Discussion is still underway as to whether to revert the entire patch
that added this function, but that discussion may not conclude before
beta1. So, in the meantime, let's do at least this much.
David Rowley
Now that Paths have their own rows field, print that rather than
the parent relation's rowcount.
Show the relid sets associated with Paths using table names rather
than numbers; since this code is able to print simple Var references
using table names, it seems a bit silly that print_relids can't.
Print the cheapest_parameterized_paths list for a RelOptInfo, and
include information about a parameterized path's required_outer rels.
Noted while trying to use this feature to debug Alexander Kirkouski's
recent bug report.
We mustn't run generate_gather_paths() during add_paths_to_joinrel(),
because that function can be invoked multiple times for the same target
joinrel. Not only is it wasteful to build GatherPaths repeatedly, but
a later add_partial_path() could delete the partial path that a previously
created GatherPath depends on. Instead establish the convention that we
do generate_gather_paths() for a rel only just before set_cheapest().
The code was accidentally not broken for baserels, because as of today there
never is more than one partial path for a baserel. But that assumption
obviously has a pretty short half-life, so move the generate_gather_paths()
calls for those cases as well.
Also add some generic comments explaining how and why this all works.
Per fuzz testing by Andreas Seltenreich.
Report: <871t5pgwdt.fsf@credativ.de>
Given a three-or-more-way equivalence class, such as X.Y = Y.Y = Z.Z,
it was possible for the planner to omit one of the quals needed to
enforce that all members of the equivalence class are actually equal.
This only happened in the case of a parameterized join node for two
of the relations, that is a plan tree like
Nested Loop
-> Scan X
-> Nested Loop
-> Scan Y
-> Scan Z
Filter: Z.Z = X.X
The eclass machinery normally expects to apply X.X = Y.Y when those
two relations are joined, but in this shape of plan tree they aren't
joined until the top node --- and, if the lower nested loop is marked
as parameterized by X, the top node will assume that the relevant eclass
condition(s) got pushed down into the lower node. On the other hand,
the scan of Z assumes that it's only responsible for constraining Z.Z
to match any one of the other eclass members. So one or another of
the required quals sometimes fell between the cracks, depending on
whether consideration of the eclass in get_joinrel_parampathinfo()
for the lower nested loop chanced to generate X.X = Y.Y or X.X = Z.Z
as the appropriate constraint there. If it generated the latter,
it'd erroneously suppose that the Z scan would take care of matters.
To fix, force X.X = Y.Y to be generated and applied at that join node
when this case occurs.
This is *extremely* hard to hit in practice, because various planner
behaviors conspire to mask the problem; starting with the fact that the
planner doesn't really like to generate a parameterized plan of the
above shape. (It might have been impossible to hit it before we
tweaked things to allow this plan shape for star-schema cases.) Many
thanks to Alexander Kirkouski for submitting a reproducible test case.
The bug can be demonstrated in all branches back to 9.2 where parameterized
paths were introduced, so back-patch that far.
Change max_parallel_degree default from 0 to 2. It is possible that
this is not a good idea, or that we should go with 1 worker rather
than 2, but we won't find out without trying it. Along the way,
reword the documentation for max_parallel_degree a little bit to
hopefully make it more clear.
Discussion: 20160420174631.3qjjhpwsvvx5bau5@alap3.anarazel.de
Now indexes (but only B-tree for now) can contain "extra" column(s) which
doesn't participate in index structure, they are just stored in leaf
tuples. It allows to use index only scan by using single index instead
of two or more indexes.
Author: Anastasia Lubennikova with minor editorializing by me
Reviewers: David Rowley, Peter Geoghegan, Jeff Janes
The code that estimates what parallel degree should be uesd for the
scan of a relation is currently rather stupid, so add a parallel_degree
reloption that can be used to override the planner's rather limited
judgement.
Julien Rouhaud, reviewed by David Rowley, James Sewell, Amit Kapila,
and me. Some further hacking by me.
We still use replacement selection for the first run of the sort only
and only when the number of tuples is relatively small. Otherwise,
the first run, and subsequent runs in all cases, are produced using
quicksort. This tends to be faster except perhaps for very small
amounts of working memory.
Peter Geoghegan, reviewed by Tomas Vondra, Jeff Janes, Mithun Cy,
Greg Stark, and me.
In cases where joins use multiple columns we currently assess each join
separately causing gross mis-estimates for join cardinality.
This patch adds use of FK information for the first time into the
planner. When FKs are present and we have multi-column join information,
plan estimates will be drastically improved. Cases with multiple FKs
are handled, though partial matches are ignored currently.
Net effect is substantial performance improvements for joins in many
common cases. Additional planning time is isolated to cases that are
currently performing poorly, measured at 0.08 - 0.15 ms.
Please watch for planner performance regressions; circumstances seem
unlikely but the law of unintended consequences may apply somewhen.
Additional complex tests welcome to prove this before release.
Tests can be performed using SET enable_fkey_estimates = on | off
using scripts provided during Hackers discussions, message id:
552335D9.3090707@2ndquadrant.com
Authors: Tomas Vondra and David Rowley
Reviewed and tested by Simon Riggs, adding comments only
Previously, the planner would reject an index-only scan if any restriction
clause for its table used a column not available from the index, even
if that restriction clause would later be dropped from the plan entirely
because it's implied by the index's predicate. This is a fairly common
situation for partial indexes because predicates using columns not included
in the index are often the most useful kind of predicate, and we have to
duplicate (or at least imply) the predicate in the WHERE clause in order
to get the index to be considered at all. So index-only scans were
essentially unavailable with such partial indexes.
To fix, we have to do detection of implied-by-predicate clauses much
earlier in the planner. This patch puts it in check_index_predicates
(nee check_partial_indexes), meaning it gets done for every partial index,
whereas we previously only considered this issue at createplan time,
so that the work was only done for an index actually selected for use.
That could result in a noticeable planning slowdown for queries against
tables with many partial indexes. However, testing suggested that there
isn't really a significant cost, especially not with reasonable numbers
of partial indexes. We do get a small additional benefit, which is that
cost_index is more accurate since it correctly discounts the evaluation
cost of clauses that will be removed. We can also avoid considering such
clauses as potential indexquals, which saves useless matching cycles in
the case where the predicate columns aren't in the index, and prevents
generating bogus plans that double-count the clause's selectivity when
the columns are in the index.
Tomas Vondra and Kyotaro Horiguchi, reviewed by Kevin Grittner and
Konstantin Knizhnik, and whacked around a little by me
cost_subplan() supposed that the given subplan must have plan_rows > 0,
which as far as I can tell was true until recent refactoring of the
code in createplan.c; but now that code allows the Result for a provably
empty subquery to have plan_rows = 0. Rather than undo that change,
put in a clamp to prevent zero divide.
get_cheapest_fractional_path() likewise supposed that best_path->rows > 0.
