Commit cbc127917e introduced tracking of unpruned relids to avoid
processing pruned relations, and changed ExecInitModifyTable() to
initialize only unpruned result relations. As a result, MERGE
statements that prune all target partitions can now lead to crashes
or incorrect behavior during execution.
The crash occurs because some executor code paths rely on
ModifyTableState.resultRelInfo[0] being present and initialized,
even when no result relations remain after pruning. For example,
ExecMerge() and ExecMergeNotMatched() use the first resultRelInfo
to determine the appropriate action. Similarly,
ExecInitPartitionInfo() assumes that at least one result relation
exists.
To preserve these assumptions, ExecInitModifyTable() now includes the
first result relation in the initialized result relation list if all
result relations for that ModifyTable were pruned. To enable that,
ExecDoInitialPruning() ensures the first relation is locked if it was
pruned and locking is necessary.
To support this exception to the pruning logic, PlannedStmt now
includes a list of RT indexes identifying the first result relation
of each ModifyTable node in the plan. This allows
ExecDoInitialPruning() to check whether each such relation was
pruned and, if so, lock it if necessary.
Bug: #18830
Reported-by: Robins Tharakan <tharakan@gmail.com>
Diagnozed-by: Tender Wang <tndrwang@gmail.com>
Diagnozed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Co-authored-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Discussion: https://postgr.es/m/18830-1f31ea1dc930d444%40postgresql.org
After pushing the bitmap iterator into table-AM specific code (as part
of making bitmap heap scan use the read stream API in 2b73a8cd33),
scan_bitmap_next_block() no longer returns the current block number.
Since scan_bitmap_next_block() isn't returning any relevant information
to bitmap table scan code, it makes more sense to get rid of it.
Now, bitmap table scan code only calls table_scan_bitmap_next_tuple(),
and the heap AM implementation of scan_bitmap_next_block() is a local
helper in heapam_handler.c.
Reviewed-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/flat/CAAKRu_ZwCwWFeL_H3ia26bP2e7HiKLWt0ZmGXPVwPO6uXq0vaA%40mail.gmail.com
When pulling up a subquery, if the subquery's target list items are
used in grouping set columns, we need to wrap them in PlaceHolderVars.
This ensures that expressions retain their separate identity so that
they will match grouping set columns when appropriate.
In 90947674f, we decided to wrap subquery outputs that are non-var
expressions in PlaceHolderVars. This prevents const-simplification
from merging them into the surrounding expressions after subquery
pullup, which could otherwise lead to failing to match those
subexpressions to grouping set columns, with the effect that they'd
not go to null when expected.
However, that left some loose ends. If the subquery's target list
contains two or more identical Var expressions, we can still fail to
match the Var expression to the expected grouping set expression.
This is not related to const-simplification, but rather to how we
match expressions to lower target items in setrefs.c.
For sort/group expressions, we use ressortgroupref matching, which
works well. For other expressions, we primarily rely on comparing the
expressions to determine if they are the same. Therefore, we need a
way to prevent setrefs.c from matching the expression to some other
identical ones.
To fix, wrap all subquery outputs in PlaceHolderVars if the parent
query uses grouping sets, ensuring that they preserve their separate
identity throughout the whole planning process.
Reported-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Discussion: https://postgr.es/m/CAMbWs4-meSahaanKskpBn0KKxdHAXC1_EJCVWHxEodqirrGJnw@mail.gmail.com
In pull_up_simple_subquery and pull_up_constant_function, there is
code that sets wrap_non_vars to true when dealing with an appendrel
member. The goal is to wrap subquery outputs that are not simple Vars
in PlaceHolderVars, ensuring that what we pull up doesn't get merged
into a surrounding expression during later processing, which could
cause it to fail to match the expression actually available from the
appendrel.
However, this is unnecessary. When pulling up an appendrel child
subquery, the only part of the upper query that could reference the
appendrel child yet is the translated_vars list of the associated
AppendRelInfo that we just made for this child. Furthermore, we do
not want to force use of PHVs in the AppendRelInfo, as there is no
outer join between. In fact, perform_pullup_replace_vars always sets
wrap_non_vars to false before performing pullup_replace_vars on the
AppendRelInfo.
This patch simply removes the code that sets wrap_non_vars to true for
UNION ALL subqueries.
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Discussion: https://postgr.es/m/CAMbWs4-VXDEi1v+hZYLxpOv0riJxHsCkCH1f46tLnhonEAyGCQ@mail.gmail.com
makeWholeRowVar() has different rules for constructing a
whole-row Var depending on the kind of RTE it's representing.
This turns out to be problematic because the rewriter and planner
can convert view RTEs and set-returning-function RTEs into
subquery RTEs; so a whole-row Var made during planning might
look different from one made by the parser. In isolation this
doesn't cause any problem, but if a query contains Vars made
both ways for the same varno, there are cross-checks in the
executor that will complain. This manifests for UPDATE, DELETE,
and MERGE queries that use whole-row table references.
To fix, we need makeWholeRowVar() to produce the same result
from an inlined RTE as it would have for the original. For
an inlined view, we can use RangeTblEntry.relid to detect
that this had been a view RTE. For inlined SRFs, make a
data structure definition change akin to commit 47bb9db75,
and say that we won't clear RangeTblEntry.functions until
the end of planning. That allows makeWholeRowVar() to
repeat what it would have done with the unmodified RTE.
Reported-by: Duncan Sands <duncan.sands@deepbluecap.com>
Reported-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Diagnosed-by: Tender Wang <tndrwang@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Discussion: https://postgr.es/m/3518c50a-ab18-482f-b916-a37263622501@deepbluecap.com
Backpatch-through: 13
Up to now we just punted on showing the window definitions used
in a plan, with window function calls represented as "OVER (?)".
To improve that, show the window definition implemented by each
WindowAgg plan node, and reference their window names in OVER.
For nameless window clauses generated by "OVER (...)", assign
unique names w1, w2, etc.
