When considering nestloop paths for individual partitions within
a partitionwise join, if the inner path is parameterized, it is
parameterized by the topmost parent of the outer rel, not the
corresponding outer rel itself. Therefore, we need to translate the
parameterization so that the inner path is parameterized by the
corresponding outer rel.
Up to now, we did this while generating join paths. However, that's
problematic because we must also translate some expressions that are
shared across all paths for a relation, such as restriction clauses
(kept in the RelOptInfo and/or IndexOptInfo) and TableSampleClauses
(kept in the RangeTblEntry). The existing code fails to translate
these at all, leading to wrong answers, odd failures such as
"variable not found in subplan target list", or executor crashes.
But we can't modify them during path generation, because that would
break things if we end up choosing some non-partitioned-join path.
So this patch postpones reparameterization of the inner path until
createplan.c, where it is safe to modify the referenced RangeTblEntry,
RelOptInfo or IndexOptInfo, because we have made a final choice of which
Path to use. We do still have to check during path generation that
the reparameterization will be possible. So we introduce a new
function path_is_reparameterizable_by_child() to detect that.
The duplication between path_is_reparameterizable_by_child() and
reparameterize_path_by_child() is a bit annoying, but there seems
no other good answer. A small benefit is that we can avoid building
useless reparameterized trees in cases where a non-partitioned join
is ultimately chosen. Also, reparameterize_path_by_child() can now
be allowed to scribble on the input paths, saving a few cycles.
This fix repairs the same problems previously addressed in the
back branches by commits 62f120203 et al.
Richard Guo, reviewed at various times by Ashutosh Bapat, Andrei
Lepikhov, Alena Rybakina, Robert Haas, and myself
Discussion: https://postgr.es/m/CAMbWs496+N=UAjOc=rcD3P7B6oJe4rZw08e_TZRUsWbPxZW3Tw@mail.gmail.com
Applying add_outer_joins_to_relids() to a child join doesn't actually
work, even if we've built a SpecialJoinInfo specialized to the child,
because that function will also compare the join's relids to elements
of the main join_info_list, which only deal in regular relids not
child relids. This mistake escaped detection by the existing
partitionwise join tests because they didn't test any cases where
add_outer_joins_to_relids() needs to add additional OJ relids (that
is, any cases where join reordering per identity 3 is possible).
Instead, let's apply adjust_child_relids() to the relids of the parent
join. This requires minor code reordering to collect the relevant
AppendRelInfo structures first, but that's work we'd do shortly anyway.
Report and fix by Richard Guo; cosmetic changes by me
Discussion: https://postgr.es/m/CAMbWs49NCNbyubZWgci3o=_OTY=snCfAPtMnM-32f3mm-K-Ckw@mail.gmail.com
Merge and hash joins can support antijoin with the non-nullable input
on the right, using very simple combinations of their existing logic
for right join and anti join. This gives the planner more freedom
about how to order the join. It's particularly useful for hash join,
since we may now have the option to hash the smaller table instead
of the larger.
Richard Guo, reviewed by Ronan Dunklau and myself
Discussion: https://postgr.es/m/CAMbWs48xh9hMzXzSy3VaPzGAz+fkxXXTUbCLohX1_L8THFRm2Q@mail.gmail.com
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
Avoid explicitly grouping by columns that we know are redundant
for sorting, for example we need group by only one of x and y in
SELECT ... WHERE x = y GROUP BY x, y
This comes up more often than you might think, as shown by the
changes in the regression tests. It's nearly free to detect too,
since we are just piggybacking on the existing logic that detects
redundant pathkeys. (In some of the existing plans that change,
it's visible that a sort step preceding the grouping step already
didn't bother to sort by the redundant column, making the old plan
a bit silly-looking.)
To do this, build processed_groupClause and processed_distinctClause
lists that omit any provably-redundant sort items, and consult those
not the originals where relevant. This means that within the
planner, one should usually consult root->processed_groupClause or
root->processed_distinctClause if one wants to know which columns
are to be grouped on; but to check whether grouping or distinct-ing
is happening at all, check non-NIL-ness of parse->groupClause or
parse->distinctClause. This is comparable to longstanding rules
about handling the HAVING clause, so I don't think it'll be a huge
maintenance problem.
nodeAgg.c also needs minor mods, because it's now possible to generate
AGG_PLAIN and AGG_SORTED Agg nodes with zero grouping columns.
Patch by me; thanks to Richard Guo and David Rowley for review.
Discussion: https://postgr.es/m/185315.1672179489@sss.pgh.pa.us
This allows left join removals and unique joins to work with partitioned
tables. The planner just lacked sufficient proofs that a given join
would not cause any row duplication. Unique indexes currently serve as
that proof, so have get_relation_info() populate the indexlist for
partitioned tables too.