This assumption has been wrong for longer. It's actually harmless given
IEEE float math, because a positive value divided by zero gives +Infinity
and compare_fractional_path_costs() will do the right thing with that.
Still, best not to assume that.
final_cost_nestloop() also seems to have some risks in this area, so
borrow the clamping logic already present in the mergejoin cost functions.
Lastly, remove unnecessary clamp_row_est() in planner.c's calls to
get_number_of_groups(). The only thing that function does with path_rows
is pass it to estimate_num_groups() which already has an internal clamp,
so we don't need the extra call; and if we did, the callers are arguably
the wrong place for it anyway.
First two items reported by Piotr Stefaniak, the others are products
of my nosing around for similar problems. No back-patch since there's
no evidence that problems arise in the back branches.
Parallel workers can now partially aggregate the data and pass the
transition values back to the leader, which can combine the partial
results to produce the final answer.
David Rowley, based on earlier work by Haribabu Kommi. Reviewed by
Álvaro Herrera, Tomas Vondra, Amit Kapila, James Sewell, and me.
In commit 19a541143a I did not make PathTarget a subtype of Node,
and embedded a RelOptInfo's reltarget directly into it rather than having
a separately-allocated Node. In hindsight that was misguided
micro-optimization, enabled by the fact that at that point we didn't have
any Paths with custom PathTargets. Now that PathTarget processing has
been fleshed out some more, it's easier to see that it's better to have
PathTarget as an indepedent Node type, even if it does cost us one more
palloc to create a RelOptInfo. So change it while we still can.
This commit just changes the representation, without doing anything more
interesting than that.
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.
In commit 1d97c19a0f and later c1d9579dd8, 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.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
In commit 19a541143a I replaced RelOptInfo.width with
RelOptInfo.reltarget.width, but I missed updating debug_print_rel()
for that because it's not compiled by default.
Reported by Salvador Fandino, patch by Michael Paquier.
This is basically a bug fix; the old code assumes that a ForeignScan
is always parallel-safe, but for postgres_fdw, for example, this is
definitely false. It should be true for file_fdw, though, since a
worker can read a file from the filesystem just as well as any other
backend process.
Original patch by Thomas Munro. Documentation, and changes to the
comments, by me.
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.
The core innovation of this patch is the introduction of the concept
of a partial path; that is, a path which if executed in parallel will
generate a subset of the output rows in each process. Gathering a
partial path produces an ordinary (complete) path. This allows us to
generate paths for parallel joins by joining a partial path for one
side (which at the baserel level is currently always a Partial Seq
Scan) to an ordinary path on the other side. This is subject to
various restrictions at present, especially that this strategy seems
unlikely to be sensible for merge joins, so only nested loops and
hash joins paths are generated.
This also allows an Append node to be pushed below a Gather node in
the case of a partitioned table.
Testing revealed that early versions of this patch made poor decisions
in some cases, which turned out to be caused by the fact that the
original cost model for Parallel Seq Scan wasn't very good. So this
patch tries to make some modest improvements in that area.
There is much more to be done in the area of generating good parallel
plans in all cases, but this seems like a useful step forward.
Patch by me, reviewed by Dilip Kumar and Amit Kapila.
This patch reduces pg_am to just two columns, a name and a handler
function. All the data formerly obtained from pg_am is now provided
in a C struct returned by the handler function. This is similar to
the designs we've adopted for FDWs and tablesample methods. There
are multiple advantages. For one, the index AM's support functions
are now simple C functions, making them faster to call and much less
error-prone, since the C compiler can now check function signatures.
For another, this will make it far more practical to define index access
methods in installable extensions.
A disadvantage is that SQL-level code can no longer see attributes
of index AMs; in particular, some of the crosschecks in the opr_sanity
regression test are no longer possible from SQL. We've addressed that
by adding a facility for the index AM to perform such checks instead.
(Much more could be done in that line, but for now we're content if the
amvalidate functions more or less replace what opr_sanity used to do.)
We might also want to expose some sort of reporting functionality, but
this patch doesn't do that.
Alexander Korotkov, reviewed by Petr Jelínek, and rather heavily
editorialized on by me.
When use_remote_estimate is enabled, consider adding ORDER BY to the
query we sending to the remote server so that we can use that ordered
data for a merge join. Commit f18c944b61
arranges to push down the query pathkeys, which seems like the case
mostly likely to be a win, but testing shows this can sometimes win,
too.
For a regular table, we know which indexes are present and therefore
test whether the ordering provided by each such index is useful. Here,
we take the opposite approach: guess what orderings would be useful if
they could be generated cheaply, and then ask the remote side what those
will cost.
Ashutosh Bapat, with very substantial cosmetic revisions by me. Also
reviewed by Rushabh Lathia.
I originally modeled this data structure on SpecialJoinInfo, but after
commit acfcd45cac 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.
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 85e5e222b1. 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.
While convincing myself that commit 7e19db0c09 would solve both of
the problems recently reported by Andreas Seltenreich, I realized that
add_paths_to_joinrel's handling of LATERAL restrictions could be made
noticeably simpler and faster if we were to retain the minimum possible
parameterization for each joinrel (that is, the set of relids supplying
unsatisfied lateral references in it). We already retain that for
baserels, in RelOptInfo.lateral_relids, so we can use that field for
joinrels too.
I re-pgindent'd the files touched here, which affects some unrelated
comments.
This is, I believe, just a minor optimization not a bug fix, so no
back-patch.
It was possible for the planner to decide to join a LATERAL subquery to
the outer side of an outer join before the outer join itself is completed.
Normally that's fine because of the associativity rules, but it doesn't
work if the subquery contains a lateral reference to the inner side of the
outer join. In such a situation the outer join *must* be done first.
join_is_legal() missed this consideration and would allow the join to be
attempted, but the actual path-building code correctly decided that no
valid join path could be made, sometimes leading to planner errors such as
"failed to build any N-way joins".
Per report from Andreas Seltenreich. Back-patch to 9.3 where LATERAL
support was added.
Add a new flag, consider_parallel, to each RelOptInfo, indicating
whether a plan for that relation could conceivably be run inside of
a parallel worker. Right now, we're pretty conservative: for example,
it might be possible to defer applying a parallel-restricted qual
in a worker, and later do it in the leader, but right now we just
don't try to parallelize access to that relation. That's probably
the right decision in most cases, anyway.