In passing, re-order the properties shown for a WindowAgg node
so that the Run Condition (if any) appears after the Window
property and before the Filter (if any). This seems more
sensible since the Run Condition is associated with the Window
and acts before the Filter.
Thanks to David G. Johnston and Álvaro Herrera for design
suggestions.
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: David Rowley <dgrowleyml@gmail.com>
Discussion: https://postgr.es/m/144530.1741469955@sss.pgh.pa.us
Recognizing the real-life complexity where columns in the table often have
functional dependencies, PostgreSQL's estimation of the number of distinct
values over a set of columns can be underestimated (or much rarely,
overestimated) when dealing with multi-clause JOIN. In the case of hash
join, it can end up with a small number of predicted hash buckets and, as
a result, picking non-optimal merge join.
To improve the situation, we introduce one additional stage of bucket size
estimation - having two or more join clauses estimator lookup for extended
statistics and use it for multicolumn estimation. Clauses are grouped into
lists, each containing expressions referencing the same relation. The result
of the multicolumn estimation made over such a list is combined with others
according to the caller's logic. Clauses that are not estimated are returned
to the caller for further estimation.
Discussion: https://postgr.es/m/52257607-57f6-850d-399a-ec33a654457b%40postgrespro.ru
Author: Andrei Lepikhov <lepihov@gmail.com>
Reviewed-by: Andy Fan <zhihui.fan1213@gmail.com>
Reviewed-by: Tomas Vondra <tomas.vondra@enterprisedb.com>
Reviewed-by: Alena Rybakina <lena.ribackina@yandex.ru>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
This change is dedicated to more active usage of IndexScan and parameterized
NestLoop paths in partitioned cases under an Append node, as it already works
with plain tables. As newly added regression tests demonstrate, it should
provide more smartness to the partitionwise technique.
With an indication of how many tuples are needed, it may be more meaningful
to use the 'fractional branch' subpaths of the Append path list, which are
more optimal for this specific number of tuples. Planning on a higher level,
if the optimizer needs all the tuples, it will choose non-fractional paths.
In the case when, during execution, Append needs to return fewer tuples than
declared by tuple_fraction, it would not be harmful to use the 'intermediate'
variant of paths. However, it will earn a considerable profit if a sensible
set of tuples is selected.
The change of the existing regression test demonstrates the positive outcome
of this feature: instead of scanning the whole table, the optimizer prefers
to use a parameterized scan, being aware of the only single tuple the join
has to produce to perform the query.
Discussion: https://www.postgresql.org/message-id/flat/CAN-LCVPxnWB39CUBTgOQ9O7Dd8DrA_tpT1EY3LNVnUuvAX1NjA%40mail.gmail.com
Author: Nikita Malakhov <hukutoc@gmail.com>
Author: Andrei Lepikhov <lepihov@gmail.com>
Reviewed-by: Andy Fan <zhihuifan1213@163.com>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
In commit b262ad440, we introduced an optimization that reduces an IS
NOT NULL qual on a column defined as NOT NULL to constant true, and an
IS NULL qual on a NOT NULL column to constant false, provided we can
prove that the input expression of the NullTest is not nullable by any
outer join. This deduction happens after we have generated multiple
clones of the same qual condition to cope with commuted-left-join
cases.
However, performing the NullTest deduction for clone clauses can be
unsafe, because we don't have a reliable way to determine if the input
expression of a NullTest is non-nullable: nullingrel bits in clone
clauses may not reflect reality, so we dare not draw conclusions from
clones about whether Vars are guaranteed not-null.
To fix, we check whether the given RestrictInfo is a clone clause in
restriction_is_always_true and restriction_is_always_false, and avoid
performing any reduction if it is.
There are several ensuing plan changes in predicate.out, and we have
to modify the tests to ensure that they continue to test what they are
intended to. Additionally, this fix causes the test case added in
f00ab1fd1 to no longer trigger the bug that commit fixed, so we also
remove that test case.
Back-patch to v17 where this bug crept in.
Reported-by: Ronald Cruz <cruz@rentec.com>
Diagnosed-by: Tom Lane <tgl@sss.pgh.pa.us>
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/f5320d3d-77af-4ce8-b9c3-4715ff33f213@rentec.com
Backpatch-through: 17
As spotted by Coverity, the calculation of ojrelid mixes signed and unsigned
types causes possible overflow and undefined behavior. Instead of trying to
fix the expression, this commit eliminates the relied local variable. The
explicit branching is used to replace the -1 value. That, in turn, requires
changing the signature of the remove_rel_from_eclass() function.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/914330.1740330169%40sss.pgh.pa.us
Reviewed-by: Andrei Lepikhov <lepihov@gmail.com>
The callback functions ReplaceVarsFromTargetList_callback and
pullup_replace_vars_callback are both used to replace Vars in an
expression tree that reference a particular RTE with items from a
targetlist, and they both need to expand whole-tuple references and
deal with OLD/NEW RETURNING list Vars. As a result, currently there
is significant code duplication between these two functions.
This patch introduces a new function, ReplaceVarFromTargetList, to
perform the replacement and calls it from both callback functions,
thereby eliminating code duplication.
Author: Dean Rasheed <dean.a.rasheed@gmail.com>
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Jian He <jian.universality@gmail.com>
Discussion: https://postgr.es/m/CAEZATCWhr=FM4X5kCPvVs-g2XEk+ceLsNtBK_zZMkqFn9vUjsw@mail.gmail.com
Commit 83ea6c540 added support for virtual generated columns that are
computed on read. All Var nodes in the query that reference virtual
generated columns must be replaced with the corresponding generation
expressions. Currently, this replacement occurs in the rewriter.
However, this approach has several issues. If a Var referencing a
virtual generated column has any varnullingrels, those varnullingrels
need to be propagated into the generation expression. Failing to do
so can lead to "wrong varnullingrels" errors and improper outer-join
removal.
