work, because there are none.
Fix some issues with wrong placement of pseudo-constant quals. initsplan.c figured that it could push Var-free qual clauses to the top of the current JoinDomain, which is okay in the abstract. But if the current domain is inside some outer join, and we later commute an inside-the-domain outer join with one outside it, we end up placing the pushed-up qual clause incorrectly. In distribute_qual_to_rels, avoid this by using the syntactic scope of the qual clause; with the exception that if we're in the top-level join domain we can still use the full query relid set, ensuring the resulting gating Result node goes to the top of the plan. (This is approximately as smart as the pre-v16 code was. Perhaps we can do better later, but it's not clear that such cases are worth a lot of sweat.) In process_implied_equality, we don't have a clear notion of syntactic scope, but we do have the results of SpecialJoinInfo construction. Thumb through those and remove any lower outer joins that might get commuted to above the join domain. Again, we can make an exception for the top-level join domain. It'd be possible to work harder here (for example, by keeping outer joins that aren't shown as potentially commutable), but I'm going to stop here for the moment. This issue has convinced me that the current representation of join domains probably needs further refinement, so I'm disinclined to write inessential dependent logic just yet. In passing, tighten the qualscope passed to process_implied_equality by generate_base_implied_equalities_no_const; there's no need for it to be larger than the rel we are currently considering. Tom Lane and Richard Guo, per report from Tender Wang. Discussion: https://postgr.es/m/CAHewXNk9eJ35ru5xATWioTV4+xZPHptjy9etdcNPjUfY9RQ+uQ@mail.gmail.com
2023-02-22 18:39:07 +01:00
* Apply it as a gating qual at the appropriate level (see comments for
* get_join_domain_min_rels).
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
*/
if (bms_is_empty(relids))
{
Fix some issues with wrong placement of pseudo-constant quals. initsplan.c figured that it could push Var-free qual clauses to the top of the current JoinDomain, which is okay in the abstract. But if the current domain is inside some outer join, and we later commute an inside-the-domain outer join with one outside it, we end up placing the pushed-up qual clause incorrectly. In distribute_qual_to_rels, avoid this by using the syntactic scope of the qual clause; with the exception that if we're in the top-level join domain we can still use the full query relid set, ensuring the resulting gating Result node goes to the top of the plan. (This is approximately as smart as the pre-v16 code was. Perhaps we can do better later, but it's not clear that such cases are worth a lot of sweat.) In process_implied_equality, we don't have a clear notion of syntactic scope, but we do have the results of SpecialJoinInfo construction. Thumb through those and remove any lower outer joins that might get commuted to above the join domain. Again, we can make an exception for the top-level join domain. It'd be possible to work harder here (for example, by keeping outer joins that aren't shown as potentially commutable), but I'm going to stop here for the moment. This issue has convinced me that the current representation of join domains probably needs further refinement, so I'm disinclined to write inessential dependent logic just yet. In passing, tighten the qualscope passed to process_implied_equality by generate_base_implied_equalities_no_const; there's no need for it to be larger than the rel we are currently considering. Tom Lane and Richard Guo, per report from Tender Wang. Discussion: https://postgr.es/m/CAHewXNk9eJ35ru5xATWioTV4+xZPHptjy9etdcNPjUfY9RQ+uQ@mail.gmail.com
2023-02-22 18:39:07 +01:00
/* eval at join domain's safe level */
relids = get_join_domain_min_rels(root, qualscope);
Invent "join domains" to replace the below_outer_join hack. EquivalenceClasses are now understood as applying within a "join domain", which is a set of inner-joined relations (possibly underneath an outer join). We no longer need to treat an EC from below an outer join as a second-class citizen. I have hopes of eventually being able to treat outer-join clauses via EquivalenceClasses, by means of only applying deductions within the EC's join domain. There are still problems in the way of that, though, so for now the reconsider_outer_join_clause logic is still here. I haven't been able to get rid of RestrictInfo.is_pushed_down either, but I wonder if that could be recast using JoinDomains. I had to hack one test case in postgres_fdw.sql to make it still test what it was meant to, because postgres_fdw is inconsistent about how it deals with quals containing non-shippable expressions; see https://postgr.es/m/1691374.1671659838@sss.pgh.pa.us. That should be improved, but I don't think it's within the scope of this patch series. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:50:25 +01:00
/* mark as gating qual */
pseudoconstant = true;
/* tell createplan.c to check for gating quals */
root->hasPseudoConstantQuals = true;
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
}
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
/*
* Build the RestrictInfo node itself.
*/
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
restrictinfo = make_restrictinfo(root,
(Expr *) clause,
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
true, /* is_pushed_down */
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
false, /* !has_clone */
false, /* !is_clone */
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
pseudoconstant,
security_level,
relids,
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
NULL, /* incompatible_relids */
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
NULL); /* outer_relids */
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
/*
* If it's a join clause, add vars used in the clause to targetlists of
* their relations, so that they will be emitted by the plan nodes that
* scan those relations (else they won't be available at the join node!).
*
* Typically, we'd have already done this when the component expressions
* were first seen by distribute_qual_to_rels; but it is possible that
* some of the Vars could have missed having that done because they only
* appeared in single-relation clauses originally. So do it here for
* safety.
*/
if (bms_membership(relids) == BMS_MULTIPLE)
{
List *vars = pull_var_clause(clause,
PVC_RECURSE_AGGREGATES |
PVC_RECURSE_WINDOWFUNCS |
PVC_INCLUDE_PLACEHOLDERS);
Use an explicit state flag to control PlaceHolderInfo creation. Up to now, callers of find_placeholder_info() were required to pass a flag indicating if it's OK to make a new PlaceHolderInfo. That'd be fine if the callers had free choice, but they do not. Once we begin deconstruct_jointree() it's no longer OK to make more PHIs; while callers before that always want to create a PHI if it's not there already. So there's no freedom of action, only the opportunity to cause bugs by creating PHIs too late. Let's get rid of that in favor of adding a state flag PlannerInfo.placeholdersFrozen, which we can set at the point where it's no longer OK to make more PHIs. This patch also simplifies a couple of call sites that were using complicated logic to avoid calling find_placeholder_info() as much as possible. Now that that lookup is O(1) thanks to the previous commit, the extra bitmap manipulations are probably a net negative. Discussion: https://postgr.es/m/1405792.1660677844@sss.pgh.pa.us
2022-08-17 21:52:53 +02:00
add_vars_to_targetlist(root, vars, relids);
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
list_free(vars);
}
/*
* Check mergejoinability. This will usually succeed, since the op came
* from an EquivalenceClass; but we could have reduced the original clause
* to a constant.
*/
check_mergejoinable(restrictinfo);
/*
* Note we don't do initialize_mergeclause_eclasses(); the caller can
* handle that much more cheaply than we can. It's okay to call
* distribute_restrictinfo_to_rels() before that happens.
*/
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
/*
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
* Push the new clause into all the appropriate restrictinfo lists.
Subselects in FROM clause, per ISO syntax: FROM (SELECT ...) [AS] alias. (Don't forget that an alias is required.) Views reimplemented as expanding to subselect-in-FROM. Grouping, aggregates, DISTINCT in views actually work now (he says optimistically). No UNION support in subselects/views yet, but I have some ideas about that. Rule-related permissions checking moved out of rewriter and into executor. INITDB REQUIRED!
2000-09-29 20:21:41 +02:00
*/
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
distribute_restrictinfo_to_rels(root, restrictinfo);
return restrictinfo;
Add an at-least-marginally-plausible method of estimating the number of groups produced by GROUP BY. This improves the accuracy of planning estimates for grouped subselects, and is needed to check whether a hashed aggregation plan risks memory overflow.