Using the new flag, generate parallel sequential scan plans for plain
baserels, meaning that we now have parallel sequential scan in
PostgreSQL. The logic here is pretty unsophisticated right now: the
costing model probably isn't right in detail, and we can't push joins
beneath Gather nodes, so the number of plans that can actually benefit
from this is pretty limited right now. Lots more work is needed.
Nevertheless, it seems time to enable this functionality so that all
this code can actually be tested easily by users and developers.
Note that, if you wish to test this functionality, it will be
necessary to set max_parallel_degree to a value greater than the
default of 0. Once a few more loose ends have been tidied up here, we
might want to consider changing the default value of this GUC, but
I'm leaving it alone for now.
Along the way, fix a bug in cost_gather: the previous coding thought
that a Gather node's transfer overhead should be costed on the basis of
the relation size rather than the number of tuples that actually need
to be passed off to the leader.
Patch by me, reviewed in earlier versions by Amit Kapila.
In addition, this path fills in a number of missing bits and pieces in
the parallel infrastructure. Paths and plans now have a parallel_aware
flag indicating whether whatever parallel-aware logic they have should
be engaged. It is believed that we will need this flag for a number of
path/plan types, not just sequential scans, which is why the flag is
generic rather than part of the SeqScan structures specifically.
Also, execParallel.c now gives parallel nodes a chance to initialize
their PlanState nodes from the DSM during parallel worker startup.
Amit Kapila, with a fair amount of adjustment by me. Review of previous
patch versions by Haribabu Kommi and others.
A Gather executor node runs any number of copies of a plan in an equal
number of workers and merges all of the results into a single tuple
stream. It can also run the plan itself, if the workers are
unavailable or haven't started up yet. It is intended to work with
the Partial Seq Scan node which will be added in future commits.
It could also be used to implement parallel query of a different sort
by itself, without help from Partial Seq Scan, if the single_copy mode
is used. In that mode, a worker executes the plan, and the parallel
leader does not, merely collecting the worker's results. So, a Gather
node could be inserted into a plan to split the execution of that plan
across two processes. Nested Gather nodes aren't currently supported,
but we might want to add support for that in the future.
There's nothing in the planner to actually generate Gather nodes yet,
so it's not quite time to break out the champagne. But we're getting
close.
Amit Kapila. Some designs suggestions were provided by me, and I also
reviewed the patch. Single-copy mode, documentation, and other minor
changes also by me.
Previously, a function call appearing at the top level of WHERE had a
hard-wired selectivity estimate of 0.3333333, a kludge conveniently dated
in the source code itself to July 1992. The expectation at the time was
that somebody would soon implement estimator support functions analogous
to those for operators; but no such code has appeared, nor does it seem
likely to in the near future. We do have an alternative solution though,
at least for immutable functions on single relations: creating an
expression index on the function call will allow ANALYZE to gather stats
about the function's selectivity. But the code in clause_selectivity()
failed to make use of such data even if it exists.
Refactor so that that will happen. I chose to make it try this technique
for any clause type for which clause_selectivity() doesn't have a special
case, not just functions. To avoid adding unnecessary overhead in the
common case where we don't learn anything new, make selfuncs.c provide an
API that hooks directly to examine_variable() and then var_eq_const(),
rather than the previous coding which laboriously constructed an OpExpr
only so that it could be expensively deconstructed again.
I preserved the behavior that the default estimate for a function call
is 0.3333333. (For any other expression node type, it's 0.5, as before.)
I had originally thought to make the default be 0.5 across the board, but
changing a default estimate that's survived for twenty-three years seems
like something not to do without a lot more testing than I care to put
into it right now.
Per a complaint from Jehan-Guillaume de Rorthais. Back-patch into 9.5,
but not further, at least for the moment.
One of the changes I made in commit 8703059c6b turns out not to have
been such a good idea: we still need the exception in join_is_legal() that
allows a join if both inputs already overlap the RHS of the special join
we're checking. Otherwise we can miss valid plans, and might indeed fail
to find a plan at all, as in recent report from Andreas Seltenreich.
That code was added way back in commit c17117649b, but I failed to
include a regression test case then; my bad. Put it back with a better
explanation, and a test this time. The logic does end up a bit different
than before though: I now believe it's appropriate to make this check
first, thereby allowing such a case whether or not we'd consider the
previous SJ(s) to commute with this one. (Presumably, we already decided
they did; but it was confusing to have this consideration in the middle
of the code that was handling the other case.)
Back-patch to all active branches, like the previous patch.
Commit 85e5e222b1 turns out not to have taken
care of all cases of the partially-evaluatable-PlaceHolderVar problem found
by Andreas Seltenreich's fuzz testing. I had set it up to check for risky
PHVs only in the event that we were making a star-schema-based exception to
the param_source_rels join ordering heuristic. However, it turns out that
the problem can occur even in joins that satisfy the param_source_rels
heuristic, in which case allow_star_schema_join() isn't consulted.
Refactor so that we check for risky PHVs whenever the proposed join has
any remaining parameterization.
Back-patch to 9.2, like the previous patch (except for the regression test
case, which only works back to 9.3 because it uses LATERAL).
Note that this discovery implies that problems of this sort could've
occurred in 9.2 and up even before the star-schema patch; though I've not
tried to prove that experimentally.
Formerly, this function would always return "true" for an appendrel child
relation, because it would think that the appendrel parent was a potential
join target for the child. In principle that should only lead to some
inefficiency in planning, but fuzz testing by Andreas Seltenreich disclosed
that it could lead to "could not find pathkey item to sort" planner errors
in odd corner cases. Specifically, we would think that all columns of a
child table's multicolumn index were interesting pathkeys, causing us to
generate a MergeAppend path that sorts by all the columns. However, if any
of those columns weren't actually used above the level of the appendrel,
they would not get added to that rel's targetlist, which would result in
being unable to resolve the MergeAppend's sort keys against its targetlist
during createplan.c.
Backpatch to 9.3. In older versions, columns of an appendrel get added
to its targetlist even if they're not mentioned above the scan level,
so that the failure doesn't occur. It might be worth back-patching this
fix to older versions anyway, but I'll refrain for the moment.
Further testing revealed that commit f69b4b9495 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.
Per the discussion in optimizer/README, it's unsafe to reassociate anything
into or out of the RHS of a SEMI or ANTI join. An example from Piotr
Stefaniak showed that join_is_legal() wasn't sufficiently enforcing this
rule, so lock it down a little harder.
I couldn't find a reasonably simple example of the optimizer trying to
do this, so no new regression test. (Piotr's example involved the random
search in GEQO accidentally trying an invalid case and triggering a sanity
check way downstream in clause selectivity estimation, which did not seem
like a sequence of events that would be useful to memorialize in a
regression test as-is.)