Additionally, if such a Var comes from the nullable side of an outer
join, we may need to wrap the generation expression in a
PlaceHolderVar to ensure that it is evaluated at the right place and
hence is forced to null when the outer join should do so. In certain
cases, such as when the query uses grouping sets, we also need a
PlaceHolderVar for anything that is not a simple Var to isolate
subexpressions. Failure to do so can result in incorrect results.
To fix these issues, this patch expands the virtual generated columns
in the planner rather than in the rewriter, and leverages the
pullup_replace_vars architecture to avoid code duplication. The
generation expressions will be correctly marked with nullingrel bits
and wrapped in PlaceHolderVars when needed by the pullup_replace_vars
callback function. This requires handling the OLD/NEW RETURNING list
Vars in pullup_replace_vars_callback, as it may now deal with Vars
referencing the result relation instead of a subquery.
The "wrong varnullingrels" error was reported by Alexander Lakhin.
The incorrect result issue and the improper outer-join removal issue
were reported by Richard Guo.
Author: Richard Guo <guofenglinux@gmail.com>
Author: Dean Rasheed <dean.a.rasheed@gmail.com>
Reviewed-by: Jian He <jian.universality@gmail.com>
Discussion: https://postgr.es/m/75eb1a6f-d59f-42e6-8a78-124ee808cda7@gmail.com
In try_partitionwise_join, we try to break down the join between two
partitioned relations into joins between matching partitions. To
achieve this, we iterate through each pair of partitions from the two
joining relations and create child join relations for them. To reduce
memory accumulation during each iteration, one step we take is freeing
the SpecialJoinInfos created for the child joins.
A child join's SpecialJoinInfo is a copy of the parent join's
SpecialJoinInfo, with some members being translated copies of their
counterparts in the parent. However, when freeing the bitmapset
members in a child join's SpecialJoinInfo, we failed to check whether
they were translated copies. As a result, we inadvertently freed the
members that were still in use by the parent SpecialJoinInfo, leading
to crashes when those freed members were accessed.
To fix, check if each member of the child join's SpecialJoinInfo is a
translated copy and free it only if that's the case. This requires
passing the parent join's SpecialJoinInfo as a parameter to
free_child_join_sjinfo.
Back-patch to v17 where this bug crept in.
Bug: #18806
Reported-by: 孟令彬 <m_lingbin@126.com>
Diagnosed-by: Tender Wang <tndrwang@gmail.com>
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Amit Langote <amitlangote09@gmail.com>
Reviewed-by: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>
Discussion: https://postgr.es/m/18806-d70b0c9fdf63dcbf@postgresql.org
Backpatch-through: 17
The Self-Join Elimination (SJE) feature removes an inner join of a plain
table to itself in the query tree if it is proven that the join can be
replaced with a scan without impacting the query result. Self-join and
inner relation get replaced with the outer in query, equivalence classes,
and planner info structures. Also, the inner restrictlist moves to the
outer one with the removal of duplicated clauses. Thus, this optimization
reduces the length of the range table list (this especially makes sense for
partitioned relations), reduces the number of restriction clauses and,
in turn, selectivity estimations, and potentially improves total planner
prediction for the query.
This feature is dedicated to avoiding redundancy, which can appear after
pull-up transformations or the creation of an EquivalenceClass-derived clause
like the below.
SELECT * FROM t1 WHERE x IN (SELECT t3.x FROM t1 t3);
SELECT * FROM t1 WHERE EXISTS (SELECT t3.x FROM t1 t3 WHERE t3.x = t1.x);
SELECT * FROM t1,t2, t1 t3 WHERE t1.x = t2.x AND t2.x = t3.x;
In the future, we could also reduce redundancy caused by subquery pull-up
after unnecessary outer join removal in cases like the one below.
SELECT * FROM t1 WHERE x IN
(SELECT t3.x FROM t1 t3 LEFT JOIN t2 ON t2.x = t1.x);
Also, it can drastically help to join partitioned tables, removing entries
even before their expansion.
The SJE proof is based on innerrel_is_unique() machinery.
We can remove a self-join when for each outer row:
1. At most, one inner row matches the join clause;
2. Each matched inner row must be (physically) the same as the outer one;
3. Inner and outer rows have the same row mark.
In this patch, we use the next approach to identify a self-join:
1. Collect all merge-joinable join quals which look like a.x = b.x;
2. Add to the list above the baseretrictinfo of the inner table;
3. Check innerrel_is_unique() for the qual list. If it returns false, skip
this pair of joining tables;
4. Check uniqueness, proved by the baserestrictinfo clauses. To prove the
possibility of self-join elimination, the inner and outer clauses must
match exactly.
The relation replacement procedure is not trivial and is partly combined
with the one used to remove useless left joins. Tests covering this feature
were added to join.sql. Some of the existing regression tests changed due
to self-join removal logic.
Discussion: https://postgr.es/m/flat/64486b0b-0404-e39e-322d-0801154901f3%40postgrespro.ru
Author: Andrey Lepikhov <a.lepikhov@postgrespro.ru>
Author: Alexander Kuzmenkov <a.kuzmenkov@postgrespro.ru>
Co-authored-by: Alexander Korotkov <aekorotkov@gmail.com>
Co-authored-by: Alena Rybakina <lena.ribackina@yandex.ru>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Simon Riggs <simon@2ndquadrant.com>
Reviewed-by: Jonathan S. Katz <jkatz@postgresql.org>
Reviewed-by: David Rowley <david.rowley@2ndquadrant.com>
Reviewed-by: Thomas Munro <thomas.munro@enterprisedb.com>
Reviewed-by: Konstantin Knizhnik <k.knizhnik@postgrespro.ru>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Hywel Carver <hywel@skillerwhale.com>
Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at>
Reviewed-by: Ronan Dunklau <ronan.dunklau@aiven.io>
Reviewed-by: vignesh C <vignesh21@gmail.com>
Reviewed-by: Zhihong Yu <zyu@yugabyte.com>
Reviewed-by: Greg Stark <stark@mit.edu>
Reviewed-by: Jaime Casanova <jcasanov@systemguards.com.ec>
Reviewed-by: Michał Kłeczek <michal@kleczek.org>
Reviewed-by: Alena Rybakina <lena.ribackina@yandex.ru>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
In set_append_rel_size(), we currently set rel->tuples to rel->rows
for an appendrel. Generally, rel->tuples is the raw number of tuples
in the relation and rel->rows is the estimated number of tuples after
the relation's restriction clauses have been applied. Although an
appendrel itself doesn't directly enforce any quals today, its child
relations may. Therefore, setting rel->tuples equal to rel->rows for
an appendrel isn't always appropriate.