Author: Arne Roland
Reviewed-by: Alvaro Herrera, Zhihong Yu, Amit Langote, David Rowley
Discussion: https://postgr.es/m/c3b2408b7a39433b8230bbcd02e9f302@index.de
When incremental sorts were added in v13 a 1.5x pessimism factor was added
to the cost modal. Seemingly this was done because the cost modal only
has an estimate of the total number of input rows and the number of
presorted groups. It assumes that the input rows will be evenly
distributed throughout the presorted groups. The 1.5x pessimism factor
was added to slightly reduce the likelihood of incremental sorts being
used in the hope to avoid performance regressions where an incremental
sort plan was picked and turned out slower due to a large skew in the
number of rows in the presorted groups.
An additional quirk with the path generation code meant that we could
consider both a sort and an incremental sort on paths with presorted keys.
This meant that with the pessimism factor, it was possible that we opted
to perform a sort rather than an incremental sort when the given path had
presorted keys.
Here we remove the 1.5x pessimism factor to allow incremental sorts to
have a fairer chance at being chosen against a full sort.
Previously we would generally create a sort path on the cheapest input
path (if that wasn't sorted already) and incremental sort paths on any
path which had presorted keys. This meant that if the cheapest input path
wasn't completely sorted but happened to have presorted keys, we would
create a full sort path *and* an incremental sort path on that input path.
Here we change this logic so that if there are presorted keys, we only
create an incremental sort path, and create sort paths only when a full
sort is required.
Both the removal of the cost pessimism factor and the changes made to the
path generation make it more likely that incremental sorts will now be
chosen. That, of course, as with teaching the planner any new tricks,
means an increased likelihood that the planner will perform an incremental
sort when it's not the best method. Our standard escape hatch for these
cases is an enable_* GUC. enable_incremental_sort already exists for
this.
This came out of a report by Pavel Luzanov where he mentioned that the
master branch was choosing to perform a Seq Scan -> Sort -> Group
Aggregate for his query with an ORDER BY aggregate function. The v15 plan
for his query performed an Index Scan -> Group Aggregate, of course, the
aggregate performed the final sort internally in nodeAgg.c for the
aggregate's ORDER BY. The ideal plan would have been to use the index,
which provided partially sorted input then use an incremental sort to
provide the aggregate with the sorted input. This was not being chosen
due to the pessimism in the incremental sort cost modal, so here we remove
that and rationalize the path generation so that sort and incremental sort
plans don't have to needlessly compete. We assume that it's senseless
to ever use a full sort on a given input path where an incremental sort
can be performed.
Reported-by: Pavel Luzanov
Reviewed-by: Richard Guo
Discussion: https://postgr.es/m/9f61ddbf-2989-1536-b31e-6459370a6baa%40postgrespro.ru
When we decide we need to make a derived clause equating a.x and
b.y, we already will re-use a previously-made clause "a.x = b.y".
But we might instead have "b.y = a.x", which is perfectly usable
because equivclass.c has never promised anything about the
operand order in clauses it builds. Saving construction of a
new RestrictInfo doesn't matter all that much in itself --- but
because we cache selectivity estimates and so on per-RestrictInfo,
there's a possibility of saving a fair amount of duplicative
effort downstream.
Hence, check for commutative matches as well as direct ones when
seeing if we have a pre-existing clause. This changes the visible
clause order in several regression test cases, but they're all
clearly-insignificant changes.
Checking for the reverse operand order is simple enough, but
if we wanted to check for operator OID match we'd need to call
get_commutator here, which is not so cheap. I concluded that
we don't really need the operator check anyway, so I just
removed it. It's unlikely that an opfamily contains more than
one applicable operator for a given pair of operand datatypes;
and if it does they had better give the same answers, so there
seems little need to insist that we use exactly the one
select_equality_operator chose.
Using the current core regression suite as a test case, I see
this change reducing the number of new join clauses built by
create_join_clause from 9673 to 5142 (out of 26652 calls).
So not quite 50% savings, but pretty close to it.
Discussion: https://postgr.es/m/78062.1666735746@sss.pgh.pa.us
This reverts commit db0d67db24 and
several follow-on fixes. The idea of making a cost-based choice
of the order of the sorting columns is not fundamentally unsound,
but it requires cost information and data statistics that we don't
really have. For example, relying on procost to distinguish the
relative costs of different sort comparators is pretty pointless
so long as most such comparator functions are labeled with cost 1.0.
Moreover, estimating the number of comparisons done by Quicksort
requires more than just an estimate of the number of distinct values
in the input: you also need some idea of the sizes of the larger
groups, if you want an estimate that's good to better than a factor of
three or so. That's data that's often unknown or not very reliable.
Worse, to arrive at estimates of the number of calls made to the
lower-order-column comparison functions, the code needs to make
estimates of the numbers of distinct values of multiple columns,
which are necessarily even less trustworthy than per-column stats.
Even if all the inputs are perfectly reliable, the cost algorithm
as-implemented cannot offer useful information about how to order
sorting columns beyond the point at which the average group size
is estimated to drop to 1.
Close inspection of the code added by db0d67db2 shows that there
are also multiple small bugs. These could have been fixed, but
there's not much point if we don't trust the estimates to be
accurate in-principle.