2002-11-20 00:22:00 +01:00
}
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
/*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* build_implied_join_equality --- build a RestrictInfo for a derived equality
Modify planner's implied-equality-deduction code so that when a set of known-equal expressions includes any constant expressions (including Params from outer queries), we actively suppress any 'var = var' clauses that are or could be deduced from the set, generating only the deducible 'var = const' clauses instead. The idea here is to push down the restrictions implied by the equality set to base relations whenever possible. Once we have applied the 'var = const' clauses, the 'var = var' clauses are redundant, and should be suppressed both to save work at execution and to avoid double-counting restrictivity.
2003-01-24 04:58:44 +01:00
*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* This overlaps the functionality of process_implied_equality(), but we
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
* must not push the RestrictInfo into the joininfo tree.
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
*
Fix planning of non-strict equivalence clauses above outer joins. If a potential equivalence clause references a variable from the nullable side of an outer join, the planner needs to take care that derived clauses are not pushed to below the outer join; else they may use the wrong value for the variable. (The problem arises only with non-strict clauses, since if an upper clause can be proven strict then the outer join will get simplified to a plain join.) The planner attempted to prevent this type of error by checking that potential equivalence clauses aren't outerjoin-delayed as a whole, but actually we have to check each side separately, since the two sides of the clause will get moved around separately if it's treated as an equivalence. Bugs of this type can be demonstrated as far back as 7.4, even though releases before 8.3 had only a very ad-hoc notion of equivalence clauses. In addition, we neglected to account for the possibility that such clauses might have nonempty nullable_relids even when not outerjoin-delayed; so the equivalence-class machinery lacked logic to compute correct nullable_relids values for clauses it constructs. This oversight was harmless before 9.2 because we were only using RestrictInfo.nullable_relids for OR clauses; but as of 9.2 it could result in pushing constructed equivalence clauses to incorrect places. (This accounts for bug #7604 from Bill MacArthur.) Fix the first problem by adding a new test check_equivalence_delay() in distribute_qual_to_rels, and fix the second one by adding code in equivclass.c and called functions to set correct nullable_relids for generated clauses. Although I believe the second part of this is not currently necessary before 9.2, I chose to back-patch it anyway, partly to keep the logic similar across branches and partly because it seems possible we might find other reasons why we need valid values of nullable_relids in the older branches. Add regression tests illustrating these problems. In 9.0 and up, also add test cases checking that we can push constants through outer joins, since we've broken that optimization before and I nearly broke it again with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
* Note: this function will copy item1 and item2, but it is caller's
* responsibility to make sure that the Relids parameters are fresh copies
* not shared with other uses.
*
Avoid creation of useless EquivalenceClasses during planning. Zoltan Boszormenyi exhibited a test case in which planning time was dominated by construction of EquivalenceClasses and PathKeys that had no actual relevance to the query (and in fact got discarded immediately). This happened because we generated PathKeys describing the sort ordering of every index on every table in the query, and only after that checked to see if the sort ordering was relevant. The EC/PK construction code is O(N^2) in the number of ECs, which is all right for the intended number of such objects, but it gets out of hand if there are ECs for lots of irrelevant indexes. To fix, twiddle the handling of mergeclauses a little bit to ensure that every interesting EC is created before we begin path generation. (This doesn't cost anything --- in fact I think it's a bit cheaper than before --- since we always eventually created those ECs anyway.) Then, if an index column can't be found in any pre-existing EC, we know that that sort ordering is irrelevant for the query. Instead of creating a useless EC, we can just not build a pathkey for the index column in the first place. The index will still be considered if it's useful for non-order-related reasons, but we will think of its output as unsorted.
2010-10-29 17:52:16 +02:00
* Note: we do not do initialize_mergeclause_eclasses() here. It is
* caller's responsibility that left_ec/right_ec be set as necessary.
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
*/
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
RestrictInfo *
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
build_implied_join_equality(PlannerInfo *root,
Oid opno,
Revise collation derivation method and expression-tree representation. All expression nodes now have an explicit output-collation field, unless they are known to only return a noncollatable data type (such as boolean or record). Also, nodes that can invoke collation-aware functions store a separate field that is the collation value to pass to the function. This avoids confusion that arises when a function has collatable inputs and noncollatable output type, or vice versa. Also, replace the parser's on-the-fly collation assignment method with a post-pass over the completed expression tree. This allows us to use a more complex (and hopefully more nearly spec-compliant) assignment rule without paying for it in extra storage in every expression node. Fix assorted bugs in the planner's handling of collations by making collation one of the defining properties of an EquivalenceClass and by converting CollateExprs into discardable RelabelType nodes during expression preprocessing.
2011-03-20 01:29:08 +01:00
Oid collation,
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
Expr *item1,
Expr *item2,
Fix planning of non-strict equivalence clauses above outer joins. If a potential equivalence clause references a variable from the nullable side of an outer join, the planner needs to take care that derived clauses are not pushed to below the outer join; else they may use the wrong value for the variable. (The problem arises only with non-strict clauses, since if an upper clause can be proven strict then the outer join will get simplified to a plain join.) The planner attempted to prevent this type of error by checking that potential equivalence clauses aren't outerjoin-delayed as a whole, but actually we have to check each side separately, since the two sides of the clause will get moved around separately if it's treated as an equivalence. Bugs of this type can be demonstrated as far back as 7.4, even though releases before 8.3 had only a very ad-hoc notion of equivalence clauses. In addition, we neglected to account for the possibility that such clauses might have nonempty nullable_relids even when not outerjoin-delayed; so the equivalence-class machinery lacked logic to compute correct nullable_relids values for clauses it constructs. This oversight was harmless before 9.2 because we were only using RestrictInfo.nullable_relids for OR clauses; but as of 9.2 it could result in pushing constructed equivalence clauses to incorrect places. (This accounts for bug #7604 from Bill MacArthur.) Fix the first problem by adding a new test check_equivalence_delay() in distribute_qual_to_rels, and fix the second one by adding code in equivclass.c and called functions to set correct nullable_relids for generated clauses. Although I believe the second part of this is not currently necessary before 9.2, I chose to back-patch it anyway, partly to keep the logic similar across branches and partly because it seems possible we might find other reasons why we need valid values of nullable_relids in the older branches. Add regression tests illustrating these problems. In 9.0 and up, also add test cases checking that we can push constants through outer joins, since we've broken that optimization before and I nearly broke it again with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
Relids qualscope,
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
Index security_level)
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
{
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
RestrictInfo *restrictinfo;
Expr *clause;
pgindent run on all C files. Java run to follow. initdb/regression tests pass.
2001-10-25 07:50:21 +02:00
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
/*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* Build the new clause. Copy to ensure it shares no substructure with
* original (this is necessary in case there are subselects in there...)
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
*/
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
clause = make_opclause(opno,
BOOLOID, /* opresulttype */
false, /* opretset */
Cast result of copyObject() to correct type copyObject() is declared to return void *, which allows easily assigning the result independent of the input, but it loses all type checking. If the compiler supports typeof or something similar, cast the result to the input type. This creates a greater amount of type safety. In some cases, where the result is assigned to a generic type such as Node * or Expr *, new casts are now necessary, but in general casts are now unnecessary in the normal case and indicate that something unusual is happening. Reviewed-by: Mark Dilger <hornschnorter@gmail.com>
2017-03-09 21:18:59 +01:00
copyObject(item1),
copyObject(item2),
Revise collation derivation method and expression-tree representation. All expression nodes now have an explicit output-collation field, unless they are known to only return a noncollatable data type (such as boolean or record). Also, nodes that can invoke collation-aware functions store a separate field that is the collation value to pass to the function. This avoids confusion that arises when a function has collatable inputs and noncollatable output type, or vice versa. Also, replace the parser's on-the-fly collation assignment method with a post-pass over the completed expression tree. This allows us to use a more complex (and hopefully more nearly spec-compliant) assignment rule without paying for it in extra storage in every expression node. Fix assorted bugs in the planner's handling of collations by making collation one of the defining properties of an EquivalenceClass and by converting CollateExprs into discardable RelabelType nodes during expression preprocessing.
2011-03-20 01:29:08 +01:00
InvalidOid,
collation);
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
/*
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
* Build the RestrictInfo node itself.