Doing so can lead to issues in cost estimates in some cases. For
instance, when estimating the number of distinct values from an
appendrel, we would not be able to adjust the estimate based on the
restriction selectivity.
This patch addresses this by setting an appendrel's tuples to the
total number of tuples accumulated from each live child, which better
aligns with reality.
This is arguably a bug, but nobody has complained about that until
now, so no back-patch.
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Reviewed-by: Alena Rybakina <a.rybakina@postgrespro.ru>
Discussion: https://postgr.es/m/CAMbWs4_TG_+kVn6fjG-5GYzzukrNK57=g9eUo4gsrUG26OFawg@mail.gmail.com
This commit introduces changes to track unpruned relations explicitly,
making it possible for top-level plan nodes, such as ModifyTable and
LockRows, to avoid processing partitions pruned during initial
pruning. Scan-level nodes, such as Append and MergeAppend, already
avoid the unnecessary processing by accessing partition pruning
results directly via part_prune_index. In contrast, top-level nodes
cannot access pruning results directly and need to determine which
partitions remain unpruned.
To address this, this commit introduces a new bitmapset field,
es_unpruned_relids, which the executor uses to track the set of
unpruned relations. This field is referenced during plan
initialization to skip initializing certain nodes for pruned
partitions. It is initialized with PlannedStmt.unprunableRelids,
a new field that the planner populates with RT indexes of relations
that cannot be pruned during runtime pruning. These include relations
not subject to partition pruning and those required for execution
regardless of pruning.
PlannedStmt.unprunableRelids is computed during set_plan_refs() by
removing the RT indexes of runtime-prunable relations, identified
from PartitionPruneInfos, from the full set of relation RT indexes.
ExecDoInitialPruning() then updates es_unpruned_relids by adding
partitions that survive initial pruning.
To support this, PartitionedRelPruneInfo and PartitionedRelPruningData
now include a leafpart_rti_map[] array that maps partition indexes to
their corresponding RT indexes. The former is used in set_plan_refs()
when constructing unprunableRelids, while the latter is used in
ExecDoInitialPruning() to convert partition indexes returned by
get_matching_partitions() into RT indexes, which are then added to
es_unpruned_relids.
These changes make it possible for ModifyTable and LockRows nodes to
process only relations that remain unpruned after initial pruning.
ExecInitModifyTable() trims lists, such as resultRelations,
withCheckOptionLists, returningLists, and updateColnosLists, to
consider only unpruned partitions. It also creates ResultRelInfo
structs only for these partitions. Similarly, child RowMarks for
pruned relations are skipped.
By avoiding unnecessary initialization of structures for pruned
partitions, these changes improve the performance of updates and
deletes on partitioned tables during initial runtime pruning.
Due to ExecInitModifyTable() changes as described above, EXPLAIN on a
plan for UPDATE and DELETE that uses runtime initial pruning no longer
lists partitions pruned during initial pruning.
Reviewed-by: Robert Haas <robertmhaas@gmail.com> (earlier versions)
Reviewed-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/CA+HiwqFGkMSge6TgC9KQzde0ohpAycLQuV7ooitEEpbKB0O_mg@mail.gmail.com
This commit allows transformation of OR-clauses into SAOP's for index scans
within nested loop joins. That required the following changes.
1. Make match_orclause_to_indexcol() and group_similar_or_args() understand
const-ness in the same way as match_opclause_to_indexcol(). This
generally makes our approach more uniform.
2. Make match_join_clauses_to_index() pass OR-clauses to
match_clause_to_index().
3. Also switch match_join_clauses_to_index() to use list_append_unique_ptr()
for adding clauses to *joinorclauses. That avoids possible duplicates
when processing the same clauses with different indexes. Previously such
duplicates were elimited in match_clause_to_index(), but now
group_similar_or_args() each time generates distinct copies of grouped
OR clauses.
Discussion: https://postgr.es/m/CAPpHfdv%2BjtNwofg-p5z86jLYZUTt6tR17Wy00ta0dL%3DwHQN3ZA%40mail.gmail.com
Reviewed-by: Andrei Lepikhov <lepihov@gmail.com>
Reviewed-by: Alena Rybakina <a.rybakina@postgrespro.ru>
Reviewed-by: Pavel Borisov <pashkin.elfe@gmail.com>
Since d4378c0005, match_clause_to_indexcol() doesn't always return NULL
for an OR clause. This commit reflects that in the function header comment.
Reported-by: Pavel Borisov <pashkin.elfe@gmail.com>
Consistently use "Size" (or size_t, or in some places int64 or double)
as the type for variables holding memory allocation sizes. In most
places variables' data types were fine already, but we had an ancient
habit of computing bytes from kilobytes-units GUCs with code like
"work_mem * 1024L". That risks overflow on Win64 where they did not
make "long" as wide as "size_t". We worked around that by restricting
such GUCs' ranges, so you couldn't set work_mem et al higher than 2GB
on Win64. This patch removes that restriction, after replacing such
calculations with "work_mem * (Size) 1024" or variants of that.
It should be noted that this patch was constructed by searching
outwards from the GUCs that have MAX_KILOBYTES as upper limit.
So I can't positively guarantee there are no other places doing
memory-size arithmetic in int or long variables. I do however feel
pretty confident that increasing MAX_KILOBYTES on Win64 is safe now.
Also, nothing in our code should be dealing in multiple-gigabyte
allocations without authorization from a relevant GUC, so it seems
pretty likely that this search caught everything that could be at
risk of overflow.