Back-patch to all active branches.
In commit b514a7460d, I changed the planner
so that it would allow nestloop paths to remain partially parameterized,
ie the inner relation might need parameters from both the current outer
relation and some upper-level outer relation. That's fine so long as we're
talking about distinct parameters; but the patch also allowed creation of
nestloop paths for cases where the inner relation's parameter was a
PlaceHolderVar whose eval_at set included the current outer relation and
some upper-level one. That does *not* work.
In principle we could allow such a PlaceHolderVar to be evaluated at the
lower join node using values passed down from the upper relation along with
values from the join's own outer relation. However, nodeNestloop.c only
supports simple Vars not arbitrary expressions as nestloop parameters.
createplan.c is also a few bricks shy of being able to handle such cases;
it misplaces the PlaceHolderVar parameters in the plan tree, which is why
the visible symptoms of this bug are "plan should not reference subplan's
variable" and "failed to assign all NestLoopParams to plan nodes" planner
errors.
Adding the necessary complexity to make this work doesn't seem like it
would be repaid in significantly better plans, because in cases where such
a PHV exists, there is probably a corresponding join order constraint that
would allow a good plan to be found without using the star-schema exception.
Furthermore, adding complexity to nodeNestloop.c would create a run-time
penalty even for plans where this whole consideration is irrelevant.
So let's just reject such paths instead.
Per fuzz testing by Andreas Seltenreich; the added regression test is based
on his example query. Back-patch to 9.2, like the previous patch.
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.
In many cases, we can implement a semijoin as a plain innerjoin by first
passing the righthand-side relation through a unique-ification step.
However, one of the cases where this does NOT work is where the RHS has
a LATERAL reference to the LHS; that makes the RHS dependent on the LHS
so that unique-ification is meaningless. joinpath.c understood this,
and so would not generate any join paths of this kind ... but join_is_legal
neglected to check for the case, so it would think that we could do it.
The upshot would be a "could not devise a query plan for the given query"
failure once we had failed to generate any join paths at all for the bogus
join pair.
Back-patch to 9.3 where LATERAL was added.
The planner generally expects that the estimated rowcount of any relation
is at least one row, *unless* it has been proven empty by constraint
exclusion or similar mechanisms, which is marked by installing a dummy path
as the rel's cheapest path (cf. IS_DUMMY_REL). When I split up
allpaths.c's processing of base rels into separate set_base_rel_sizes and
set_base_rel_pathlists steps, the intention was that dummy rels would get
marked as such during the "set size" step; this is what justifies an Assert
in indxpath.c's get_loop_count that other relations should either be dummy
or have positive rowcount. Unfortunately I didn't get that quite right
for append relations: if all the child rels have been proven empty then
set_append_rel_size would come up with a rowcount of zero, which is
correct, but it didn't then do set_dummy_rel_pathlist. (We would have
ended up with the right state after set_append_rel_pathlist, but that's
too late, if we generate indexpaths for some other rel first.)
In addition to fixing the actual bug, I installed an Assert enforcing this
convention in set_rel_size; that then allows simplification of a couple
of now-redundant tests for zero rowcount in set_append_rel_size.
Also, to cover the possibility that third-party FDWs have been careless
about not returning a zero rowcount estimate, apply clamp_row_est to
whatever an FDW comes up with as the rows estimate.
Per report from Andreas Seltenreich. Back-patch to 9.2. Earlier branches
did not have the separation between set_base_rel_sizes and
set_base_rel_pathlists steps, so there was no intermediate state where an
appendrel would have had inconsistent rowcount and pathlist. It's possible
that adding the Assert to set_rel_size would be a good idea in older
branches too; but since they're not under development any more, it's likely
not worth the trouble.
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.
UPDATE ... WHERE CURRENT OF would not work in conjunction with
RLS. Arrange to allow the CURRENT OF expression to be pushed down.
Issue noted by Peter Geoghegan. Patch by Dean Rasheed. Back patch
to 9.5 where RLS was introduced.
When the inner side of a nestloop SEMI or ANTI join is an indexscan that
uses all the join clauses as indexquals, it can be presumed that both
matched and unmatched outer rows will be processed very quickly: for
matched rows, we'll stop after fetching one row from the indexscan, while
for unmatched rows we'll have an indexscan that finds no matching index
entries, which should also be quick. The planner already knew about this,
but it was nonetheless charging for at least one full run of the inner
indexscan, as a consequence of concerns about the behavior of materialized
inner scans --- but those concerns don't apply in the fast case. If the
inner side has low cardinality (many matching rows) this could make an
indexscan plan look far more expensive than it actually is. To fix,
rearrange the work in initial_cost_nestloop/final_cost_nestloop so that we
don't add the inner scan cost until we've inspected the indexquals, and
then we can add either the full-run cost or just the first tuple's cost as
appropriate.
Experimentation with this fix uncovered another problem: add_path and
friends were coded to disregard cheap startup cost when considering
parameterized paths. That's usually okay (and desirable, because it thins
the path herd faster); but in this fast case for SEMI/ANTI joins, it could
result in throwing away the desired plain indexscan path in favor of a
bitmap scan path before we ever get to the join costing logic. In the
many-matching-rows cases of interest here, a bitmap scan will do a lot more
work than required, so this is a problem. To fix, add a per-relation flag
consider_param_startup that works like the existing consider_startup flag,
but applies to parameterized paths, and set it for relations that are the
inside of a SEMI or ANTI join.
To make this patch reasonably safe to back-patch, care has been taken to
avoid changing the planner's behavior except in the very narrow case of
SEMI/ANTI joins with inner indexscans. There are places in
compare_path_costs_fuzzily and add_path_precheck that are not terribly
consistent with the new approach, but changing them will affect planner
decisions at the margins in other cases, so we'll leave that for a
HEAD-only fix.
Back-patch to 9.3; before that, the consider_startup flag didn't exist,
meaning that the second aspect of the patch would be too invasive.
Per a complaint from Peter Holzer and analysis by Tomas Vondra.
Use "a" and "an" correctly, mostly in comments. Two error messages were
also fixed (they were just elogs, so no translation work required). Two
function comments in pg_proc.h were also fixed. Etsuro Fujita reported one
of these, but I found a lot more with grep.
Also fix a few other typos spotted while grepping for the a/an typos.
For example, "consists out of ..." -> "consists of ...". Plus a "though"/
"through" mixup reported by Euler Taveira.
Many of these typos were in old code, which would be nice to backpatch to
make future backpatching easier. But much of the code was new, and I didn't
feel like crafting separate patches for each branch. So no backpatching.