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
*/
Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar. Previously, pull_varnos() took the relids of a PlaceHolderVar as being equal to the relids in its contents, but that fails to account for the possibility that we have to postpone evaluation of the PHV due to outer joins. This could result in a malformed plan. The known cases end up triggering the "failed to assign all NestLoopParams to plan nodes" sanity check in createplan.c, but other symptoms may be possible. The right value to use is the join level we actually intend to evaluate the PHV at. We can get that from the ph_eval_at field of the associated PlaceHolderInfo. However, there are some places that call pull_varnos() before the PlaceHolderInfos have been created; in that case, fall back to the conservative assumption that the PHV will be evaluated at its syntactic level. (In principle this might result in missing some legal optimization, but I'm not aware of any cases where it's an issue in practice.) Things are also a bit ticklish for calls occurring during deconstruct_jointree(), but AFAICS the ph_eval_at fields should have reached their final values by the time we need them. The main problem in making this work is that pull_varnos() has no way to get at the PlaceHolderInfos. We can fix that easily, if a bit tediously, in HEAD by passing it the planner "root" pointer. In the back branches that'd cause an unacceptable API/ABI break for extensions, so leave the existing entry points alone and add new ones with the additional parameter. (If an old entry point is called and encounters a PHV, it'll fall back to using the syntactic level, again possibly missing some valid optimization.) Back-patch to v12. The computation is surely also wrong before that, but it appears that we cannot reach a bad plan thanks to join order restrictions imposed on the subquery that the PlaceHolderVar came from. The error only became reachable when commit 4be058fe9 allowed trivial subqueries to be collapsed out completely, eliminating their join order restrictions. Per report from Stephan Springl. Discussion: https://postgr.es/m/171041.1610849523@sss.pgh.pa.us
2021-01-21 21:37:23 +01:00
restrictinfo = make_restrictinfo(root,
clause,
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
true, /* is_pushed_down */
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
false, /* !has_clone */
false, /* !is_clone */
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
false, /* pseudoconstant */
Improve RLS planning by marking individual quals with security levels. In an RLS query, we must ensure that security filter quals are evaluated before ordinary query quals, in case the latter contain "leaky" functions that could expose the contents of sensitive rows. The original implementation of RLS planning ensured this by pushing the scan of a secured table into a sub-query that it marked as a security-barrier view. Unfortunately this results in very inefficient plans in many cases, because the sub-query cannot be flattened and gets planned independently of the rest of the query. To fix, drop the use of sub-queries to enforce RLS qual order, and instead mark each qual (RestrictInfo) with a security_level field establishing its priority for evaluation. Quals must be evaluated in security_level order, except that "leakproof" quals can be allowed to go ahead of quals of lower security_level, if it's helpful to do so. This has to be enforced within the ordering of any one list of quals to be evaluated at a table scan node, and we also have to ensure that quals are not chosen for early evaluation (i.e., use as an index qual or TID scan qual) if they're not allowed to go ahead of other quals at the scan node. This is sufficient to fix the problem for RLS quals, since we only support RLS policies on simple tables and thus RLS quals will always exist at the table scan level only. Eventually these qual ordering rules should be enforced for join quals as well, which would permit improving planning for explicit security-barrier views; but that's a task for another patch. Note that FDWs would need to be aware of these rules --- and not, for example, send an insecure qual for remote execution --- but since we do not yet allow RLS policies on foreign tables, the case doesn't arise. This will need to be addressed before we can allow such policies. Patch by me, reviewed by Stephen Frost and Dean Rasheed. Discussion: https://postgr.es/m/8185.1477432701@sss.pgh.pa.us
2017-01-18 18:58:20 +01:00
security_level, /* security_level */
Fix planner to restore its previous level of intelligence about pushing constants through full joins, as in select * from tenk1 a full join tenk1 b using (unique1) where unique1 = 42; which should generate a fairly cheap plan where we apply the constraint unique1 = 42 in each relation scan. This had been broken by my patch of 2008-06-27, which is now reverted in favor of a more invasive but hopefully less incorrect approach. That patch was meant to prevent incorrect extraction of OR'd indexclauses from OR conditions above an outer join. To do that correctly we need more information than the outerjoin_delay flag can provide, so add a nullable_relids field to RestrictInfo that records exactly which relations are nulled by outer joins that are underneath a particular qual clause. A side benefit is that we can make the test in create_or_index_quals more specific: it is now smart enough to extract an OR'd indexclause into the outer side of an outer join, even though it must not do so in the inner side. The old coding couldn't distinguish these cases so it could not do either.
2009-04-16 22:42:16 +02:00
qualscope, /* required_relids */
Fix filtering of "cloned" outer-join quals some more. We've had multiple issues with the clause_is_computable_at logic that I introduced in 2489d76c4: it's been known to accept more than one clone of the same qual at the same plan node, and also to accept no clones at all. It's looking impractical to get it 100% right on the basis of the currently-stored information, so fix it by introducing a new RestrictInfo field "incompatible_relids" that explicitly shows which outer joins a given clone mustn't be pushed above. In principle we could populate this field in every RestrictInfo, but that would cost space and there doesn't presently seem to be a need for it in general. Also, while deconstruct_distribute_oj_quals can easily fill the field with the remaining members of the commutative join set that it's considering, computing it in the general case seems again pretty complicated. So for now, just fill it for clone quals. Along the way, fix a bug that may or may not be only latent: equivclass.c was generating replacement clauses with is_pushed_down and has_clone/is_clone markings that didn't match their required_relids. This led me to conclude that leaving the clone flags out of make_restrictinfo's purview wasn't such a great idea after all, so add them. Per report from Richard Guo. Discussion: https://postgr.es/m/CAMbWs48EYi_9-pSd0ORes1kTmTeAjT4Q3gu49hJtYCbSn2JyeA@mail.gmail.com
2023-05-25 16:28:33 +02:00
NULL, /* incompatible_relids */
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
NULL); /* outer_relids */
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
Support RIGHT and FULL OUTER JOIN in hash joins. This is advantageous first because it allows us to hash the smaller table regardless of the outer-join type, and second because hash join can be more flexible than merge join in dealing with arbitrary join quals in a FULL join. For merge join all the join quals have to be mergejoinable, but hash join will work so long as there's at least one hashjoinable qual --- the others can be any condition. (This is true essentially because we don't keep per-inner-tuple match flags in merge join, while hash join can do so.) To do this, we need a has-it-been-matched flag for each tuple in the hashtable, not just one for the current outer tuple. The key idea that makes this practical is that we can store the match flag in the tuple's infomask, since there are lots of bits there that are of no interest for a MinimalTuple. So we aren't increasing the size of the hashtable at all for the feature. To write this without turning the hash code into even more of a pile of spaghetti than it already was, I rewrote ExecHashJoin in a state-machine style, similar to ExecMergeJoin. Other than that decision, it was pretty straightforward.
2010-12-31 02:24:55 +01:00
/* Set mergejoinability/hashjoinability flags */
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
check_mergejoinable(restrictinfo);
Support RIGHT and FULL OUTER JOIN in hash joins. This is advantageous first because it allows us to hash the smaller table regardless of the outer-join type, and second because hash join can be more flexible than merge join in dealing with arbitrary join quals in a FULL join. For merge join all the join quals have to be mergejoinable, but hash join will work so long as there's at least one hashjoinable qual --- the others can be any condition. (This is true essentially because we don't keep per-inner-tuple match flags in merge join, while hash join can do so.) To do this, we need a has-it-been-matched flag for each tuple in the hashtable, not just one for the current outer tuple. The key idea that makes this practical is that we can store the match flag in the tuple's infomask, since there are lots of bits there that are of no interest for a MinimalTuple. So we aren't increasing the size of the hashtable at all for the feature. To write this without turning the hash code into even more of a pile of spaghetti than it already was, I rewrote ExecHashJoin in a state-machine style, similar to ExecMergeJoin. Other than that decision, it was pretty straightforward.