Author: Vladlen Popolitov <v.popolitov@postgrespro.ru>
Co-authored-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/1a01f0-66ec2d80-3b-68487680@27595217
This moves PartitionPruneInfo from plan nodes to PlannedStmt,
simplifying traversal by centralizing all PartitionPruneInfo
structures in a single list in it, which holds all instances for the
main query and its subqueries. Instead of plan nodes (Append or
MergeAppend) storing PartitionPruneInfo pointers, they now reference
an index in this list.
A bitmapset field is added to PartitionPruneInfo to store the RT
indexes corresponding to the apprelids field in Append or MergeAppend.
This allows execution pruning logic to verify that it operates on the
correct plan node, mainly to facilitate debugging.
Duplicated code in set_append_references() and
set_mergeappend_references() is refactored into a new function,
register_pruneinfo(). This updates RT indexes by applying rtoffet
and adds PartitionPruneInfo to the global list in PlannerGlobal.
By allowing pruning to be performed without traversing the plan tree,
this change lays the groundwork for runtime initial pruning to occur
independently of plan tree initialization.
Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org> (earlier version)
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/CA+HiwqFGkMSge6TgC9KQzde0ohpAycLQuV7ooitEEpbKB0O_mg@mail.gmail.com
If a column is omitted in an INSERT, and there's no column default,
the code in preptlist.c generates a NULL Const to be inserted.
Furthermore, if the column is of a domain type, we wrap the Const
in CoerceToDomain, so as to throw a run-time error if the domain
has a NOT NULL constraint. That's fine as far as it goes, but
there are two problems:
1. We're being sloppy about the type/typmod that the Const is
labeled with. It really should have the domain's base type/typmod,
since it's the input to CoerceToDomain not the output. This can
result in coerce_to_domain inserting a useless length-coercion
function (useless because it's being applied to a null). The
coercion would typically get const-folded away later, but it'd
be better not to create it in the first place.
2. We're not applying expression preprocessing (specifically,
eval_const_expressions) to the resulting expression tree.
The planner's primary expression-preprocessing pass already happened,
so that means the length coercion step and CoerceToDomain node miss
preprocessing altogether.
This is at the least inefficient, since it means the length coercion
and CoerceToDomain will actually be executed for each inserted row,
though they could be const-folded away in most cases. Worse, it
seems possible that missing preprocessing for the length coercion
could result in an invalid plan (for example, due to failing to
perform default-function-argument insertion). I'm not aware of
any live bug of that sort with core datatypes, and it might be
unreachable for extension types as well because of restrictions of
CREATE CAST, but I'm not entirely convinced that it's unreachable.
Hence, it seems worth back-patching the fix (although I only went
back to v14, as the patch doesn't apply cleanly at all in v13).
There are several places in the rewriter that are building null
domain constants the same way as preptlist.c. While those are
before the planner and hence don't have any reachable bug, they're
still applying a length coercion that will be const-folded away
later, uselessly wasting cycles. Hence, make a utility routine
that all of these places can call to do it right.
Making this code more careful about the typmod assigned to the
generated NULL constant has visible but cosmetic effects on some
of the plans shown in contrib/postgres_fdw's regression tests.
Discussion: https://postgr.es/m/1865579.1738113656@sss.pgh.pa.us
Backpatch-through: 14
We should run the expression subtrees of PartitionedRelPruneInfo
structs through fix_scan_expr. Failure to do so means that
AlternativeSubPlans within those expressions won't be cleaned up
properly, resulting in "unrecognized node type" errors since v14.
It seems fairly likely that at least some of the other steps done
by fix_scan_expr are important here as well, resulting in as-yet-
undetected bugs. Therefore, I've chosen to back-patch this to
all supported branches including v13, even though the known
symptom doesn't manifest in v13.
Per bug #18778 from Alexander Lakhin.
Discussion: https://postgr.es/m/18778-24cd399df6c806af@postgresql.org
This allows the RETURNING list of INSERT/UPDATE/DELETE/MERGE queries
to explicitly return old and new values by using the special aliases
"old" and "new", which are automatically added to the query (if not
already defined) while parsing its RETURNING list, allowing things
like:
RETURNING old.colname, new.colname, ...
RETURNING old.*, new.*
Additionally, a new syntax is supported, allowing the names "old" and
"new" to be changed to user-supplied alias names, e.g.:
RETURNING WITH (OLD AS o, NEW AS n) o.colname, n.colname, ...
This is useful when the names "old" and "new" are already defined,
such as inside trigger functions, allowing backwards compatibility to
be maintained -- the interpretation of any existing queries that
happen to already refer to relations called "old" or "new", or use
those as aliases for other relations, is not changed.
For an INSERT, old values will generally be NULL, and for a DELETE,
new values will generally be NULL, but that may change for an INSERT
with an ON CONFLICT ... DO UPDATE clause, or if a query rewrite rule
changes the command type. Therefore, we put no restrictions on the use
of old and new in any DML queries.
Dean Rasheed, reviewed by Jian He and Jeff Davis.
Discussion: https://postgr.es/m/CAEZATCWx0J0-v=Qjc6gXzR=KtsdvAE7Ow=D=mu50AgOe+pvisQ@mail.gmail.com
RowCompareType served as a way to describe the fundamental meaning of
an operator, notionally independent of an operator class (although so
far this was only really supported for btrees). Its original purpose
was for use inside RowCompareExpr, and it has also found some small
use outside, such as for get_op_btree_interpretation().
We want to expand this now, as a more general way to describe operator
semantics for other index access methods, including gist (to improve
GistTranslateStratnum()) and others not written yet. To avoid future
confusion, we rename the type to CompareType and the symbols from
ROWCOMPARE_XXX to COMPARE_XXX to reflect their more general purpose.
Reviewed-by: Mark Dilger <mark.dilger@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/flat/E72EAA49-354D-4C2E-8EB9-255197F55330@enterprisedb.com
66c0185a3 gave the planner the ability to have union child queries
provide the union planner with pre-sorted input so that UNION queries
could be more efficiently implemented using Merge Append.