Finally, the changes in cost_sort's behavior made for very large
changes (often a factor of 2 or so) in the cost estimates for all
sorting operations, not only those for multi-column GROUP BY.
That naturally changes plan choices in many situations, and there's
precious little evidence to show that the changes are for the better.
Given the above doubts about whether the new estimates are really
trustworthy, it's hard to summon much confidence that these changes
are better on the average.
Since we're hard up against the release deadline for v15, let's
revert these changes for now. We can always try again later.
Note: in v15, I left T_PathKeyInfo in place in nodes.h even though
it's unreferenced. Removing it would be an ABI break, and it seems
a bit late in the release cycle for that.
Discussion: https://postgr.es/m/TYAPR01MB586665EB5FB2C3807E893941F5579@TYAPR01MB5866.jpnprd01.prod.outlook.com
When evaluating a query with a multi-column GROUP BY clause using sort,
the cost may be heavily dependent on the order in which the keys are
compared when building the groups. Grouping does not imply any ordering,
so we're allowed to compare the keys in arbitrary order, and a Hash Agg
leverages this. But for Group Agg, we simply compared keys in the order
as specified in the query. This commit explores alternative ordering of
the keys, trying to find a cheaper one.
In principle, we might generate grouping paths for all permutations of
the keys, and leave the rest to the optimizer. But that might get very
expensive, so we try to pick only a couple interesting orderings based
on both local and global information.
When planning the grouping path, we explore statistics (number of
distinct values, cost of the comparison function) for the keys and
reorder them to minimize comparison costs. Intuitively, it may be better
to perform more expensive comparisons (for complex data types etc.)
last, because maybe the cheaper comparisons will be enough. Similarly,
the higher the cardinality of a key, the lower the probability we’ll
need to compare more keys. The patch generates and costs various
orderings, picking the cheapest ones.
The ordering of group keys may interact with other parts of the query,
some of which may not be known while planning the grouping. E.g. there
may be an explicit ORDER BY clause, or some other ordering-dependent
operation, higher up in the query, and using the same ordering may allow
using either incremental sort or even eliminate the sort entirely.
The patch generates orderings and picks those minimizing the comparison
cost (for various pathkeys), and then adds orderings that might be
useful for operations higher up in the plan (ORDER BY, etc.). Finally,
it always keeps the ordering specified in the query, on the assumption
the user might have additional insights.
This introduces a new GUC enable_group_by_reordering, so that the
optimization may be disabled if needed.
The original patch was proposed by Teodor Sigaev, and later improved and
reworked by Dmitry Dolgov. Reviews by a number of people, including me,
Andrey Lepikhov, Claudio Freire, Ibrar Ahmed and Zhihong Yu.
Author: Dmitry Dolgov, Teodor Sigaev, Tomas Vondra
Reviewed-by: Tomas Vondra, Andrey Lepikhov, Claudio Freire, Ibrar Ahmed, Zhihong Yu
Discussion: https://postgr.es/m/7c79e6a5-8597-74e8-0671-1c39d124c9d6%40sigaev.ru
Discussion: https://postgr.es/m/CA%2Bq6zcW_4o2NC0zutLkOJPsFt80megSpX_dVRo6GK9PC-Jx_Ag%40mail.gmail.com
When building append paths, we've been looking only at startup and total
costs for the paths. When building fractional paths that may eliminate
the cheapest one, because it may be dominated by two separate paths (one
for startup, one for total cost).
This extends generate_orderedappend_paths() to also consider which paths
have lowest fractional cost. Currently we only consider paths matching
pathkeys - in the future this may be improved by also considering paths
that are only partially sorted, with an incremental sort on top.
Original report of an issue by Arne Roland, patch by me (based on a
suggestion by Tom Lane).
Reviewed-by: Arne Roland, Zhihong Yu
Discussion: https://postgr.es/m/e8f9ec90-546d-e948-acce-0525f3e92773%40enterprisedb.com
Discussion: https://postgr.es/m/1581042da8044e71ada2d6e3a51bf7bb%40index.de
This patch makes two closely related sets of changes:
1. For UPDATE, the subplan of the ModifyTable node now only delivers
the new values of the changed columns (i.e., the expressions computed
in the query's SET clause) plus row identity information such as CTID.
ModifyTable must re-fetch the original tuple to merge in the old
values of any unchanged columns. The core advantage of this is that
the changed columns are uniform across all tables of an inherited or
partitioned target relation, whereas the other columns might not be.
A secondary advantage, when the UPDATE involves joins, is that less
data needs to pass through the plan tree. The disadvantage of course
is an extra fetch of each tuple to be updated. However, that seems to
be very nearly free in context; even worst-case tests don't show it to
add more than a couple percent to the total query cost. At some point
it might be interesting to combine the re-fetch with the tuple access
that ModifyTable must do anyway to mark the old tuple dead; but that
would require a good deal of refactoring and it seems it wouldn't buy
all that much, so this patch doesn't attempt it.