This SQL standard functionality allows to aggregate data by different
GROUP BY clauses at once. Each grouping set returns rows with columns
grouped by in other sets set to NULL.
This could previously be achieved by doing each grouping as a separate
query, conjoined by UNION ALLs. Besides being considerably more concise,
grouping sets will in many cases be faster, requiring only one scan over
the underlying data.
The current implementation of grouping sets only supports using sorting
for input. Individual sets that share a sort order are computed in one
pass. If there are sets that don't share a sort order, additional sort &
aggregation steps are performed. These additional passes are sourced by
the previous sort step; thus avoiding repeated scans of the source data.
The code is structured in a way that adding support for purely using
hash aggregation or a mix of hashing and sorting is possible. Sorting
was chosen to be supported first, as it is the most generic method of
implementation.
Instead of, as in an earlier versions of the patch, representing the
chain of sort and aggregation steps as full blown planner and executor
nodes, all but the first sort are performed inside the aggregation node
itself. This avoids the need to do some unusual gymnastics to handle
having to return aggregated and non-aggregated tuples from underlying
nodes, as well as having to shut down underlying nodes early to limit
memory usage. The optimizer still builds Sort/Agg node to describe each
phase, but they're not part of the plan tree, but instead additional
data for the aggregation node. They're a convenient and preexisting way
to describe aggregation and sorting. The first (and possibly only) sort
step is still performed as a separate execution step. That retains
similarity with existing group by plans, makes rescans fairly simple,
avoids very deep plans (leading to slow explains) and easily allows to
avoid the sorting step if the underlying data is sorted by other means.
A somewhat ugly side of this patch is having to deal with a grammar
ambiguity between the new CUBE keyword and the cube extension/functions
named cube (and rollup). To avoid breaking existing deployments of the
cube extension it has not been renamed, neither has cube been made a
reserved keyword. Instead precedence hacking is used to make GROUP BY
cube(..) refer to the CUBE grouping sets feature, and not the function
cube(). To actually group by a function cube(), unlikely as that might
be, the function name has to be quoted.
Needs a catversion bump because stored rules may change.
Author: Andrew Gierth and Atri Sharma, with contributions from Andres Freund
Reviewed-By: Andres Freund, Noah Misch, Tom Lane, Svenne Krap, Tomas
Vondra, Erik Rijkers, Marti Raudsepp, Pavel Stehule
Discussion: CAOeZVidmVRe2jU6aMk_5qkxnB7dfmPROzM7Ur8JPW5j8Y5X-Lw@mail.gmail.com
For upcoming BRIN opclasses, it's convenient to have strategy numbers
defined in a single place. Since there's nothing appropriate, create
it. The StrategyNumber typedef now lives there, as well as existing
strategy numbers for B-trees (from skey.h) and R-tree-and-friends (from
gist.h). skey.h is forced to include stratnum.h because of the
StrategyNumber typedef, but gist.h is not; extensions that currently
rely on gist.h for rtree strategy numbers might need to add a new
A few .c files can stop including skey.h and/or gist.h, which is a nice
side benefit.
Per discussion:
https://www.postgresql.org/message-id/20150514232132.GZ2523@alvh.no-ip.org
Authored by Emre Hasegeli and Álvaro.
(It's not clear to me why bootscanner.l has any #include lines at all.)
Add a TABLESAMPLE clause to SELECT statements that allows
user to specify random BERNOULLI sampling or block level
SYSTEM sampling. Implementation allows for extensible
sampling functions to be written, using a standard API.
Basic version follows SQLStandard exactly. Usable
concrete use cases for the sampling API follow in later
commits.
Petr Jelinek
Reviewed by Michael Paquier and Simon Riggs
Commit e7cb7ee145 included some design
decisions that seem pretty questionable to me, and there was quite a lot
of stuff not to like about the documentation and comments. Clean up
as follows:
* Consider foreign joins only between foreign tables on the same server,
rather than between any two foreign tables with the same underlying FDW
handler function. In most if not all cases, the FDW would simply have had
to apply the same-server restriction itself (far more expensively, both for
lack of caching and because it would be repeated for each combination of
input sub-joins), or else risk nasty bugs. Anyone who's really intent on
doing something outside this restriction can always use the
set_join_pathlist_hook.
* Rename fdw_ps_tlist/custom_ps_tlist to fdw_scan_tlist/custom_scan_tlist
to better reflect what they're for, and allow these custom scan tlists
to be used even for base relations.
* Change make_foreignscan() API to include passing the fdw_scan_tlist
value, since the FDW is required to set that. Backwards compatibility
doesn't seem like an adequate reason to expect FDWs to set it in some
ad-hoc extra step, and anyway existing FDWs can just pass NIL.
* Change the API of path-generating subroutines of add_paths_to_joinrel,
and in particular that of GetForeignJoinPaths and set_join_pathlist_hook,
so that various less-used parameters are passed in a struct rather than
as separate parameter-list entries. The objective here is to reduce the
probability that future additions to those parameter lists will result in
source-level API breaks for users of these hooks. It's possible that this
is even a small win for the core code, since most CPU architectures can't
pass more than half a dozen parameters efficiently anyway. I kept root,
joinrel, outerrel, innerrel, and jointype as separate parameters to reduce
code churn in joinpath.c --- in particular, putting jointype into the
struct would have been problematic because of the subroutines' habit of
changing their local copies of that variable.
* Avoid ad-hocery in ExecAssignScanProjectionInfo. It was probably all
right for it to know about IndexOnlyScan, but if the list is to grow
we should refactor the knowledge out to the callers.
* Restore nodeForeignscan.c's previous use of the relcache to avoid
extra GetFdwRoutine lookups for base-relation scans.
* Lots of cleanup of documentation and missed comments. Re-order some
code additions into more logical places.
Foreign data wrappers can use this capability for so-called "join
pushdown"; that is, instead of executing two separate foreign scans
and then joining the results locally, they can generate a path which
performs the join on the remote server and then is scanned locally.
This commit does not extend postgres_fdw to take advantage of this
capability; it just provides the infrastructure.
Custom scan providers can use this in a similar way. Previously,
it was only possible for a custom scan provider to scan a single
relation. Now, it can scan an entire join tree, provided of course
that it knows how to produce the same results that the join would
have produced if executed normally.
KaiGai Kohei, reviewed by Shigeru Hanada, Ashutosh Bapat, and me.
The original security barrier view implementation, on which RLS is
built, prevented all non-leakproof functions from being pushed down to
below the view, even when the function was not receiving any data from
the view. This optimization improves on that situation by, instead of
checking strictly for non-leakproof functions, it checks for Vars being
passed to non-leakproof functions and allows functions which do not
accept arguments or whose arguments are not from the current query level
(eg: constants can be particularly useful) to be pushed down.