That commit overlooked checking that the UNION target list and the union
child target list's types all match. In some corner cases, this could
result in the planner producing sorts using the sort operator of the
top-level UNION's target list type rather than of the union child's
target list's type. The implications of this range from silently
working correctly, despite using the wrong sort operator all the way up
to a segmentation fault.
Here we fix by adjusting the planner so it makes no attempt to have the
subquery produce pre-sorted results when the data type of the UNION
target list and the types from the subquery target list don't match
exactly.
Backpatch to 17, where 66c0185a3 was introduced.
Reported-by: Jason Smith <dqetool@126.com>
Diagnosed-by: Tom Lane <tgl@sss.pgh.pa.us>
Bug: 18764
Discussion: https://postgr.es/m/18764-63ad667ea26e877a%40postgresql.org
Backpatch-through: 17
The new compact_attrs array stores a few select fields from
FormData_pg_attribute in a more compact way, using only 16 bytes per
column instead of the 104 bytes that FormData_pg_attribute uses. Using
CompactAttribute allows performance-critical operations such as tuple
deformation to be performed without looking at the FormData_pg_attribute
element in TupleDesc which means fewer cacheline accesses.
For some workloads, tuple deformation can be the most CPU intensive part
of processing the query. Some testing with 16 columns on a table
where the first column is variable length showed around a 10% increase in
transactions per second for an OLAP type query performing aggregation on
the 16th column. However, in certain cases, the increases were much
higher, up to ~25% on one AMD Zen4 machine.
This also makes pg_attribute.attcacheoff redundant. A follow-on commit
will remove it, thus shrinking the FormData_pg_attribute struct by 4
bytes.
Author: David Rowley
Reviewed-by: Andres Freund, Victor Yegorov
Discussion: https://postgr.es/m/CAApHDvrBztXP3yx=NKNmo3xwFAFhEdyPnvrDg3=M0RhDs+4vYw@mail.gmail.com
Remove the code for inserting flag columns in the inputs of a SetOp.
That was the only reason why there would be resjunk columns in a
set-operations plan tree, so we can get rid of some code that
supported that, too.
Get rid of choose_hashed_setop() in favor of building Paths for
the hashed and sorted alternatives, and letting them fight it out
within add_path().
Remove set_operation_ordered_results_useful(), which was giving wrong
answers due to examining the wrong ancestor node: we need to examine
the immediate SetOperationStmt parent not the topmost node. Instead
make each caller of recurse_set_operations() pass down the relevant
parent node. (This thinko seems to have led only to wasted planning
cycles and possibly-inferior plans, not wrong query answers. Perhaps
we should back-patch it, but I'm not doing so right now.)
Teach generate_nonunion_paths() to consider pre-sorted inputs for
sorted SetOps, rather than always generating a Sort node.
Patch by me; thanks to Richard Guo and David Rowley for review.
Discussion: https://postgr.es/m/1850138.1731549611@sss.pgh.pa.us
The original design for set operations involved appending the two
input relations into one and adding a flag column that allows
distinguishing which side each row came from. Then the SetOp node
pries them apart again based on the flag. This is bizarre. The
only apparent reason to do it is that when sorting, we'd only need
one Sort node not two. But since sorting is at least O(N log N),
sorting all the data is actually worse than sorting each side
separately --- plus, we have no chance of taking advantage of
presorted input. On top of that, adding the flag column frequently
requires an additional projection step that adds cycles, and then
the Append node isn't free either. Let's get rid of all of that
and make the SetOp node have two separate children, using the
existing outerPlan/innerPlan infrastructure.
This initial patch re-implements nodeSetop.c and does a bare minimum
of work on the planner side to generate correctly-shaped plans.
In particular, I've tried not to change the cost estimates here,
so that the visible changes in the regression test results will only
involve removal of useless projection steps and not any changes in
whether to use sorted vs hashed mode.
For SORTED mode, we combine successive identical tuples from each
input into groups, and then merge-join the groups. The tuple
comparisons now use SortSupport instead of simple equality, but
the group-formation part should involve roughly the same number of
tuple comparisons as before. The cross-comparisons between left and
right groups probably add to that, but I'm not sure to quantify how
many more comparisons we might need.
For HASHED mode, nodeSetop's logic is almost the same as before,
just refactored into two separate loops instead of one loop that
has an assumption that it will see all the left-hand inputs first.
In both modes, I added early-exit logic to not bother reading the
right-hand relation if the left-hand input is empty, since neither
INTERSECT nor EXCEPT modes can produce any output if the left input
is empty. This could have been done before in the hashed mode, but
not in sorted mode. Sorted mode can also stop as soon as it exhausts
the left input; any remaining right-hand tuples cannot have matches.
Also, this patch adds some infrastructure for detecting whether
child plan nodes all output the same type of tuple table slot.
If they do, the hash table logic can use slightly more efficient
code based on assuming that that's the input slot type it will see.
We'll make use of that infrastructure in other plan node types later.
Patch by me; thanks to Richard Guo and David Rowley for review.
Discussion: https://postgr.es/m/1850138.1731549611@sss.pgh.pa.us
Commit b437571714 allowed parallel builds for BRIN, but left behind
two comments claiming only btree indexes support parallel builds.
Reported by Egor Rogov, along with similar issues in SGML docs.
Backpatch to 17, where parallel builds for BRIN were introduced.
Reported-by: Egor Rogov
Backpatch-through: 17
Discussion: https://postgr.es/m/114e2d5d-125e-07d8-94aa-5ad175fb7443@postgrespro.ru
When pulling up a subquery that is under an outer join, if the
subquery's target list contains a strict expression that uses a
subquery variable, it's okay to pull up the expression without
wrapping it in a PlaceHolderVar: if the subquery variable is forced to
NULL by the outer join, the expression result will come out as NULL
too.