2. For inherited UPDATE/DELETE, instead of generating a separate
subplan for each target relation, we now generate a single subplan
that is just exactly like a SELECT's plan, then stick ModifyTable
on top of that. To let ModifyTable know which target relation a
given incoming row refers to, a tableoid junk column is added to
the row identity information. This gets rid of the horrid hack
that was inheritance_planner(), eliminating O(N^2) planning cost
and memory consumption in cases where there were many unprunable
target relations.
Point 2 of course requires point 1, so that there is a uniform
definition of the non-junk columns to be returned by the subplan.
We can't insist on uniform definition of the row identity junk
columns however, if we want to keep the ability to have both
plain and foreign tables in a partitioning hierarchy. Since
it wouldn't scale very far to have every child table have its
own row identity column, this patch includes provisions to merge
similar row identity columns into one column of the subplan result.
In particular, we can merge the whole-row Vars typically used as
row identity by FDWs into one column by pretending they are type
RECORD. (It's still okay for the actual composite Datums to be
labeled with the table's rowtype OID, though.)
There is more that can be done to file down residual inefficiencies
in this patch, but it seems to be committable now.
FDW authors should note several API changes:
* The argument list for AddForeignUpdateTargets() has changed, and so
has the method it must use for adding junk columns to the query. Call
add_row_identity_var() instead of manipulating the parse tree directly.
You might want to reconsider exactly what you're adding, too.
* PlanDirectModify() must now work a little harder to find the
ForeignScan plan node; if the foreign table is part of a partitioning
hierarchy then the ForeignScan might not be the direct child of
ModifyTable. See postgres_fdw for sample code.
* To check whether a relation is a target relation, it's no
longer sufficient to compare its relid to root->parse->resultRelation.
Instead, check it against all_result_relids or leaf_result_relids,
as appropriate.
Amit Langote and Tom Lane
Discussion: https://postgr.es/m/CA+HiwqHpHdqdDn48yCEhynnniahH78rwcrv1rEX65-fsZGBOLQ@mail.gmail.com
reparameterize_path_by_child() failed to reparameterize BitmapAnd
and BitmapOr paths. This matters only if such a path is chosen as
the inside of a nestloop partition-wise join, where we have to pass
in parameters from the outside of the nestloop. If that did happen,
we generated a bad plan that would likely lead to crashes at execution.
This is not entirely reparameterize_path_by_child()'s fault though;
it's the victim of an ancient decision (my ancient decision, I think)
to not bother filling in param_info in BitmapAnd/Or path nodes. That
caused the function to believe that such nodes and their children
contain no parameter references and so need not be processed.
In hindsight that decision looks pretty penny-wise and pound-foolish:
while it saves a few cycles during path node setup, we do commonly
need the information later. In particular, by reversing the decision
and requiring valid param_info data in all nodes of a bitmap path
tree, we can get rid of indxpath.c's get_bitmap_tree_required_outer()
function, which computed the data on-demand. It's not unlikely that
that nets out as a savings of cycles in many scenarios. A couple
of other things in indxpath.c can be simplified as well.
While here, get rid of some cases in reparameterize_path_by_child()
that are visibly dead or useless, given that we only care about
reparameterizing paths that can be on the inside of a parameterized
nestloop. This case reminds one of the maxim that untested code
probably does not work, so I'm unwilling to leave unreachable code
in this function. (I did leave the T_Gather case in place even
though it's not reached in the regression tests. It's not very
clear to me when the planner might prefer to put Gather below
rather than above a nestloop, but at least in principle the case
might be interesting.)
Per bug #16536, originally from Arne Roland but with a test case
by Andrew Gierth. Back-patch to v11 where this code came in.
Discussion: https://postgr.es/m/16536-2213ee0b3aad41fd@postgresql.org
This case didn't work because columns merged by FULL JOIN USING are
represented in the parse tree by COALESCE expressions, and the logic
for recognizing a partitionable join failed to match upper-level join
clauses to such expressions. To fix, synthesize suitable COALESCE
expressions and add them to the nullable_partexprs lists. This is
pretty ugly and brute-force, but it gets the job done. (I have
ambitions of rethinking the way outer-join output Vars are
represented, so maybe that will provide a cleaner solution someday.
For now, do this.)
Amit Langote, reviewed by Justin Pryzby, Richard Guo, and myself
Discussion: https://postgr.es/m/CA+HiwqG2WVUGmLJqtR0tPFhniO=H=9qQ+Z3L_ZC+Y3-EVQHFGg@mail.gmail.com
Previously, the partitionwise join technique only allowed partitionwise
join when input partitioned tables had exactly the same partition
bounds. This commit extends the technique to some cases when the tables
have different partition bounds, by using an advanced partition-matching
algorithm introduced by this commit. For both the input partitioned
tables, the algorithm checks whether every partition of one input
partitioned table only matches one partition of the other input
partitioned table at most, and vice versa. In such a case the join
between the tables can be broken down into joins between the matching
partitions, so the algorithm produces the pairs of the matching
partitions, plus the partition bounds for the join relation, to allow
partitionwise join for computing the join. Currently, the algorithm
works for list-partitioned and range-partitioned tables, but not
hash-partitioned tables. See comments in partition_bounds_merge().
