Mar 2025

1 commit
Make group_similar_or_args() reorder clause list as little as possible
28 Mar 2025Alexander Korotkovmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #3233 of 3 reviewers saw the final version

Implements the groupindex and second-sort approach from #323 in indxpath.c, but changes the group-key comparisons from OR to AND and adds regression tests and comment edits.

  • Pavel Borisovsaw the final versiontrailerReviewed both approaches and supplied test and comment feedback on the final submitted implementation in #324.
  • Andrei Lepikhovsaw the final versiontrailerReviewed the replacement for his original fix and accepted the additional sorting cost in #325 and #327.
  • Alena Rybakinasaw the final versiontrailerExamined both approaches and resolved her remaining concern about the final submission in #328 and #329.

Feb 2025

1 commit
Allow usage of match_orclause_to_indexcol() for joins
4 Feb 2025Alexander Korotkovmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #3173 of 3 reviewers saw the final version

Matches the second patch of #317, including relation-relative constant detection and join-clause deduplication; the separate header-comment patch is outside this commit.

  • Andrei Lepikhovsaw the final versiontrailerReported duplicate clauses and a crash, supplied #303, and reviewed the near-final version in #315, after which only tests changed.
  • Alena Rybakinasaw the final versiontrailerReviewed correctness and performance cases and accepted the final submission in #318.
  • Pavel Borisovsaw the final versiontrailerRequested explanation and test improvements in #310 and reviewed the final version in #319.

Nov 2024

4 commits
Skip not SOAP-supported indexes while transforming an OR clause into SAOP
29 Nov 2024Alexander Korotkovmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surelanded as postedfrom patchset #2952 of 3 reviewers saw the final version

Matches #295 modulo whitespace, adding amsearcharray checks to match_orclause_to_indexcol() and group_similar_or_args(); the commit explicitly credits Alena, whose original fix is #291.

  • Ranier Vilelasaw an earlier versiontrailerSuggested moving the guard before the loop in #293, before the additional grouping-function guard was introduced.
  • Alexander Korotkovsaw the final versiontrailerExtended Alena's fix with the grouping-function check in #295 and reviewed the committed version as committer.
  • Andrei Lepikhovsaw the final versiontrailerReviewed the complete revised fix and reported that it looked good in #296.
Remove the wrong assertion from match_orclause_to_indexcol()
25 Nov 2024Alexander Korotkovmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Richard Guo reported that a valid zero constant triggers the assertion in #287; Korotkov removed it and added regression coverage in #288, without an attached patchset.

  • Richard Guosaw the final versionLLM analysis, 100% sureReported the zero-constant assertion failure in #287, directly prompting the assertion removal and regression test.
Teach bitmap path generation about transforming OR-clauses to SAOP's
23 Nov 2024Alexander Korotkovmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #2833 of 5 reviewers saw the final version

Takes the second patch of #283, adding bitmap OR grouping and make_plain_restrictinfo(), with only a comment typo corrected; sender Alexander Korotkov is a credited coauthor.

  • Alena Rybakinasaw the final versiontrailerTested grouping and recheck behavior through #280, after which the accepted bitmap implementation received only light cleanup.
  • Andrei Lepikhovsaw the final versiontrailerIdentified RestrictInfo and bitmap-path issues, supplied corrections, and reviewed the final pair in #284.
  • Jian hesaw the final versiontrailerReviewed grouping, rechecks and index qualification through #279, covering the implementation retained in the final submission.
  • Robert Haassaw an earlier versiontrailerReviewed semantics, comments and planning behavior in #250, before subsequent bitmap fast-path and list-copy changes.
  • Peter Geoghegansaw an earlier versiontrailerExamined type-related performance and planning tradeoffs in #245 and #253, before the later bitmap logic revisions.
Transform OR-clauses to SAOP's during index matching
23 Nov 2024Alexander Korotkovmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #2835 of 7 reviewers saw the final version

Takes the first patch of #283, implementing match_orclause_to_indexcol() and its tests, with a one-word comment edit; sender Alexander Korotkov is a credited coauthor.

  • Peter Geoghegansaw the final versiontrailerReviewed the evolving design and tested mixed-type limitations of the index-matching version in #245, accepting its restricted scope in #263.
  • Ranier Vilelasaw an earlier versiontrailerSupplied parser-stage cleanup and correctness feedback in #22, #31, #41 and #45, before the architectural rewrites.
  • Alexander Korotkovsaw the final versiontrailerReviewed and substantially revised the implementation throughout the thread, including #206, and reviewed the final version as committer.
  • Robert Haassaw the final versiontrailerDrove operator and transformation-stage corrections and reviewed the late index-matching design in #250 and #267, followed by light core changes.
  • Jian Hesaw the final versiontrailerReported volatility and type issues and continued reviewing original-clause rechecks and lossiness through #279.
  • Tom Lanesaw the final versiontrailerReviewed the late transformation's coercion semantics in #264, after which the core retained its restricted type behavior.
  • Nikolay Shaplovsaw an earlier versiontrailerReviewed naming, duplicated branches, tests and index-matching explanations in #196, #198 and #209, before later correctness changes.

Apr 2024

2 commits
Fix the value of or_to_any_transform_limit in postgresql.conf.sample
8 Apr 2024Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

Changes the sample or_to_any_transform_limit value from 0 to 5 in response to Justin Pryzby's report #177; no attached patchset contains this standalone correction.

  • Justin Pryzbysaw the final versionLLM analysis, 100% sureIdentified the incorrect postgresql.conf.sample default in #177 before the correction was committed.
Transform OR clauses to ANY expression
7 Apr 2024Alexander Korotkovmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #1751 of 5 reviewers saw the final version

Commits only the first component of #175, whose header credits Alena and Andrei; #176 confirms omission of the reverse-conversion component. Edits rename the GUC, change its default from 0 to 5 and its threshold comparison, and revise documentation and tests.

  • Peter Geoghegansaw an earlier versiontrailerReviewed SAOP representation, normalization and threshold choices through #104, before conversion moved from parsing to planner preprocessing.
  • Ranier Vilelasaw an earlier versiontrailerSupplied early parser cleanup and correctness feedback through #45, before substantial implementation and placement changes.
  • Alexander Korotkovsaw the final versiontrailerReviewed complexity and planning regressions, proposed the threshold, and explained his final commit edits in #176.
  • Robert Haassaw an earlier versiontrailerChallenged parser-stage optimization, operator lookup and coercions in #103, #106 and #111, before the planner-stage rewrite.
  • Jian Hesaw an earlier versiontrailerReported volatility, type and hash issues and reviewed later cleanup through #161, before the planner relocation and final threshold changes.

Credits across 8 commits

Author
Alexander Korotkov2 LLM4
Alena Rybakina3
Andrei Lepikhov3
Reviewer
Andrei Lepikhov4
Alena Rybakina3
Alexander Korotkov3
jian he3
Peter Geoghegan3
Ranier Vilela3
Robert Haas3
Pavel Borisov2
Nikolay Shaplov1
Tom Lane1
Reported by
Alexander Lakhin1
Andrei Lepikhov1
Justin Pryzby1
Richard Guo1
Committer
Alexander Korotkov8

N LLM means N of the credits come from LLM analysis, not from commit trailers.