2010-12-31 02:24:55 +01:00
check_hashjoinable(restrictinfo);
Change the name of the Result Cache node to Memoize "Result Cache" was never a great name for this node, but nobody managed to come up with another name that anyone liked enough. That was until David Johnston mentioned "Node Memoization", which Tom Lane revised to just "Memoize". People seem to like "Memoize", so let's do the rename. Reviewed-by: Justin Pryzby Discussion: https://postgr.es/m/20210708165145.GG1176@momjian.us Backpatch-through: 14, where Result Cache was introduced
2021-07-14 02:43:58 +02:00
check_memoizable(restrictinfo);
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
return restrictinfo;
Extend code that deduces implied equality clauses to detect whether a clause being added to a particular restriction-clause list is redundant with those already in the list. This avoids useless work at runtime, and (perhaps more importantly) keeps the selectivity estimation routines from generating too-small estimates of numbers of output rows. Also some minor improvements in OPTIMIZER_DEBUG displays.
2001-10-18 18:11:42 +02:00
}
Fix some issues with wrong placement of pseudo-constant quals. initsplan.c figured that it could push Var-free qual clauses to the top of the current JoinDomain, which is okay in the abstract. But if the current domain is inside some outer join, and we later commute an inside-the-domain outer join with one outside it, we end up placing the pushed-up qual clause incorrectly. In distribute_qual_to_rels, avoid this by using the syntactic scope of the qual clause; with the exception that if we're in the top-level join domain we can still use the full query relid set, ensuring the resulting gating Result node goes to the top of the plan. (This is approximately as smart as the pre-v16 code was. Perhaps we can do better later, but it's not clear that such cases are worth a lot of sweat.) In process_implied_equality, we don't have a clear notion of syntactic scope, but we do have the results of SpecialJoinInfo construction. Thumb through those and remove any lower outer joins that might get commuted to above the join domain. Again, we can make an exception for the top-level join domain. It'd be possible to work harder here (for example, by keeping outer joins that aren't shown as potentially commutable), but I'm going to stop here for the moment. This issue has convinced me that the current representation of join domains probably needs further refinement, so I'm disinclined to write inessential dependent logic just yet. In passing, tighten the qualscope passed to process_implied_equality by generate_base_implied_equalities_no_const; there's no need for it to be larger than the rel we are currently considering. Tom Lane and Richard Guo, per report from Tender Wang. Discussion: https://postgr.es/m/CAHewXNk9eJ35ru5xATWioTV4+xZPHptjy9etdcNPjUfY9RQ+uQ@mail.gmail.com
2023-02-22 18:39:07 +01:00
/*
* get_join_domain_min_rels
* Identify the appropriate join level for derived quals belonging
* to the join domain with the given relids.
*
* When we derive a pseudoconstant (Var-free) clause from an EquivalenceClass,
* we'd ideally apply the clause at the top level of the EC's join domain.
* However, if there are any outer joins inside that domain that get commuted
* with joins outside it, that leads to not finding a correct place to apply
* the clause. Instead, remove any lower outer joins from the relid set,
* and apply the clause to just the remaining rels. This still results in a
* correct answer, since if the clause produces FALSE then the LHS of these
* joins will be empty leading to an empty join result.
*
* However, there's no need to remove outer joins if this is the top-level
* join domain of the query, since then there's nothing else to commute with.
*
* Note: it's tempting to use this in distribute_qual_to_rels where it's
* dealing with pseudoconstant quals; but we can't because the necessary
* SpecialJoinInfos aren't all formed at that point.
*
* The result is always freshly palloc'd; we do not modify domain_relids.
*/
static Relids
get_join_domain_min_rels(PlannerInfo *root, Relids domain_relids)
{
Relids result = bms_copy(domain_relids);
ListCell *lc;
/* Top-level join domain? */
if (bms_equal(result, root->all_query_rels))
return result;
/* Nope, look for lower outer joins that could potentially commute out */
foreach(lc, root->join_info_list)
{
SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc);
if (sjinfo->jointype == JOIN_LEFT &&
bms_is_member(sjinfo->ojrelid, result))
{
result = bms_del_member(result, sjinfo->ojrelid);
result = bms_del_members(result, sjinfo->syn_righthand);
}
}
return result;
}
Deduce equality constraints that are implied by transitivity of mergejoinable qual clauses, and add them to the query quals. For example, WHERE a = b AND b = c will cause us to add AND a = c. This is necessary to ensure that it's safe to use these variables as interchangeable sort keys, which is something 7.0 knows how to do. Should provide a useful improvement in planning ability, too.
2000-07-24 05:11:01 +02:00
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
/*
* match_foreign_keys_to_quals
* Match foreign-key constraints to equivalence classes and join quals
*
* The idea here is to see which query join conditions match equality
* constraints of a foreign-key relationship. For such join conditions,
* we can use the FK semantics to make selectivity estimates that are more
* reliable than estimating from statistics, especially for multiple-column
* FKs, where the normal assumption of independent conditions tends to fail.
*
* In this function we annotate the ForeignKeyOptInfos in root->fkey_list
* with info about which eclasses and join qual clauses they match, and
* discard any ForeignKeyOptInfos that are irrelevant for the query.
*/
void
match_foreign_keys_to_quals(PlannerInfo *root)
{
List *newlist = NIL;
ListCell *lc;
foreach(lc, root->fkey_list)
{
ForeignKeyOptInfo *fkinfo = (ForeignKeyOptInfo *) lfirst(lc);
Fix match_foreign_keys_to_quals for FKs linking to unused rtable entries. Since get_relation_foreign_keys doesn't try to determine whether RTEs are actually part of the query semantics, it might make FK info records linking to RTEs that won't have a RelOptInfo at all. Cope with that. Per bug #14219 from Andrew Gierth. Report: <20160629183338.1397.43514@wrigleys.postgresql.org>
2016-06-29 22:02:08 +02:00
RelOptInfo *con_rel;
RelOptInfo *ref_rel;
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
int colno;
Fix match_foreign_keys_to_quals for FKs linking to unused rtable entries. Since get_relation_foreign_keys doesn't try to determine whether RTEs are actually part of the query semantics, it might make FK info records linking to RTEs that won't have a RelOptInfo at all. Cope with that. Per bug #14219 from Andrew Gierth. Report: <20160629183338.1397.43514@wrigleys.postgresql.org>
2016-06-29 22:02:08 +02:00
/*
* Either relid might identify a rel that is in the query's rtable but
When removing a relation from the query, drop its RelOptInfo. In commit b78f6264e I opined that it was "too risky" to delete a relation's RelOptInfo from the planner's data structures when we have realized that we don't need to join to it; so instead we just marked it as a dead relation. In hindsight that judgment seems flawed: any subsequent access to such a dead relation is arguably a bug in itself, so leaving the RelOptInfo present just helps to mask bugs. Let's delete it instead, allowing removal of the whole notion of a "dead relation". So far as the regression tests can find, this requires no other code changes, except for one Assert in equivclass.c that was very dubiously not complaining about access to a dead rel. Discussion: https://postgr.es/m/229905.1676062220@sss.pgh.pa.us
2023-02-13 19:35:38 +01:00
* isn't referenced by the jointree, or has been removed by join
* removal, so that it won't have a RelOptInfo. Hence don't use
* find_base_rel() here. We can ignore such FKs.