Ashutosh Bapat and Etsuro Fujita, most of regression tests by Rajkumar
Raghuwanshi, some of the tests by Mark Dilger and Amul Sul, reviewed by
Dmitry Dolgov and Amul Sul, with additional review at various points by
Ashutosh Bapat, Mark Dilger, Robert Haas, Antonin Houska, Amit Langote,
Justin Pryzby, and Tomas Vondra
Discussion: https://postgr.es/m/CAFjFpRdjQvaUEV5DJX3TW6pU5eq54NCkadtxHX2JiJG_GvbrCA@mail.gmail.com
This patch causes EXPLAIN to always assign a separate table alias to the
parent RTE of an append relation (inheritance set); before, such RTEs
were ignored if not actually scanned by the plan. Since the child RTEs
now always have that same alias to start with (cf. commit 55a1954da),
the net effect is that the parent RTE usually gets the alias used or
implied by the query text, and the children all get that alias with "_N"
appended. (The exception to "usually" is if there are duplicate aliases
in different subtrees of the original query; then some of those original
RTEs will also have "_N" appended.)
This results in more uniform output for partitioned-table plans than
we had before: the partitioned table itself gets the original alias,
and all child tables have aliases with "_N", rather than the previous
behavior where one of the children would get an alias without "_N".
The reason for giving the parent RTE an alias, even if it isn't scanned
by the plan, is that we now use the parent's alias to qualify Vars that
refer to an appendrel output column and appear above the Append or
MergeAppend that computes the appendrel. But below the append, Vars
refer to some one of the child relations, and are displayed that way.
This seems clearer than the old behavior where a Var that could carry
values from any child relation was displayed as if it referred to only
one of them.
While at it, change ruleutils.c so that the code paths used by EXPLAIN
deal in Plan trees not PlanState trees. This effectively reverts a
decision made in commit 1cc29fe7c, which seemed like a good idea at
the time to make ruleutils.c consistent with explain.c. However,
it's problematic because we'd really like to allow executor startup
pruning to remove all the children of an append node when possible,
leaving no child PlanState to resolve Vars against. (That's not done
here, but will be in the next patch.) This requires different handling
of subplans and initplans than before, but is otherwise a pretty
straightforward change.
Discussion: https://postgr.es/m/001001d4f44b$2a2cca50$7e865ef0$@lab.ntt.co.jp
If an inheritance/partitioning parent table is assigned some column
alias names in the query, EXPLAIN mapped those aliases onto the
child tables' columns by physical position, resulting in bogus output
if a child table's columns aren't one-for-one with the parent's.
To fix, make expand_single_inheritance_child() generate a correctly
re-mapped column alias list, rather than just copying the parent
RTE's alias node. (We have to fill the alias field, not just
adjust the eref field, because ruleutils.c will ignore eref in
favor of looking at the real column names.)
This means that child tables will now always have alias fields in
plan rtables, where before they might not have. That results in
a rather substantial set of regression test output changes:
EXPLAIN will now always show child tables with aliases that match
the parent table (usually with "_N" appended for uniqueness).
But that seems like a net positive for understandability, since
the parent alias corresponds to something that actually appeared
in the original query, while the child table names didn't.
(Note that this does not change anything for cases where an explicit
table alias was written in the query for the parent table; it
just makes cases without such aliases behave similarly to that.)
Hence, while we could avoid these subsidiary changes if we made
inherit.c more complicated, we choose not to.
Discussion: https://postgr.es/m/12424.1575168015@sss.pgh.pa.us
Commit d25ea0127 got rid of what I thought were entirely unnecessary
derived child expressions in EquivalenceClasses for EC members that
mention multiple baserels. But it turns out that some of the child
expressions that code created are necessary for partitionwise joins,
else we fail to find matching pathkeys for Sort nodes. (This happens
only for certain shapes of the resulting plan; it may be that
partitionwise aggregation is also necessary to show the failure,
though I'm not sure of that.)
Reverting that commit entirely would be quite painful performance-wise
for large partition sets. So instead, add code that explicitly
generates child expressions that match only partitionwise child join
rels we have actually generated.
Per report from Justin Pryzby. (Amit Langote noticed the problem
earlier, though it's not clear if he recognized then that it could
result in a planner error, not merely failure to exploit partitionwise
join, in the code as-committed.) Back-patch to v12 where commit
d25ea0127 came in.
Amit Langote, with lots of kibitzing from me
Discussion: https://postgr.es/m/CA+HiwqG2WVUGmLJqtR0tPFhniO=H=9qQ+Z3L_ZC+Y3-EVQHFGg@mail.gmail.com
Discussion: https://postgr.es/m/20191011143703.GN10470@telsasoft.com
While trying to plan a partitionwise join, we may be faced with cases
where one or both input partitions for a particular segment of the join
have been pruned away. In HEAD and v11, this is problematic because
earlier processing didn't bother to make a pruned RelOptInfo fully
valid. With an upcoming patch to make partition pruning more efficient,
this'll be even more problematic because said RelOptInfo won't exist at
all.