As discussed, this does mean that a function which is pushed down might
gain some idea that there are rows meeting a certain criteria based on
the number of times the function is called, but this isn't a
particularly new issue and the documentation in rules.sgml already
addressed similar covert-channel risks. That documentation is updated
to reflect that non-leakproof functions may be pushed down now, if
they meet the above-described criteria.
Author: Dean Rasheed, with a bit of rework to make things clearer,
along with comment and documentation updates from me.
The cross-reference to set_append_rel_pathlist() was obsoleted by
commit e2fa76d80b, which split what
had been set_rel_pathlist() and child routines into two sets of
functions. But I (tgl) evidently missed updating this comment.
Back-patch to 9.2 to avoid unnecessary divergence among branches.
Amit Langote
This adds a new GiST opclass method, 'fetch', which is used to reconstruct
the original Datum from the value stored in the index. Also, the 'canreturn'
index AM interface function gains a new 'attno' argument. That makes it
possible to use index-only scans on a multi-column index where some of the
opclasses support index-only scans but some do not.
This patch adds support in the box and point opclasses. Other opclasses
can added later as follow-on patches (btree_gist would be particularly
interesting).
Anastasia Lubennikova, with additional fixes and modifications by me.
While poking at David Kubečka's issue I noticed an ancient logic error
in get_loop_count(): it used 1.0 as a "no data yet" indicator, but since
that is actually a valid rowcount estimate, this doesn't work. If we
have one input relation with 1.0 as rowcount and then another one with
a larger rowcount, we should use 1.0 as the result, but we picked the
larger rowcount instead. (I think when I coded this, I recognized the
conflict, but mistakenly thought that the logic would pick the desired
count anyway.)
Fixing this changed the plan for one existing regression test case.
Since the point of that test is to exercise creation of a particular
shape of nestloop plan, I tweaked the query a little bit so it still
results in the same plan choice.
This is definitely a bug, but I'm hesitant to back-patch since it might
change plan choices unexpectedly, and anyway failure to implement a
heuristic precisely as intended is a pretty low-grade bug.
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.
This code relied on pointer equality to identify which restriction clauses
also appear in the indexquals (and, therefore, don't need to be applied as
simple filter conditions). That was okay once upon a time, years ago,
before we introduced the equivalence-class machinery. Now there's about a
50-50 chance that an equality clause appearing in the indexquals will be
the mirror image (commutator) of its mate in the restriction list. When
that happens, we'd erroneously think that the clause would be re-evaluated
at each visited row, and therefore inflate the cost estimate for the
indexscan by the clause's cost.
Add some logic to catch this case. It seems to me that it continues not to
be worthwhile to expend the extra predicate-proof work that createplan.c
will do on the finally-selected plan, but this case is common enough and
cheap enough to handle that we should do so.
This will make a small difference (about one cpu_operator_cost per row)
in simple cases; but in situations where there's an expensive function in
the indexquals, it can make a very large difference, as seen in recent
example from Jeff Janes.
This is a long-standing bug, but I'm hesitant to back-patch because of the
possibility of destabilizing plan choices that people may be happy with.
Part of the intent of the parameterized-path mechanism was to handle
star-schema queries efficiently, but some overly-restrictive search
limiting logic added in commit e2fa76d80b
prevented such cases from working as desired. Fix that and add a
regression test about it. Per gripe from Marc Cousin.
This is arguably a bug rather than a new feature, so back-patch to 9.2
where parameterized paths were introduced.
This requires changing quite a few places that were depending on
sizeof(HeapTupleHeaderData), but it seems for the best.
Michael Paquier, some adjustments by me
This patch adds a function that replaces a bms_membership() test followed
by a bms_singleton_member() call, performing both the test and the
extraction of a singleton set's member in one scan of the bitmapset.
The performance advantage over the old way is probably minimal in current
usage, but it seems worthwhile on notational grounds anyway.
David Rowley
This patch adds a way of iterating through the members of a bitmapset
nondestructively, unlike the old way with bms_first_member(). While
bms_next_member() is very slightly slower than bms_first_member()
(at least for typical-size bitmapsets), eliminating the need to palloc
and pfree a temporary copy of the target bitmapset is a significant win.
So this method should be preferred in all cases where a temporary copy
would be necessary.
Tom Lane, with suggestions from Dean Rasheed and David Rowley
As pointed out by Robert, we should really have named pg_rowsecurity
pg_policy, as the objects stored in that catalog are policies. This
patch fixes that and updates the column names to start with 'pol' to
match the new catalog name.
The security consideration for COPY with row level security, also
pointed out by Robert, has also been addressed by remembering and
re-checking the OID of the relation initially referenced during COPY
processing, to make sure it hasn't changed under us by the time we
finish planning out the query which has been built.
Robert and Alvaro also commented on missing OCLASS and OBJECT entries
for POLICY (formerly ROWSECURITY or POLICY, depending) in various
places. This patch fixes that too, which also happens to add the
ability to COMMENT on policies.
In passing, attempt to improve the consistency of messages, comments,
and documentation as well. This removes various incarnations of
'row-security', 'row-level security', 'Row-security', etc, in favor
of 'policy', 'row level security' or 'row_security' as appropriate.
Happy Thanksgiving!
Instead of register_custom_path_provider and a CreateCustomScanPath
callback, let's just provide a standard function hook in set_rel_pathlist.
This is more flexible than what was previously committed, is more like the
usual conventions for planner hooks, and requires less support code in the
core. We had discussed this design (including centralizing the
set_cheapest() calls) back in March or so, so I'm not sure why it wasn't
done like this already.
This allows extension modules to define their own methods for
scanning a relation, and get the core code to use them. It's
unclear as yet how much use this capability will find, but we
won't find out if we never commit it.
KaiGai Kohei, reviewed at various times and in various levels
of detail by Shigeru Hanada, Tom Lane, Andres Freund, Álvaro
Herrera, and myself.
Since we taught btree to handle ScalarArrayOpExpr quals natively (commit
9e8da0f757), the planner has always included
ScalarArrayOpExpr quals in index conditions if possible. However, if the
qual is for a non-first index column, this could result in an inferior plan
because we can no longer take advantage of index ordering (cf. commit
807a40c551). It can be better to omit the
ScalarArrayOpExpr qual from the index condition and let it be done as a
filter, so that the output doesn't need to get sorted. Indeed, this is
true for the query introduced as a test case by the latter commit.
To fix, restructure get_index_paths and build_index_paths so that we
consider paths both with and without ScalarArrayOpExpr quals in non-first
index columns. Redesign the API of build_index_paths so that it reports
what it found, saving useless second or third calls.