If the strict expression does not contain any subquery variables, the
current code always wraps it in a PlaceHolderVar. While this is not
incorrect, the analysis could be tighter: if the strict expression
contains any variables of rels that are under the same lowest nulling
outer join as the subquery, we can also avoid wrapping it. This is
safe because if the subquery variable is forced to NULL by the outer
join, the variables of rels that are under the same lowest nulling
outer join will also be forced to NULL, resulting in the expression
evaluating to NULL as well. Therefore, it's not necessary to force
the expression to be evaluated below the outer join. It could be
beneficial to get rid of such PHVs because they could imply lateral
dependencies, which force us to resort to nestloop joins.
This patch checks if the lateral references in the strict expression
contain any variables of rels under the same lowest nulling outer join
as the subquery, and avoids wrapping the expression in that case.
This is fundamentally a generalization of the optimizations for bare
Vars and PHVs introduced in commit f64ec81a8.
No backpatch as this could result in plan changes.
Author: Richard Guo
Discussion: https://postgr.es/m/CAMbWs4_ENtfRdLaM_bXAxiKRYO7DmwDBDG4_2=VTDi0mJP-jAw@mail.gmail.com
d4c3a156c added support that when the GROUP BY contained all of the
columns belonging to a relation's PRIMARY KEY, all other columns
belonging to that relation would be removed from the GROUP BY clause.
That's possible because all other columns are functionally dependent on
the PRIMARY KEY and those columns alone ensure the groups are distinct.
Here we expand on that optimization and allow it to work for any unique
indexes on the table rather than just the PRIMARY KEY index. This
normally requires that all columns in the index are defined with NOT NULL,
however, we can relax that requirement when the index is defined with
NULLS NOT DISTINCT.
When there are multiple suitable indexes to allow columns to be removed,
we prefer the index with the least number of columns as this allows us
to remove the highest number of GROUP BY columns. One day, we may want to
revisit that decision as it may make more sense to use the narrower set of
columns in terms of the width of the data types and stored/queried data.
This also adjusts the code to make use of RelOptInfo.indexlist rather
than looking up the catalog tables.
In passing, add another short-circuit path to allow bailing out earlier
in cases where it's certainly not possible to remove redundant GROUP BY
columns. This early exit is now cheaper to do than when this code was
originally written as 00b41463c made it cheaper to check for empty
Bitmapsets.
Patch originally by Zhang Mingli and later worked on by jian he, but after
I (David) worked on it, there was very little of the original left.
Author: Zhang Mingli, jian he, David Rowley
Reviewed-by: jian he, Andrei Lepikhov
Discussion: https://postgr.es/m/327990c8-b9b2-4b0c-bffb-462249f82de0%40Spark
Traditionally, remove_useless_groupby_columns() was called during
grouping_planner() directly after the call to preprocess_groupclause().
While in many ways, it made sense to populate the field and remove the
functionally dependent columns from processed_groupClause at the same
time, it's just that doing so had the disadvantage that
remove_useless_groupby_columns() was being called before the RelOptInfos
were populated for the relations mentioned in the query. Not having
RelOptInfos available meant we needed to manually query the catalog tables
to get the required details about the primary key constraint for the
table.
Here we move the remove_useless_groupby_columns() call to
query_planner() and put it directly after the RelOptInfos are populated.
This is fine to do as processed_groupClause still isn't final at this
point as it can still be modified inside standard_qp_callback() by
make_pathkeys_for_sortclauses_extended().
This commit is just a refactor and simply moves
remove_useless_groupby_columns() into initsplan.c. A planned follow-up
commit will adjust that function so it uses RelOptInfo instead of doing
catalog lookups and also teach it how to use unique indexes as proofs to
expand the cases where we can remove functionally dependent columns from
the GROUP BY.
Reviewed-by: Andrei Lepikhov, jian he
Discussion: https://postgr.es/m/CAApHDvqLezKwoEBBQd0dp4Y9MDkFBDbny0f3SzEeqOFoU7Z5+A@mail.gmail.com
When pulling up a lateral subquery that is under an outer join, the
current code always wraps a Var or PHV in the subquery's targetlist
into a new PlaceHolderVar if it is a lateral reference to something
outside the subquery. This is necessary when the Var/PHV references
the non-nullable side of the outer join from the nullable side: we
need to ensure that it is evaluated at the right place and hence is
forced to null when the outer join should do so. However, if the
referenced rel is under the same lowest nulling outer join, we can
actually omit the wrapping. That's safe because if the subquery
variable is forced to NULL by the outer join, the lateral reference
variable will come out as NULL too. It could be beneficial to get rid
of such PHVs because they imply lateral dependencies, which force us
to resort to nestloop joins.
This patch leverages the newly introduced nullingrel_info to check if
the nullingrels of the subquery RTE are a subset of those of the
laterally referenced rel, in order to determine if the referenced rel
is under the same lowest nulling outer join.
No backpatch as this could result in plan changes.
Author: Richard Guo
Reviewed-by: James Coleman, Dmitry Dolgov, Andrei Lepikhov
Discussion: https://postgr.es/m/CAMbWs48uk6C7Z9m_FNT8_21CMCk68hrgAsz=z6zpP1PNZMkeoQ@mail.gmail.com
As I'd feared, commit 5c9d8636d was still a few bricks shy of a load.
We can't just leave pulled-up lateral-reference Vars with no new
nullingrels: we have to carefully compute what subset of the
to-be-replaced Var's nullingrels apply to them, else we still get
"wrong varnullingrels" errors. This is a bit tedious, but it looks
like we can use the nullingrel data this patch computes for other
purposes, enabling better optimization. We don't want to inject
unnecessary plan changes into stable branches though, so leave that
idea for a later HEAD-only patch.
Patch by me, but thanks to Richard Guo for devising a test case that
broke 5c9d8636d, and for preliminary investigation about how to fix
it. As before, back-patch to v16.
Discussion: https://postgr.es/m/E1tGn4j-0003zi-MP@gemulon.postgresql.org
There is no point in transforming OR-clauses into SAOP's if the target index
doesn't support SAOP scans anyway. This commit adds corresponding checks
to match_orclause_to_indexcol() and group_similar_or_args(). The first check
fixes the actual bug, while the second just saves some cycles.