The existing code attempts to deal with this by retroactively making the
RelOptInfo fully valid, but that causes crashes under GEQO because join
planning is done in a short-lived memory context. In v11 we could
probably have fixed this by switching to the planner's main context
while fixing up the RelOptInfo, but that idea doesn't scale well to the
upcoming patch. It would be better not to mess with the base-relation
data structures during join planning, anyway --- that's just a recipe
for order-of-operations bugs.
In many cases, though, we don't actually need the child RelOptInfo,
because if the input is certainly empty then the join segment's result
is certainly empty, so we can skip making a join plan altogether. (The
existing code ultimately arrives at the same conclusion, but only after
doing a lot more work.) This approach works except when the pruned-away
partition is on the nullable side of a LEFT, ANTI, or FULL join, and the
other side isn't pruned. But in those cases the existing code leaves a
lot to be desired anyway --- the correct output is just the result of
the unpruned side of the join, but we were emitting a useless outer join
against a dummy Result. Pending somebody writing code to handle that
more nicely, let's just abandon the partitionwise-join optimization in
such cases.
When the modified code skips making a join plan, it doesn't make a
join RelOptInfo either; this requires some upper-level code to
cope with nulls in part_rels[] arrays. We would have had to have
that anyway after the upcoming patch.
Back-patch to v11 since the crash is demonstrable there.
Discussion: https://postgr.es/m/8305.1553884377@sss.pgh.pa.us
If there's only one child relation, the Append or MergeAppend isn't
doing anything useful, and can be elided. It does have a purpose
during planning though, which is to serve as a buffer between parent
and child Var numbering. Therefore we keep it all the way through
to setrefs.c, and get rid of it only after fixing references in the
plan level(s) above it. This works largely the same as setrefs.c's
ancient hack to get rid of no-op SubqueryScan nodes, and can even
share some code with that.
Note the change to make setrefs.c use apply_tlist_labeling rather than
ad-hoc code. This has the effect of propagating the child's resjunk
and ressortgroupref labels, which formerly weren't propagated when
removing a SubqueryScan. Doing that is demonstrably necessary for
the [Merge]Append cases, and seems harmless for SubqueryScan, if only
because trivial_subqueryscan is afraid to collapse cases where the
resjunk marking differs. (I suspect that restriction could now be
removed, though it's unclear that it'd make any new matches possible,
since the outer query can't have references to a child resjunk column.)
David Rowley, reviewed by Alvaro Herrera and Tomas Vondra
Discussion: https://postgr.es/m/CAKJS1f_7u8ATyJ1JGTMHFoKDvZdeF-iEBhs+sM_SXowOr9cArg@mail.gmail.com
In place of three separate but interrelated lists (indexclauses,
indexquals, and indexqualcols), an IndexPath now has one list
"indexclauses" of IndexClause nodes. This holds basically the same
information as before, but in a more useful format: in particular, there
is now a clear connection between an indexclause (an original restriction
clause from WHERE or JOIN/ON) and the indexquals (directly usable index
conditions) derived from it.
We also change the ground rules a bit by mandating that clause commutation,
if needed, be done up-front so that what is stored in the indexquals list
is always directly usable as an index condition. This gets rid of repeated
re-determination of which side of the clause is the indexkey during costing
and plan generation, as well as repeated lookups of the commutator
operator. To minimize the added up-front cost, the typical case of
commuting a plain OpExpr is handled by a new special-purpose function
commute_restrictinfo(). For RowCompareExprs, generating the new clause
properly commuted to begin with is not really any more complex than before,
it's just different --- and we can save doing that work twice, as the
pretty-klugy original implementation did.
Tracking the connection between original and derived clauses lets us
also track explicitly whether the derived clauses are an exact or lossy
translation of the original. This provides a cheap solution to getting
rid of unnecessary rechecks of boolean index clauses, which previously
seemed like it'd be more expensive than it was worth.
Another pleasant (IMO) side-effect is that EXPLAIN now always shows
index clauses with the indexkey on the left; this seems less confusing.
This commit leaves expand_indexqual_conditions() and some related
functions in a slightly messy state. I didn't bother to change them
any more than minimally necessary to work with the new data structure,
because all that code is going to be refactored out of existence in
a follow-on patch.
Discussion: https://postgr.es/m/22182.1549124950@sss.pgh.pa.us
We should never estimate the output of a semijoin to be more rows than
we estimate for an inner join with the same input rels and join condition;
it's obviously impossible for that to happen. However, given the
relatively poor quality of our semijoin selectivity estimates ---
particularly, but not only, in cases where we punt and return a default
estimate --- we did often deliver such estimates. To improve matters,
calculate both estimates inside eqjoinsel() and take the smaller one.
The bulk of this patch is just mechanical refactoring to avoid repetitive
information lookup when we call both eqjoinsel_semi and eqjoinsel_inner.