Report and patch by Andrew Gierth (though rather heavily modified by me).
Back-patch to 9.2 where this code was introduced, since the issue can
result in significant performance regressions compared to plans produced
by 9.1 and earlier.
As of commit a87c72915 (which later got backpatched as far as 9.1),
we're explicitly supporting the notion that append relations can be
nested; this can occur when UNION ALL constructs are nested, or when
a UNION ALL contains a table with inheritance children.
Bug #11457 from Nelson Page, as well as an earlier report from Elvis
Pranskevichus, showed that there were still nasty bugs associated with such
cases: in particular the EquivalenceClass mechanism could try to generate
"join" clauses connecting an appendrel child to some grandparent appendrel,
which would result in assertion failures or bogus plans.
Upon investigation I concluded that all current callers of
find_childrel_appendrelinfo() need to be fixed to explicitly consider
multiple levels of parent appendrels. The most complex fix was in
processing of "broken" EquivalenceClasses, which are ECs for which we have
been unable to generate all the derived equality clauses we would like to
because of missing cross-type equality operators in the underlying btree
operator family. That code path is more or less entirely untested by
the regression tests to date, because no standard opfamilies have such
holes in them. So I wrote a new regression test script to try to exercise
it a bit, which turned out to be quite a worthwhile activity as it exposed
existing bugs in all supported branches.
The present patch is essentially the same as far back as 9.2, which is
where parameterized paths were introduced. In 9.0 and 9.1, we only need
to back-patch a small fragment of commit 5b7b5518d, which fixes failure to
propagate out the original WHERE clauses when a broken EC contains constant
members. (The regression test case results show that these older branches
are noticeably stupider than 9.2+ in terms of the quality of the plans
generated; but we don't really care about plan quality in such cases,
only that the plan not be outright wrong. A more invasive fix in the
older branches would not be a good idea anyway from a plan-stability
standpoint.)
We can allow this even without any specific knowledge of the semantics
of the window function, so long as pushed-down quals will either accept
every row in a given window partition, or reject every such row. Because
window functions act only within a partition, such a case can't result
in changing the window functions' outputs for any surviving row.
Eliminating entire partitions in this way obviously can reduce the cost
of the window-function computations substantially.
The fly in the ointment is that it's hard to be entirely sure whether
this is true for an arbitrary qual condition. This patch allows pushdown
if (a) the qual references only partitioning columns, and (b) the qual
contains no volatile functions. We are at risk of incorrect results if
the qual can produce different answers for values that the partitioning
equality operator sees as equal. While it's not hard to invent cases
for which that can happen, it seems to seldom be a problem in practice,
since no one has complained about a similar assumption that we've had
for many years with respect to DISTINCT. The potential performance
gains seem to be worth the risk.
David Rowley, reviewed by Vik Fearing; some credit is due also to
Thomas Mayer who did considerable preliminary investigation.
A WHERE clause applied to the output of a subquery with DISTINCT should
theoretically be applied only once per distinct row; but if we push it
into the subquery then it will be evaluated at each row before duplicate
elimination occurs. If the qual is volatile this can give rise to
observably wrong results, so don't do that.
While at it, refactor a little bit to allow subquery_is_pushdown_safe
to report more than one kind of restrictive condition without indefinitely
expanding its argument list.
Although this is a bug fix, it seems unwise to back-patch it into released
branches, since it might de-optimize plans for queries that aren't giving
any trouble in practice. So apply to 9.4 but not further back.
If a sub-select-in-FROM gets flattened into the upper query, then we
naturally get rid of any output columns that are defined in the sub-select
text but not actually used in the upper query. However, this doesn't
happen when it's not possible to flatten the subquery, for example because
it contains GROUP BY, LIMIT, etc. Allowing the subquery to compute useless
output columns is often fairly harmless, but sometimes it has significant
performance cost: the unused output might be an expensive expression,
or it might be a Var from a relation that we could remove entirely (via
the join-removal logic) if only we realized that we didn't really need
that Var. Situations like this are common when expanding views, so it
seems worth taking the trouble to detect and remove unused outputs.
Because the upper query's Var numbering for subquery references depends on
positions in the subquery targetlist, we don't want to renumber the items
we leave behind. Instead, we can implement "removal" by replacing the
unwanted expressions with simple NULL constants. This wastes a few cycles
at runtime, but not enough to justify more work in the planner.
This reverts commit ee1e5662d8, as well as
a remarkably large number of followup commits, which were mostly concerned
with the fact that the implementation didn't work terribly well. It still
doesn't: we probably need some rather basic work in the GUC infrastructure
if we want to fully support GUCs whose default varies depending on the
value of another GUC. Meanwhile, it also emerged that there wasn't really
consensus in favor of the definition the patch tried to implement (ie,
effective_cache_size should default to 4 times shared_buffers). So whack
it all back to where it was. In a followup commit, I'll do what was
recently agreed to, which is to simply change the default to a higher
value.
The original coding of EquivalenceClasses didn't foresee that appendrel
child relations might themselves be appendrels; but this is possible for
example when a UNION ALL subquery scans a table with inheritance children.
The oversight led to failure to optimize ordering-related issues very well
for the grandchild tables. After some false starts involving explicitly
flattening the appendrel representation, we found that this could be fixed
easily by removing a few implicit assumptions about appendrel parent rels
not being children themselves.
Kyotaro Horiguchi and Tom Lane, reviewed by Noah Misch
The previous method was overly complex and underly correct; in particular,
by assigning the default value with PGC_S_OVERRIDE, it prevented later
attempts to change the setting in postgresql.conf, as noted by Jeff Janes.
We should just assign the default value with source PGC_S_DYNAMIC_DEFAULT,
which will have the desired priority relative to the boot_val as well as
user-set values.
There is still a gap in this method: if there's an explicit assignment of
effective_cache_size = -1 in the postgresql.conf file, and that assignment
appears before shared_buffers is assigned, the code will substitute 4 times
the bootstrap default for shared_buffers, and that value will then persist
(since it will have source PGC_S_FILE). I don't see any very nice way
to avoid that though, and it's not a case to be expected in practice.
The existing comments in guc-file.l look forward to a redesign of the
DYNAMIC_DEFAULT mechanism; if that ever happens, we should consider this
case as one of the things we'd like to improve.
Fix integer overflow issue noted by Magnus Hagander, as well as a bunch
of other infelicities in commit ee1e5662d8
and its unreasonably large number of followups.
We don't need make_restrictinfo_from_bitmapqual() anymore at all.
generate_bitmap_or_paths() doesn't need to be exported, and we can
drop its rather klugy restriction_only flag.