Fix match_foreign_keys_to_quals for FKs linking to unused rtable entries. Since get_relation_foreign_keys doesn't try to determine whether RTEs are actually part of the query semantics, it might make FK info records linking to RTEs that won't have a RelOptInfo at all. Cope with that. Per bug #14219 from Andrew Gierth. Report: <20160629183338.1397.43514@wrigleys.postgresql.org>
2016-06-29 22:02:08 +02:00
*/
if (fkinfo->con_relid >= root->simple_rel_array_size ||
fkinfo->ref_relid >= root->simple_rel_array_size)
continue; /* just paranoia */
con_rel = root->simple_rel_array[fkinfo->con_relid];
if (con_rel == NULL)
continue;
ref_rel = root->simple_rel_array[fkinfo->ref_relid];
if (ref_rel == NULL)
continue;
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
/*
* Ignore FK unless both rels are baserels. This gets rid of FKs that
When removing a relation from the query, drop its RelOptInfo. In commit b78f6264e I opined that it was "too risky" to delete a relation's RelOptInfo from the planner's data structures when we have realized that we don't need to join to it; so instead we just marked it as a dead relation. In hindsight that judgment seems flawed: any subsequent access to such a dead relation is arguably a bug in itself, so leaving the RelOptInfo present just helps to mask bugs. Let's delete it instead, allowing removal of the whole notion of a "dead relation". So far as the regression tests can find, this requires no other code changes, except for one Assert in equivclass.c that was very dubiously not complaining about access to a dead rel. Discussion: https://postgr.es/m/229905.1676062220@sss.pgh.pa.us
2023-02-13 19:35:38 +01:00
* link to inheritance child rels (otherrels).
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
*/
if (con_rel->reloptkind != RELOPT_BASEREL ||
ref_rel->reloptkind != RELOPT_BASEREL)
continue;
/*
* Scan the columns and try to match them to eclasses and quals.
*
* Note: for simple inner joins, any match should be in an eclass.
* "Loose" quals that syntactically match an FK equality must have
* been rejected for EC status because they are outer-join quals or
Do assorted mop-up in the planner. Remove RestrictInfo.nullable_relids, along with a good deal of infrastructure that calculated it. One use-case for it was in join_clause_is_movable_to, but we can now replace that usage with a check to see if the clause's relids include any outer join that can null the target relation. The other use-case was in join_clause_is_movable_into, but that test can just be dropped entirely now that the clause's relids include outer joins. Furthermore, join_clause_is_movable_into should now be accurate enough that it will accept anything returned by generate_join_implied_equalities, so we can restore the Assert that was diked out in commit 95f4e59c3. Remove the outerjoin_delayed mechanism. We needed this before to prevent quals from getting evaluated below outer joins that should null some of their vars. Now that we consider varnullingrels while placing quals, that's taken care of automatically, so throw the whole thing away. Teach remove_useless_result_rtes to also remove useless FromExprs. Having done that, the delay_upper_joins flag serves no purpose any more and we can remove it, largely reverting 11086f2f2. Use constant TRUE for "dummy" clauses when throwing back outer joins. This improves on a hack I introduced in commit 6a6522529. If we have a left-join clause l.x = r.y, and a WHERE clause l.x = constant, we generate r.y = constant and then don't really have a need for the join clause. But we must throw the join clause back anyway after marking it redundant, so that the join search heuristics won't think this is a clauseless join and avoid it. That was a kluge introduced under time pressure, and after looking at it I thought of a better way: let's just introduce constant-TRUE "join clauses" instead, and get rid of them at the end. This improves the generated plans for such cases by not having to test a redundant join clause. We can also get rid of the ugly hack used to mark such clauses as redundant for selectivity estimation. Patch by me; thanks to Richard Guo for review. Discussion: https://postgr.es/m/830269.1656693747@sss.pgh.pa.us
2023-01-30 19:44:36 +01:00
* similar. We can still consider them to match the FK.
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
*/
for (colno = 0; colno < fkinfo->nkeys; colno++)
{
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
EquivalenceClass *ec;
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
AttrNumber con_attno,
ref_attno;
Oid fpeqop;
ListCell *lc2;
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
ec = match_eclasses_to_foreign_key_col(root, fkinfo, colno);
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
/* Don't bother looking for loose quals if we got an EC match */
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
if (ec != NULL)
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
{
fkinfo->nmatched_ec++;
Fix foreign-key selectivity estimation in the presence of constants. get_foreign_key_join_selectivity() looks for join clauses that equate the two sides of the FK constraint. However, if we have a query like "WHERE fktab.a = pktab.a and fktab.a = 1", it won't find any such join clause, because equivclass.c replaces the given clauses with "fktab.a = 1 and pktab.a = 1", which can be enforced at the scan level, leaving nothing to be done for column "a" at the join level. We can fix that expectation without much trouble, but then a new problem arises: applying the foreign-key-based selectivity rule produces a rowcount underestimate, because we're effectively double-counting the selectivity of the "fktab.a = 1" clause. So we have to cancel that selectivity out of the estimate. To fix, refactor process_implied_equality() so that it can pass back the new RestrictInfo to its callers in equivclass.c, allowing the generated "fktab.a = 1" clause to be saved in the EquivalenceClass's ec_derives list. Then it's not much trouble to dig out the relevant RestrictInfo when we need to adjust an FK selectivity estimate. (While at it, we can also remove the expensive use of initialize_mergeclause_eclasses() to set up the new RestrictInfo's left_ec and right_ec pointers. The equivclass.c code can set those basically for free.) This seems like clearly a bug fix, but I'm hesitant to back-patch it, first because there's some API/ABI risk for extensions and second because we're usually loath to destabilize plan choices in stable branches. Per report from Sigrid Ehrenreich. Discussion: https://postgr.es/m/1019549.1603770457@sss.pgh.pa.us Discussion: https://postgr.es/m/AM6PR02MB5287A0ADD936C1FA80973E72AB190@AM6PR02MB5287.eurprd02.prod.outlook.com
2020-10-28 16:15:47 +01:00
if (ec->ec_has_const)
fkinfo->nconst_ec++;
Restore foreign-key-aware estimation of join relation sizes. This patch provides a new implementation of the logic added by commit 137805f89 and later removed by 77ba61080. It differs from the original primarily in expending much less effort per joinrel in large queries, which it accomplishes by doing most of the matching work once per query not once per joinrel. Hopefully, it's also less buggy and better commented. The never-documented enable_fkey_estimates GUC remains gone. There remains work to be done to make the selectivity estimates account for nulls in FK referencing columns; but that was true of the original patch as well. We may be able to address this point later in beta. In the meantime, any error should be in the direction of overestimating rather than underestimating joinrel sizes, which seems like the direction we want to err in. Tomas Vondra and Tom Lane Discussion: <31041.1465069446@sss.pgh.pa.us>
2016-06-18 21:22:34 +02:00
continue;
}
/*
* Scan joininfo list for relevant clauses. Either rel's joininfo
* list would do equally well; we use con_rel's.
*/
con_attno = fkinfo->conkey[colno];
ref_attno = fkinfo->confkey[colno];
fpeqop = InvalidOid; /* we'll look this up only if needed */
foreach(lc2, con_rel->joininfo)
{
RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc2);
OpExpr *clause = (OpExpr *) rinfo->clause;
Var *leftvar;
Var *rightvar;
/* Only binary OpExprs are useful for consideration */
if (!IsA(clause, OpExpr) ||
list_length(clause->args) != 2)
continue;
leftvar = (Var *) get_leftop((Expr *) clause);
rightvar = (Var *) get_rightop((Expr *) clause);
/* Operands must be Vars, possibly with RelabelType */
while (leftvar && IsA(leftvar, RelabelType))
leftvar = (Var *) ((RelabelType *) leftvar)->arg;
if (!(leftvar && IsA(leftvar, Var)))
continue;
while (rightvar && IsA(rightvar, RelabelType))
rightvar = (Var *) ((RelabelType *) rightvar)->arg;
if (!(rightvar && IsA(rightvar, Var)))
continue;
/* Now try to match the vars to the current foreign key cols */
if (fkinfo->ref_relid == leftvar->varno &&
ref_attno == leftvar->varattno &&
fkinfo->con_relid == rightvar->varno &&
con_attno == rightvar->varattno)
{
/* Vars match, but is it the right operator? */
if (clause->opno == fkinfo->conpfeqop[colno])
{
fkinfo->rinfos[colno] = lappend(fkinfo->rinfos[colno],
rinfo);
fkinfo->nmatched_ri++;
}
}
else if (fkinfo->ref_relid == rightvar->varno &&
ref_attno == rightvar->varattno &&
fkinfo->con_relid == leftvar->varno &&
con_attno == leftvar->varattno)
{
/*
* Reverse match, must check commutator operator. Look it
* up if we didn't already. (In the worst case we might
* do multiple lookups here, but that would require an FK
* equality operator without commutator, which is
* unlikely.)