The actual new behavior is just
selec = Min(selec, inner_rel->rows * selec_inner);
which looks a bit odd but is correct because of our different definitions
for inner and semi join selectivity.
There is one ensuing plan change in the regression tests, but it looks
reasonable enough (and checking the actual row counts shows that the
estimate moved closer to reality, not further away).
Per bug #15160 from Alexey Ermakov. Although this is arguably a bug fix,
I won't risk destabilizing plan choices in stable branches by
back-patching.
Tom Lane, reviewed by Melanie Plageman
Discussion: https://postgr.es/m/152395805004.19366.3107109716821067806@wrigleys.postgresql.org
Commit f49842d, which added support for partitionwise joins, built the
child's tlist by applying adjust_appendrel_attrs() to the parent's. So in
the case where the parent's included a whole-row Var for the parent, the
child's contained a ConvertRowtypeExpr. To cope with that, that commit
added code to the planner, such as setrefs.c, but some code paths still
assumed that the tlist for a scan (or join) rel would only include Vars
and PlaceHolderVars, which was true before that commit, causing errors:
* When creating an explicit sort node for an input path for a mergejoin
path for a child join, prepare_sort_from_pathkeys() threw the 'could not
find pathkey item to sort' error.
* When deparsing a relation participating in a pushed down child join as a
subquery in contrib/postgres_fdw, get_relation_column_alias_ids() threw
the 'unexpected expression in subquery output' error.
* When performing set_plan_references() on a local join plan generated by
contrib/postgres_fdw for EvalPlanQual support for a pushed down child
join, fix_join_expr() threw the 'variable not found in subplan target
lists' error.
To fix these, two approaches have been proposed: one by Ashutosh Bapat and
one by me. While the former keeps building the child's tlist with a
ConvertRowtypeExpr, the latter builds it with a whole-row Var for the
child not to violate the planner assumption, and tries to fix it up later,
But both approaches need more work, so refuse to generate partitionwise
join paths when whole-row Vars are involved, instead. We don't need to
handle ConvertRowtypeExprs in the child's tlists for now, so this commit
also removes the changes to the planner.
Previously, partitionwise join computed attr_needed data for each child
separately, and built the child join's tlist using that data, which also
required an extra step for adding PlaceHolderVars to that tlist, but it
would be more efficient to build it from the parent join's tlist through
the adjust_appendrel_attrs() transformation. So this commit builds that
list that way, and simplifies build_joinrel_tlist() and placeholder.c as
well as part of set_append_rel_size() to basically what they were before
partitionwise join went in.
Back-patch to PG11 where partitionwise join was introduced.
Report by Rajkumar Raghuwanshi. Analysis by Ashutosh Bapat, who also
provided some of regression tests. Patch by me, reviewed by Robert Haas.
Discussion: https://postgr.es/m/CAKcux6ktu-8tefLWtQuuZBYFaZA83vUzuRd7c1YHC-yEWyYFpg@mail.gmail.com
When considering a partitioning parent rel, we should stop processing that
subroot as soon as we've done adjust_appendrel_attrs and any securityQuals
updates. The rest of this is unnecessary, and indeed adding duplicate
subquery RTEs to the subroot is *wrong*. As the code stood, the children
of that partition ended up with two sets of copied subquery RTEs, confusing
matters greatly. Even more hilarity ensued if all of the children got
excluded by constraint exclusion, so that the extra RTEs didn't make it
back into the parent rtable.
Per fuzz testing by Andreas Seltenreich. Back-patch to v11 where this
got broken (by commit 0a480502b, it looks like).
Discussion: https://postgr.es/m/87va8g7vq0.fsf@ansel.ydns.eu
If the toplevel scan/join target list is parallel-safe, postpone
generating Gather (or Gather Merge) paths until after the toplevel has
been adjusted to return it. This (correctly) makes queries with
expensive functions in the target list more likely to choose a
parallel plan, since the cost of the plan now reflects the fact that
the evaluation will happen in the workers rather than the leader.
The original complaint about this problem was from Jeff Janes.
If the toplevel scan/join relation is partitioned, recursively apply
the changes to all partitions. This sometimes allows us to get rid of
Result nodes, because Append is not projection-capable but its
children may be. It also cleans up what appears to be incorrect SRF
handling from commit e2f1eb0ee30d144628ab523432320f174a2c8966: the old
code had no knowledge of SRFs for child scan/join rels.
Because we now use create_projection_path() in some cases where we
formerly used apply_projection_to_path(), this changes the ordering
of columns in some queries generated by postgres_fdw. Update
regression outputs accordingly.
Patch by me, reviewed by Amit Kapila and by Ashutosh Bapat. Other
fixes for this problem (substantially different from this version)
were reviewed by Dilip Kumar, Amit Khandekar, and Marina Polyakova.
Discussion: http://postgr.es/m/CAMkU=1ycXNipvhWuweUVpKuyu6SpNjF=yHWu4c4US5JgVGxtZQ@mail.gmail.com
Commit 7d8ac9814b adjusted these
tests in the hope of preserving the plan shape, but I failed to
notice that the three partitions were, on my local machine, choosing
two different plan shapes. This is probably related to the fact
that all three tables have exactly the same row count. Try to
improve the situation by making pht1_e about half as large as
the other two.