It's possible to extract a restriction OR clause from a join clause that
has the form of an OR-of-ANDs, if each sub-AND includes a clause that
mentions only one specific relation. While PG has been aware of that idea
for many years, the code previously only did it if it could extract an
indexable OR clause. On reflection, though, that seems a silly limitation:
adding a restriction clause can be a win by reducing the number of rows
that have to be filtered at the join step, even if we have to test the
clause as a plain filter clause during the scan. This should be especially
useful for foreign tables, where the change can cut the number of rows that
have to be retrieved from the foreign server; but testing shows it can win
even on local tables. Per a suggestion from Robert Haas.
As a heuristic, I made the code accept an extracted restriction clause
if its estimated selectivity is less than 0.9, which will probably result
in accepting extracted clauses just about always. We might need to tweak
that later based on experience.
Since the code no longer has even a weak connection to Path creation,
remove orindxpath.c and create a new file optimizer/util/orclauses.c.
There's some additional janitorial cleanup of now-dead code that needs
to happen, but it seems like that's a fit subject for a separate commit.
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
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 db9f0e1d9a, 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.
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 4da6439bd8, 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.
This is SQL-standard with a few extensions, namely support for
subqueries and outer references in clause expressions.
catversion bump due to change in Aggref and WindowFunc.
David Fetter, reviewed by Dean Rasheed.
The code in set_append_rel_pathlist() for building parameterized paths
for append relations (inheritance and UNION ALL combinations) supposed
that the cheapest regular path for a child relation would still be cheapest
when reparameterized. Which might not be the case, particularly if the
added join conditions are expensive to compute, as in a recent example from
Jeff Janes. Fix it to compare child path costs *after* reparameterizing.
We can short-circuit that if the cheapest pre-existing path is already
parameterized correctly, which seems likely to be true often enough to be
worth checking for.
Back-patch to 9.2 where parameterized paths were introduced.
The planner is aware that it mustn't push down upper-level quals into
subqueries if the quals reference subquery output columns that contain
set-returning functions or volatile functions, or are non-DISTINCT outputs
of a DISTINCT ON subquery. However, it missed making this check when
there were one or more levels of UNION or INTERSECT above the dangerous
expression. This could lead to "set-valued function called in context that
cannot accept a set" errors, as seen in bug #8213 from Eric Soroos, or to
silently wrong answers in the other cases.
To fix, refactor the checks so that we make the column-is-unsafe checks
during subquery_is_pushdown_safe(), which already has to recursively
inspect all arms of a set-operation tree. This makes
qual_is_pushdown_safe() considerably simpler, at the cost that we will
spend some cycles checking output columns that possibly aren't referenced
in any upper qual. But the cases where this code gets executed at all
are already nontrivial queries, so it's unlikely anybody will notice any
slowdown of planning.
This has been broken since commit 05f916e6ad,
which makes the bug over ten years old. A bit surprising nobody noticed it
before now.
This reverts the code changes in 50c137487c,
which turned out to induce crashes and not completely fix the problem
anyway. That commit only considered single subqueries that were excluded
by constraint-exclusion logic, but actually the problem also exists for
subqueries that are appendrel members (ie part of a UNION ALL list). In
such cases we can't add a dummy subpath to the appendrel's AppendPath list
without defeating the logic that recognizes when an appendrel is completely
excluded. Instead, fix the problem by having setrefs.c scan the rangetable
an extra time looking for subqueries that didn't get into the plan tree.
(This approach depends on the 9.2 change that made set_subquery_pathlist
generate dummy paths for excluded single subqueries, so that the exclusion
behavior is the same for single subqueries and appendrel members.)
Note: it turns out that the appendrel form of the missed-permissions-checks
bug exists as far back as 8.4. However, since the practical effect of that
bug seems pretty minimal, consensus is to not attempt to fix it in the back
branches, at least not yet. Possibly we could back-port this patch once
it's gotten a reasonable amount of testing in HEAD. For the moment I'm
just going to revert the previous patch in 9.2.
A view defined as "select <something> where false" had the curious property
that the system wouldn't check whether users had the privileges necessary
to select from it. More generally, permissions checks could be skipped
for tables referenced in sub-selects or views that were proven empty by
constraint exclusion (although some quick testing suggests this seldom
happens in cases of practical interest). This happened because the planner
failed to include rangetable entries for such tables in the finished plan.
This was noticed in connection with erroneous handling of materialized
views, but actually the issue is quite unrelated to matviews. Therefore,
revert commit 200ba1667b in favor of a more
direct test for the real problem.
Back-patch to 9.2 where the bug was introduced (by commit
7741dd6590).
This patch gets rid of the concept of, and infrastructure for,
non-canonical PathKeys; we now only ever create canonical pathkey lists.
The need for non-canonical pathkeys came from the desire to have
grouping_planner initialize query_pathkeys and related pathkey lists before
calling query_planner. However, since query_planner didn't actually *do*
anything with those lists before they'd been made canonical, we can get rid
of the whole mess by just not creating the lists at all until the point
where we formerly canonicalized them.
There are several ways in which we could implement that without making
query_planner itself deal with grouping/sorting features (which are
supposed to be the province of grouping_planner). I chose to add a
callback function to query_planner's API; other alternatives would have
required adding more fields to PlannerInfo, which while not bad in itself
would create an ABI break for planner-related plugins in the 9.2 release
series. This still breaks ABI for anything that calls query_planner
directly, but it seems somewhat unlikely that there are any such plugins.
I had originally conceived of this change as merely a step on the way to
fixing bug #8049 from Teun Hoogendoorn; but it turns out that this fixes
that bug all by itself, as per the added regression test. The reason is
that now get_eclass_for_sort_expr is adding the ORDER BY expression at the
end of EquivalenceClass creation not the start, and so anything that is in
a multi-member EquivalenceClass has already been created with correct
em_nullable_relids. I am suspicious that there are related scenarios in
which we still need to teach get_eclass_for_sort_expr to compute correct
nullable_relids, but am not eager to risk destabilizing either 9.2 or 9.3
to fix bugs that are only hypothetical. So for the moment, do this and
stop here.
Back-patch to 9.2 but not to earlier branches, since they don't exhibit
this bug for lack of join-clause-movement logic that depends on
em_nullable_relids being correct. (We might have to revisit that choice
if any related bugs turn up.) In 9.2, don't change the signature of
make_pathkeys_for_sortclauses nor remove canonicalize_pathkeys, so as
not to risk more plugin breakage than we have to.
Robert Haas
Tom Lane
Teodor Sigaev
Simon Riggs
Bruce Momjian
Heikki Linnakangas
Joe Conway
Andres Freund
Alvaro Herrera
Stephen Frost
Noah Misch