*/
if (!OidIsValid(fpeqop))
fpeqop = get_commutator(fkinfo->conpfeqop[colno]);
if (clause->opno == fpeqop)
{
fkinfo->rinfos[colno] = lappend(fkinfo->rinfos[colno],
rinfo);
fkinfo->nmatched_ri++;
}
}
}
/* If we found any matching loose quals, count col as matched */
if (fkinfo->rinfos[colno])
fkinfo->nmatched_rcols++;
}
/*
* Currently, we drop multicolumn FKs that aren't fully matched to the
* query. Later we might figure out how to derive some sort of
* estimate from them, in which case this test should be weakened to
* "if ((fkinfo->nmatched_ec + fkinfo->nmatched_rcols) > 0)".
*/
if ((fkinfo->nmatched_ec + fkinfo->nmatched_rcols) == fkinfo->nkeys)
newlist = lappend(newlist, fkinfo);
}
/* Replace fkey_list, thereby discarding any useless entries */
root->fkey_list = newlist;
}
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
/*****************************************************************************
*
New cost model for planning, incorporating a penalty for random page accesses versus sequential accesses, a (very crude) estimate of the effects of caching on random page accesses, and cost to evaluate WHERE- clause expressions. Export critical parameters for this model as SET variables. Also, create SET variables for the planner's enable flags (enable_seqscan, enable_indexscan, etc) so that these can be controlled more conveniently than via PGOPTIONS. Planner now estimates both startup cost (cost before retrieving first tuple) and total cost of each path, so it can optimize queries with LIMIT on a reasonable basis by interpolating between these costs. Same facility is a win for EXISTS(...) subqueries and some other cases. Redesign pathkey representation to achieve a major speedup in planning (I saw as much as 5X on a 10-way join); also minor changes in planner to reduce memory consumption by recycling discarded Path nodes and not constructing unnecessary lists. Minor cleanups to display more-plausible costs in some cases in EXPLAIN output. Initdb forced by change in interface to index cost estimation functions.
2000-02-15 21:49:31 +01:00
* CHECKS FOR MERGEJOINABLE AND HASHJOINABLE CLAUSES
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*
*****************************************************************************/
/*
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
* check_mergejoinable
* If the restrictinfo's clause is mergejoinable, set the mergejoin
* info fields in the restrictinfo.
*
* Currently, we support mergejoin for binary opclauses where
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
* the operator is a mergejoinable operator. The arguments can be
* anything --- as long as there are no volatile functions in them.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
static void
check_mergejoinable(RestrictInfo *restrictinfo)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
Expr *clause = restrictinfo->clause;
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
Oid opno;
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
Node *leftarg;
Massive commit to run PGINDENT on all *.c and *.h files.
1997-09-07 07:04:48 +02:00
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
if (restrictinfo->pseudoconstant)
return;
Phase 2 of read-only-plans project: restructure expression-tree nodes so that all executable expression nodes inherit from a common supertype Expr. This is somewhat of an exercise in code purity rather than any real functional advance, but getting rid of the extra Oper or Func node formerly used in each operator or function call should provide at least a little space and speed improvement. initdb forced by changes in stored-rules representation.
2002-12-12 16:49:42 +01:00
if (!is_opclause(clause))
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
return;
Use the new List API function names throughout the backend, and disable the list compatibility API by default. While doing this, I decided to keep the llast() macro around and introduce llast_int() and llast_oid() variants.
2004-05-31 01:40:41 +02:00
if (list_length(((OpExpr *) clause)->args) != 2)
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
return;
Fix a number of places that made faulty assumptions about what is_opclause will accept.
1999-02-15 02:06:59 +01:00
Phase 2 of read-only-plans project: restructure expression-tree nodes so that all executable expression nodes inherit from a common supertype Expr. This is somewhat of an exercise in code purity rather than any real functional advance, but getting rid of the extra Oper or Func node formerly used in each operator or function call should provide at least a little space and speed improvement. initdb forced by changes in stored-rules representation.
2002-12-12 16:49:42 +01:00
opno = ((OpExpr *) clause)->opno;
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
leftarg = linitial(((OpExpr *) clause)->args);
Fix a number of places that made faulty assumptions about what is_opclause will accept.
1999-02-15 02:06:59 +01:00
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
if (op_mergejoinable(opno, exprType(leftarg)) &&
Cache if PathTarget and RestrictInfos contain volatile functions Here we aim to reduce duplicate work done by contain_volatile_functions() by caching whether PathTargets and RestrictInfos contain any volatile functions the first time contain_volatile_functions() is called for them. Any future calls for these nodes just use the cached value rather than going to the trouble of recursively checking the sub-node all over again. Thanks to Tom Lane for the idea. Any locations in the code which make changes to a PathTarget or RestrictInfo which could change the outcome of the volatility check must change the cached value back to VOLATILITY_UNKNOWN again. contain_volatile_functions() is the only code in charge of setting the cache value to either VOLATILITY_VOLATILE or VOLATILITY_NOVOLATILE. Some existing code does benefit from this additional caching, however, this change is mainly aimed at an upcoming patch that must check for volatility during the join search. Repeated volatility checks in that case can become very expensive when the join search contains more than a few relations. Author: David Rowley Discussion: https://postgr.es/m/3795226.1614059027@sss.pgh.pa.us
2021-03-29 03:47:05 +02:00
!contain_volatile_functions((Node *) restrictinfo))
Refactor planner's pathkeys data structure to create a separate, explicit representation of equivalence classes of variables. This is an extensive rewrite, but it brings a number of benefits: * planner no longer fails in the presence of "incomplete" operator families that don't offer operators for every possible combination of datatypes. * avoid generating and then discarding redundant equality clauses. * remove bogus assumption that derived equalities always use operators named "=". * mergejoins can work with a variety of sort orders (e.g., descending) now, instead of tying each mergejoinable operator to exactly one sort order. * better recognition of redundant sort columns. * can make use of equalities appearing underneath an outer join.
2007-01-20 21:45:41 +01:00
restrictinfo->mergeopfamilies = get_mergejoin_opfamilies(opno);
/*
* Note: op_mergejoinable is just a hint; if we fail to find the operator
* in any btree opfamilies, mergeopfamilies remains NIL and so the clause
* is not treated as mergejoinable.
*/
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
/*
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
* check_hashjoinable
* If the restrictinfo's clause is hashjoinable, set the hashjoin
* info fields in the restrictinfo.
*
* Currently, we support hashjoin for binary opclauses where
Allow merge and hash joins to occur on arbitrary expressions (anything not containing a volatile function), rather than only on 'Var = Var' clauses as before. This makes it practical to do flatten_join_alias_vars at the start of planning, which in turn eliminates a bunch of klugery inside the planner to deal with alias vars. As a free side effect, we now detect implied equality of non-Var expressions; for example in SELECT ... WHERE a.x = b.y and b.y = 42 we will deduce a.x = 42 and use that as a restriction qual on a. Also, we can remove the restriction introduced 12/5/02 to prevent pullup of subqueries whose targetlists contain sublinks. Still TODO: make statistical estimation routines in selfuncs.c and costsize.c smarter about expressions that are more complex than plain Vars. The need for this is considerably greater now that we have to be able to estimate the suitability of merge and hash join techniques on such expressions.
2003-01-15 20:35:48 +01:00
* the operator is a hashjoinable operator. The arguments can be
* anything --- as long as there are no volatile functions in them.
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
*/
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
static void
check_hashjoinable(RestrictInfo *restrictinfo)
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
{
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
Expr *clause = restrictinfo->clause;
Oid opno;
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
Node *leftarg;
Fix a number of places that made faulty assumptions about what is_opclause will accept.