Per Tom Lane and the buildfarm.
Discussion: http://postgr.es/m/25380.1519277713@sss.pgh.pa.us
Previously, Append didn't charge anything at all, and MergeAppend
charged only cpu_operator_cost, about half the value used here. This
change might make MergeAppend plans slightly more likely to be chosen
than before, since this commit increases the assumed cost for Append
-- with default values -- by 0.005 per tuple but MergeAppend by only
0.0025 per tuple. Since the comparisons required by MergeAppend are
costed separately, it's not clear why MergeAppend needs to be
otherwise more expensive than Append, so hopefully this is OK.
Prior to partition-wise join, it didn't really matter whether or not
an Append node had any cost of its own, because every plan had to use
the same number of Append or MergeAppend nodes and in the same places.
Only the relative cost of Append vs. MergeAppend made a difference.
Now, however, it is possible to avoid some of the Append nodes using a
partition-wise join, so it's worth making an effort. Pending patches
for partition-wise aggregate care too, because an Append of Aggregate
nodes will incur the Append overhead fewer times than an Aggregate
over an Append. Although in most cases this change will favor the use
of partition-wise techniques, it does the opposite when the join
cardinality is greater than the sum of the input cardinalities. Since
this situation arises in an existing regression test, I [rhaas]
adjusted it to keep the overall plan shape approximately the same.
Jeevan Chalke, per a suggestion from David Rowley. Reviewed by
Ashutosh Bapat. Some changes by me. The larger patch series of which
this patch is a part was also reviewed and tested by Antonin Houska,
Rajkumar Raghuwanshi, David Rowley, Dilip Kumar, Konstantin Knizhnik,
Pascal Legrand, Rafia Sabih, and me.
Discussion: http://postgr.es/m/CAKJS1f9UXdk6ZYyqbJnjFO9a9hyHKGW7B=ZRh-rxy9qxfPA5Gw@mail.gmail.com
The previous code assumed that we'd always succeed in creating
child-joins for a joinrel for which partition-wise join was considered,
but that's not guaranteed, at least in the case where dummy rels
are involved.
Ashutosh Bapat, with some wordsmithing by me.
Discussion: http://postgr.es/m/CAFjFpRf8=uyMYYfeTBjWDMs1tR5t--FgOe2vKZPULxxdYQ4RNw@mail.gmail.com
Hash partitioning is useful when you want to partition a growing data
set evenly. This can be useful to keep table sizes reasonable, which
makes maintenance operations such as VACUUM faster, or to enable
partition-wise join.
At present, we still depend on constraint exclusion for partitioning
pruning, and the shape of the partition constraints for hash
partitioning is such that that doesn't work. Work is underway to fix
that, which should both improve performance and make partitioning
pruning work with hash partitioning.
Amul Sul, reviewed and tested by Dilip Kumar, Ashutosh Bapat, Yugo
Nagata, Rajkumar Raghuwanshi, Jesper Pedersen, and by me. A few
final tweaks also by me.
Discussion: http://postgr.es/m/CAAJ_b96fhpJAP=ALbETmeLk1Uni_GFZD938zgenhF49qgDTjaQ@mail.gmail.com
Instead of joining two partitioned tables in their entirety we can, if
it is an equi-join on the partition keys, join the matching partitions
individually. This involves teaching the planner about "other join"
rels, which are related to regular join rels in the same way that
other member rels are related to baserels. This can use significantly
more CPU time and memory than regular join planning, because there may
now be a set of "other" rels not only for every base relation but also
for every join relation. In most practical cases, this probably
shouldn't be a problem, because (1) it's probably unusual to join many
tables each with many partitions using the partition keys for all
joins and (2) if you do that scenario then you probably have a big
enough machine to handle the increased memory cost of planning and (3)
the resulting plan is highly likely to be better, so what you spend in
planning you'll make up on the execution side. All the same, for now,
turn this feature off by default.
Currently, we can only perform joins between two tables whose
partitioning schemes are absolutely identical. It would be nice to
cope with other scenarios, such as extra partitions on one side or the
other with no match on the other side, but that will have to wait for
a future patch.
Ashutosh Bapat, reviewed and tested by Rajkumar Raghuwanshi, Amit
Langote, Rafia Sabih, Thomas Munro, Dilip Kumar, Antonin Houska, Amit
Khandekar, and by me. A few final adjustments by me.
Discussion: http://postgr.es/m/CAFjFpRfQ8GrQvzp3jA2wnLqrHmaXna-urjm_UY9BqXj=EaDTSA@mail.gmail.com
Discussion: http://postgr.es/m/CAFjFpRcitjfrULr5jfuKWRPsGUX0LQ0k8-yG0Qw2+1LBGNpMdw@mail.gmail.com
Tom Lane
David Rowley
Tomas Vondra
Etsuro Fujita
Jeff Davis
Robert Haas
Peter Eisentraut