Reported-by: Alexander Lakhin
Discussion: https://postgr.es/m/8174de69-9e1a-0827-0e81-ef97f56a5939%40gmail.com
Author: Alena Rybakina
Reviewed-by: Ranier Vilela, Alexander Korotkov, Andrei Lepikhov
If we are pulling up a subquery that's under an outer join, and
the subquery's target list contains a strict expression that uses
both a subquery variable and a lateral-reference variable, it's okay
to pull up the expression without wrapping it in a PlaceHolderVar.
That's safe because if the subquery variable is forced to NULL
by the outer join, the expression result will come out as NULL too,
so we don't have to force that outcome by evaluating the expression
below the outer join. It'd be correct to wrap in a PHV, but that can
lead to very significantly worse plans, since we'd then have to use
a nestloop plan to pass down the lateral reference to where the
expression will be evaluated.
However, when we do that, we should not mark the lateral reference
variable as being nulled by the outer join, because it isn't after
we pull up the expression in this way. So the marking logic added
by cb8e50a4a was incorrect in this detail, leading to "wrong
varnullingrels" errors from the consistency-checking logic in
setrefs.c. It seems to be sufficient to just not mark lateral
references at all in this case. (I have a nagging feeling that more
complexity may be needed in cases where there are several levels of
outer join, but some attempts to break it with that didn't succeed.)
Per report from Bertrand Mamasam. Back-patch to v16, as the previous
patch was.
Discussion: https://postgr.es/m/CACZ67_UA_EVrqiFXJu9XK50baEpH=ofEPJswa2kFxg6xuSw-ww@mail.gmail.com
The function get_param_path_clause_serials() is used to get the set of
pushed-down clauses enforced within a parameterized Path. Since we
don't currently support parameterized MergeAppend paths, and it
doesn't look like that is going to change anytime soon (as explained
in the comments for generate_orderedappend_paths), we don't need to
consider MergeAppendPath in this function.
This change won't make any measurable difference in performance; it's
just for clarity's sake.
Author: Richard Guo
Reviewed-by: Andrei Lepikhov
Discussion: https://postgr.es/m/CAMbWs4_Puie4DQ2ODvjQB_3CxYkUODnrJm8jn_ObMAcrjYNW7Q@mail.gmail.com
The ordering of DISTINCT items is semantically insignificant, so we
can reorder them as needed. In fact, in the parser, we absorb the
sorting semantics of the sortClause as much as possible into the
distinctClause, ensuring that one clause is a prefix of the other.
This can help avoid a possible need to re-sort.
In this commit, we attempt to adjust the DISTINCT keys to match the
input path's pathkeys. This can likewise help avoid re-sorting, or
allow us to use incremental-sort to save efforts.
For DISTINCT ON expressions, the parser already ensures that they
match the initial ORDER BY expressions. When reordering the DISTINCT
keys, we must ensure that the resulting pathkey list matches the
initial distinctClause pathkeys.
This introduces a new GUC, enable_distinct_reordering, which allows
the optimization to be disabled if needed.
Author: Richard Guo
Reviewed-by: Andrei Lepikhov
Discussion: https://postgr.es/m/CAMbWs48dR26cCcX0f=8bja2JKQPcU64136kHk=xekHT9xschiQ@mail.gmail.com
When optimizer generates bitmap paths, it considers breaking OR-clause
arguments one-by-one. But now, a group of similar OR-clauses can be
transformed into SAOP during index matching. So, bitmap paths should
keep up.
This commit teaches bitmap paths generation machinery to group similar
OR-clauses into dedicated RestrictInfos. Those RestrictInfos are considered
both to match index as a whole (as SAOP), or to match as a set of individual
OR-clause argument one-by-one (the old way).
Therefore, bitmap path generation will takes advantage of OR-clauses to SAOP's
transformation. The old way of handling them is also considered. So, there
shouldn't be planning regression.
Discussion: https://postgr.es/m/CAPpHfdu5iQOjF93vGbjidsQkhHvY2NSm29duENYH_cbhC6x%2BMg%40mail.gmail.com
Author: Alexander Korotkov, Andrey Lepikhov
Reviewed-by: Alena Rybakina, Andrei Lepikhov, Jian he, Robert Haas
Reviewed-by: Peter Geoghegan
This commit makes match_clause_to_indexcol() match
"(indexkey op C1) OR (indexkey op C2) ... (indexkey op CN)" expression
to the index while transforming it into "indexkey op ANY(ARRAY[C1, C2, ...])"
(ScalarArrayOpExpr node).
This transformation allows handling long OR-clauses with single IndexScan
avoiding diving them into a slower BitmapOr.
We currently restrict Ci to be either Const or Param to apply this
transformation only when it's clearly beneficial. However, in the future,
we might switch to a liberal understanding of constants, as it is in other
cases.
Discussion: https://postgr.es/m/567ED6CA.2040504%40sigaev.ru
Author: Alena Rybakina, Andrey Lepikhov, Alexander Korotkov
Reviewed-by: Peter Geoghegan, Ranier Vilela, Alexander Korotkov, Robert Haas
Reviewed-by: Jian He, Tom Lane, Nikolay Shaplov
In the dim past we figured it was okay to ignore collations
when combining UNION set-operation nodes into a single N-way
UNION operation. I believe that was fine at the time, but
it stopped being fine when we added nondeterministic collations:
the semantics of distinct-ness are affected by those. v17 made
it even less fine by allowing per-child sorting operations to
be merged via MergeAppend, although I think we accidentally
avoided any live bug from that.
Add a check that collations match before deciding that two
UNION nodes are equivalent. I also failed to resist the
temptation to comment plan_union_children() a little better.
Back-patch to all supported branches (v13 now), since they
all have nondeterministic collations.
Discussion: https://postgr.es/m/3605568.1731970579@sss.pgh.pa.us