1999-02-15 02:06:59 +01:00
Revise the planner's handling of "pseudoconstant" WHERE clauses, that is clauses containing no variables and no volatile functions. Such a clause can be used as a one-time qual in a gating Result plan node, to suppress plan execution entirely when it is false. Even when the clause is true, putting it in a gating node wins by avoiding repeated evaluation of the clause. In previous PG releases, query_planner() would do this for pseudoconstant clauses appearing at the top level of the jointree, but there was no ability to generate a gating Result deeper in the plan tree. To fix it, get rid of the special case in query_planner(), and instead process pseudoconstant clauses through the normal RestrictInfo qual distribution mechanism. When a pseudoconstant clause is found attached to a path node in create_plan(), pull it out and generate a gating Result at that point. This requires special-casing pseudoconstants in selectivity estimation and cost_qual_eval, but on the whole it's pretty clean. It probably even makes the planner a bit faster than before for the normal case of no pseudoconstants, since removing pull_constant_clauses saves one useless traversal of the qual tree. Per gripe from Phil Frost.
2006-07-01 20:38:33 +02:00
if (restrictinfo->pseudoconstant)
return;
Phase 2 of read-only-plans project: restructure expression-tree nodes so that all executable expression nodes inherit from a common supertype Expr. This is somewhat of an exercise in code purity rather than any real functional advance, but getting rid of the extra Oper or Func node formerly used in each operator or function call should provide at least a little space and speed improvement. initdb forced by changes in stored-rules representation.
2002-12-12 16:49:42 +01:00
if (!is_opclause(clause))
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
return;
Use the new List API function names throughout the backend, and disable the list compatibility API by default. While doing this, I decided to keep the llast() macro around and introduce llast_int() and llast_oid() variants.
2004-05-31 01:40:41 +02:00
if (list_length(((OpExpr *) clause)->args) != 2)
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
return;
Phase 2 of read-only-plans project: restructure expression-tree nodes so that all executable expression nodes inherit from a common supertype Expr. This is somewhat of an exercise in code purity rather than any real functional advance, but getting rid of the extra Oper or Func node formerly used in each operator or function call should provide at least a little space and speed improvement. initdb forced by changes in stored-rules representation.
2002-12-12 16:49:42 +01:00
opno = ((OpExpr *) clause)->opno;
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
leftarg = linitial(((OpExpr *) clause)->args);
Fix a number of places that made faulty assumptions about what is_opclause will accept.
1999-02-15 02:06:59 +01:00
Provide hashing support for arrays. The core of this patch is hash_array() and associated typcache infrastructure, which works just about exactly like the existing support for array comparison. In addition I did some work to ensure that the planner won't think that an array type is hashable unless its element type is hashable, and similarly for sorting. This includes adding a datatype parameter to op_hashjoinable and op_mergejoinable, and adding an explicit "hashable" flag to SortGroupClause. The lack of a cross-check on the element type was a pre-existing bug in mergejoin support --- but it didn't matter so much before, because if you couldn't sort the element type there wasn't any good alternative to failing anyhow. Now that we have the alternative of hashing the array type, there are cases where we can avoid a failure by being picky at the planner stage, so it's time to be picky. The issue of exactly how to combine the per-element hash values to produce an array hash is still open for discussion, but the rest of this is pretty solid, so I'll commit it as-is.
2010-10-31 02:55:20 +01:00
if (op_hashjoinable(opno, exprType(leftarg)) &&
Cache if PathTarget and RestrictInfos contain volatile functions Here we aim to reduce duplicate work done by contain_volatile_functions() by caching whether PathTargets and RestrictInfos contain any volatile functions the first time contain_volatile_functions() is called for them. Any future calls for these nodes just use the cached value rather than going to the trouble of recursively checking the sub-node all over again. Thanks to Tom Lane for the idea. Any locations in the code which make changes to a PathTarget or RestrictInfo which could change the outcome of the volatility check must change the cached value back to VOLATILITY_UNKNOWN again. contain_volatile_functions() is the only code in charge of setting the cache value to either VOLATILITY_VOLATILE or VOLATILITY_NOVOLATILE. Some existing code does benefit from this additional caching, however, this change is mainly aimed at an upcoming patch that must check for volatility during the join search. Repeated volatility checks in that case can become very expensive when the join search contains more than a few relations. Author: David Rowley Discussion: https://postgr.es/m/3795226.1614059027@sss.pgh.pa.us
2021-03-29 03:47:05 +02:00
!contain_volatile_functions((Node *) restrictinfo))
Major planner/optimizer revision: get rid of PathOrder node type, store all ordering information in pathkeys lists (which are now lists of lists of PathKeyItem nodes, not just lists of lists of vars). This was a big win --- the code is smaller and IMHO more understandable than it was, even though it handles more cases. I believe the node changes will not force an initdb for anyone; planner nodes don't show up in stored rules.
1999-08-16 04:17:58 +02:00
restrictinfo->hashjoinoperator = opno;
Postgres95 1.01 Distribution - Virgin Sources
1996-07-09 08:22:35 +02:00
}
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
/*
Change the name of the Result Cache node to Memoize "Result Cache" was never a great name for this node, but nobody managed to come up with another name that anyone liked enough. That was until David Johnston mentioned "Node Memoization", which Tom Lane revised to just "Memoize". People seem to like "Memoize", so let's do the rename. Reviewed-by: Justin Pryzby Discussion: https://postgr.es/m/20210708165145.GG1176@momjian.us Backpatch-through: 14, where Result Cache was introduced
2021-07-14 02:43:58 +02:00
* check_memoizable
* If the restrictinfo's clause is suitable to be used for a Memoize node,
Fix various typos and incorrect/outdated name references Author: Alexander Lakhin Discussion: https://postgr.es/m/699beab4-a6ca-92c9-f152-f559caf6dc25@gmail.com
2023-04-19 03:50:33 +02:00
* set the left_hasheqoperator and right_hasheqoperator to the hash equality
Fix incorrect hash equality operator bug in Memoize In v14, because we don't have a field in RestrictInfo to cache both the left and right type's hash equality operator, we just restrict the scope of Memoize to only when the left and right types of a RestrictInfo are the same. In master we add another field to RestrictInfo and cache both hash equality operators. Reported-by: Jaime Casanova Author: David Rowley Discussion: https://postgr.es/m/20210929185544.GB24346%40ahch-to Backpatch-through: 14
2021-11-08 02:40:33 +01:00
* operator that will be needed during caching.
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
*/
static void
Change the name of the Result Cache node to Memoize "Result Cache" was never a great name for this node, but nobody managed to come up with another name that anyone liked enough. That was until David Johnston mentioned "Node Memoization", which Tom Lane revised to just "Memoize". People seem to like "Memoize", so let's do the rename. Reviewed-by: Justin Pryzby Discussion: https://postgr.es/m/20210708165145.GG1176@momjian.us Backpatch-through: 14, where Result Cache was introduced
2021-07-14 02:43:58 +02:00
check_memoizable(RestrictInfo *restrictinfo)
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
{
TypeCacheEntry *typentry;
Expr *clause = restrictinfo->clause;
Fix incorrect hash equality operator bug in Memoize In v14, because we don't have a field in RestrictInfo to cache both the left and right type's hash equality operator, we just restrict the scope of Memoize to only when the left and right types of a RestrictInfo are the same. In master we add another field to RestrictInfo and cache both hash equality operators. Reported-by: Jaime Casanova Author: David Rowley Discussion: https://postgr.es/m/20210929185544.GB24346%40ahch-to Backpatch-through: 14
2021-11-08 02:40:33 +01:00
Oid lefttype;
Oid righttype;
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
if (restrictinfo->pseudoconstant)
return;
if (!is_opclause(clause))
return;
if (list_length(((OpExpr *) clause)->args) != 2)
return;
Fix incorrect hash equality operator bug in Memoize In v14, because we don't have a field in RestrictInfo to cache both the left and right type's hash equality operator, we just restrict the scope of Memoize to only when the left and right types of a RestrictInfo are the same. In master we add another field to RestrictInfo and cache both hash equality operators. Reported-by: Jaime Casanova Author: David Rowley Discussion: https://postgr.es/m/20210929185544.GB24346%40ahch-to Backpatch-through: 14
2021-11-08 02:40:33 +01:00
lefttype = exprType(linitial(((OpExpr *) clause)->args));
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
Fix incorrect hash equality operator bug in Memoize In v14, because we don't have a field in RestrictInfo to cache both the left and right type's hash equality operator, we just restrict the scope of Memoize to only when the left and right types of a RestrictInfo are the same. In master we add another field to RestrictInfo and cache both hash equality operators. Reported-by: Jaime Casanova Author: David Rowley Discussion: https://postgr.es/m/20210929185544.GB24346%40ahch-to Backpatch-through: 14
2021-11-08 02:40:33 +01:00
typentry = lookup_type_cache(lefttype, TYPECACHE_HASH_PROC |
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
TYPECACHE_EQ_OPR);
Fix incorrect hash equality operator bug in Memoize In v14, because we don't have a field in RestrictInfo to cache both the left and right type's hash equality operator, we just restrict the scope of Memoize to only when the left and right types of a RestrictInfo are the same. In master we add another field to RestrictInfo and cache both hash equality operators. Reported-by: Jaime Casanova Author: David Rowley Discussion: https://postgr.es/m/20210929185544.GB24346%40ahch-to Backpatch-through: 14
2021-11-08 02:40:33 +01:00
if (OidIsValid(typentry->hash_proc) && OidIsValid(typentry->eq_opr))
restrictinfo->left_hasheqoperator = typentry->eq_opr;
righttype = exprType(lsecond(((OpExpr *) clause)->args));
/*
* Lookup the right type, unless it's the same as the left type, in which
* case typentry is already pointing to the required TypeCacheEntry.
*/
if (lefttype != righttype)
typentry = lookup_type_cache(righttype, TYPECACHE_HASH_PROC |
TYPECACHE_EQ_OPR);
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
Fix incorrect hash equality operator bug in Memoize In v14, because we don't have a field in RestrictInfo to cache both the left and right type's hash equality operator, we just restrict the scope of Memoize to only when the left and right types of a RestrictInfo are the same. In master we add another field to RestrictInfo and cache both hash equality operators. Reported-by: Jaime Casanova Author: David Rowley Discussion: https://postgr.es/m/20210929185544.GB24346%40ahch-to Backpatch-through: 14
2021-11-08 02:40:33 +01:00
if (OidIsValid(typentry->hash_proc) && OidIsValid(typentry->eq_opr))
restrictinfo->right_hasheqoperator = typentry->eq_opr;
Add Result Cache executor node (take 2) Here we add a new executor node type named "Result Cache". The planner can include this node type in the plan to have the executor cache the results from the inner side of parameterized nested loop joins. This allows caching of tuples for sets of parameters so that in the event that the node sees the same parameter values again, it can just return the cached tuples instead of rescanning the inner side of the join all over again. Internally, result cache uses a hash table in order to quickly find tuples that have been previously cached. For certain data sets, this can significantly improve the performance of joins. The best cases for using this new node type are for join problems where a large portion of the tuples from the inner side of the join have no join partner on the outer side of the join. In such cases, hash join would have to hash values that are never looked up, thus bloating the hash table and possibly causing it to multi-batch. Merge joins would have to skip over all of the unmatched rows. If we use a nested loop join with a result cache, then we only cache tuples that have at least one join partner on the outer side of the join. The benefits of using a parameterized nested loop with a result cache increase when there are fewer distinct values being looked up and the number of lookups of each value is large. Also, hash probes to lookup the cache can be much faster than the hash probe in a hash join as it's common that the result cache's hash table is much smaller than the hash join's due to result cache only caching useful tuples rather than all tuples from the inner side of the join. This variation in hash probe performance is more significant when the hash join's hash table no longer fits into the CPU's L3 cache, but the result cache's hash table does. The apparent "random" access of hash buckets with each hash probe can cause a poor L3 cache hit ratio for large hash tables. Smaller hash tables generally perform better. The hash table used for the cache limits itself to not exceeding work_mem * hash_mem_multiplier in size. We maintain a dlist of keys for this cache and when we're adding new tuples and realize we've exceeded the memory budget, we evict cache entries starting with the least recently used ones until we have enough memory to add the new tuples to the cache. For parameterized nested loop joins, we now consider using one of these result cache nodes in between the nested loop node and its inner node. We determine when this might be useful based on cost, which is primarily driven off of what the expected cache hit ratio will be. Estimating the cache hit ratio relies on having good distinct estimates on the nested loop's parameters. For now, the planner will only consider using a result cache for parameterized nested loop joins. This works for both normal joins and also for LATERAL type joins to subqueries. It is possible to use this new node for other uses in the future. For example, to cache results from correlated subqueries. However, that's not done here due to some difficulties obtaining a distinct estimation on the outer plan to calculate the estimated cache hit ratio. Currently we plan the inner plan before planning the outer plan so there is no good way to know if a result cache would be useful or not since we can't estimate the number of times the subplan will be called until the outer plan is generated. The functionality being added here is newly introducing a dependency on the return value of estimate_num_groups() during the join search. Previously, during the join search, we only ever needed to perform selectivity estimations. With this commit, we need to use estimate_num_groups() in order to estimate what the hit ratio on the result cache will be. In simple terms, if we expect 10 distinct values and we expect 1000 outer rows, then we'll estimate the hit ratio to be 99%. Since cache hits are very cheap compared to scanning the underlying nodes on the inner side of the nested loop join, then this will significantly reduce the planner's cost for the join. However, it's fairly easy to see here that things will go bad when estimate_num_groups() incorrectly returns a value that's significantly lower than the actual number of distinct values. If this happens then that may cause us to make use of a nested loop join with a result cache instead of some other join type, such as a merge or hash join. Our distinct estimations have been known to be a source of trouble in the past, so the extra reliance on them here could cause the planner to choose slower plans than it did previous to having this feature. Distinct estimations are also fairly hard to estimate accurately when several tables have been joined already or when a WHERE clause filters out a set of values that are correlated to the expressions we're estimating the number of distinct value for. For now, the costing we perform during query planning for result caches does put quite a bit of faith in the distinct estimations being accurate. When these are accurate then we should generally see faster execution times for plans containing a result cache. However, in the real world, we may find that we need to either change the costings to put less trust in the distinct estimations being accurate or perhaps even disable this feature by default. There's always an element of risk when we teach the query planner to do new tricks that it decides to use that new trick at the wrong time and causes a regression. Users may opt to get the old behavior by turning the feature off using the enable_resultcache GUC. Currently, this is enabled by default. It remains to be seen if we'll maintain that setting for the release. Additionally, the name "Result Cache" is the best name I could think of for this new node at the time I started writing the patch. Nobody seems to strongly dislike the name. A few people did suggest other names but no other name seemed to dominate in the brief discussion that there was about names. Let's allow the beta period to see if the current name pleases enough people. If there's some consensus on a better name, then we can change it before the release. Please see the 2nd discussion link below for the discussion on the "Result Cache" name. Author: David Rowley Reviewed-by: Andy Fan, Justin Pryzby, Zhihong Yu, Hou Zhijie Tested-By: Konstantin Knizhnik Discussion: https://postgr.es/m/CAApHDvrPcQyQdWERGYWx8J%2B2DLUNgXu%2BfOSbQ1UscxrunyXyrQ%40mail.gmail.com Discussion: https://postgr.es/m/CAApHDvq=yQXr5kqhRviT2RhNKwToaWr9JAN5t+5_PzhuRJ3wvg@mail.gmail.com
2021-04-02 03:10:56 +02:00
}