Mar 2021

5 commits
Doc: Update description for parallel insert reloption.
18 Mar 2021Amit Kapilamaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #3542 of 2 reviewers saw the final version

Uses Justin's create_table.sgml correction from #354, with line wrapping and an additional singular-to-plural correction in alter_table.sgml proposed in #355.

  • Hou, Zhijiesaw the final versiontrailerReviewed Justin's correction and Amit's additional documentation fix in #356.
  • Amit Kapilasaw the final versiontrailerIdentified the additional alter_table.sgml typo in #355 and reviewed the final version as committer.
Add a new GUC and a reloption to enable inserts in parallel-mode.
18 Mar 2021Amit Kapilamaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1Also in: Determine parallel-safety of partition relations for Inserts (not this thread)
LLM analysis, 99% suremodified before commitlight changesfrom patchset #3493 of 4 reviewers saw the final version

Matches the configuration implementation forwarded by Amit in #349, with regression tests changed from SET to RESET. Hou's latest implementation here postdates all supplied attachments in thread #43541, establishing this thread as the landing source.

  • Greg Nancarrowsaw the final versiontrailerReported checking and testing the revised patches in #321 and endorsed the subsequent default-on behavior in #335.
  • Amit Langotesaw the final versiontrailerProposed the PartitionedTableRdOptions name in #341, which was adopted before cosmetic finalization.
  • Takayuki Tsunakawasaw an earlier versiontrailerDiscussed the option naming and default rationale in #333, before the default-on implementation in #336.
  • Amit Kapilasaw the final versiontrailerReviewed Hou's configuration patch, posted cosmetic revisions in #349, and committed the final version.
Improve FK trigger parallel-safety check added by 05c8482f7f.
13 Mar 2021Amit Kapilamaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 90% surepart of a posted patchsetlight changesfrom patchset #3421 of 1 reviewer saw the final version

Implements the second fix attached in #342, replacing clauses.c FK special casing with parallel-restricted pg_proc.dat entries and a catalog-version bump. The attachment extraction failed, preventing exact comparison; light finalization is inferred from the discussion and narrowly matching commit.

  • Amit Kapilasaw the final versiontrailerHandled Greg's second attached fix from #342 and reviewed the final version as committer.
Fix use of relcache TriggerDesc field introduced by commit 05c8482f7f.
12 Mar 2021Amit Kapilamaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 92% surepart of a posted patchsetlight changesfrom patchset #3422 of 2 reviewers saw the final version

Implements the first fix attached in #342, passing Relation into trigger hazard checking and avoiding stale TriggerDesc pointers across cache accesses. Attachment extraction failed, so exact edits cannot be measured; the matching description and subsequent tests support light finalization.

  • Hou Zhijiesaw the final versiontrailerTested the attached fix with CLOBBER_CACHE_ALWAYS and reported the result in #343.
  • Amit Kapilasaw the final versiontrailerReported equivalent cache-clobber testing in #345 and reviewed the committed version.
Enable parallel SELECT for "INSERT INTO ... SELECT ...".
10 Mar 2021Amit Kapilamaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1Also in: Bug in query rewriter - hasModifyingCTE not getting set (not this thread)
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #3263 of 9 reviewers saw the final version

Commits the SELECT implementation and basic tests from the first component of #326, changing only FK-comment wrapping while leaving configuration and advanced tests pending. The attachments in thread #43663 concern the separately discovered rewriter bug, consistent with that thread's not_this_thread verdict.

  • Amit Langotesaw the final versiontrailerContributed hazard-walker and partition-locking fixes, then explicitly reviewed the final submission in #328.
  • Hou Zhijiesaw the final versiontrailerReviewed safety checks, memory handling, tests, and index predicates, and confirmed examination of the final SELECT patch in #329.
  • Takayuki Tsunakawasaw an earlier versiontrailerReviewed locking, planning overhead, and command-ID behavior in #96, #164, #259, and #290, before the final partition-tracking and lock-mode changes.
  • Antonin Houskasaw an earlier versiontrailerIdentified the nested-parallel-execution assertion problem in #109 and verified its correction in #111, before later hazard and partition-cache restructuring.
  • Bharath Rupireddysaw an earlier versiontrailerReviewed unsafe targets, triggers, and partition cases in #11, #35, and #37, before the safety framework was substantially revised.
  • Dilip Kumarsaw an earlier versiontrailerExamined SELECT-only ON CONFLICT and transaction-ID restrictions in #90 and #99, then reviewed safety-check logic in #297 before further restructuring.
  • Vignesh Csaw an earlier versiontrailerReviewed temporary-table exclusions, reusable hazard checking, error handling, and tests in #21, #82, #88, #97, and #113.
  • Zhihong Yusaw an earlier versiontrailerReviewed safety-check cleanup and partition-OID serialization in #144, #241, and #315, before the final lock-mode corrections.
  • Amit Kapilasaw the final versiontrailerDirected the safety and locking reviews, supplied the CTE fix in #277, prepared the polished series in #314, and committed the final SELECT patch.

Oct 2020

1 commit
Fix estimates for ModifyTable paths without RETURNING.
12 Oct 2020Thomas Munromaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #641 of 1 reviewer saw the final version

Uses the returningLists condition from #64 in src/backend/optimizer/util/pathnode.c, but retains and revises the comment about inaccurate RETURNING widths rather than deleting it, as explained in #70.

  • Greg Nancarrowsaw the final versiontrailerExplicitly checked and tested Thomas's attached costing correction in #66.

Credits across 6 commits

Author
Greg Nancarrow3
Amit Kapila1
Amit Langote1
Hou, Zhijie1
Justin Pryzby1
Thomas MunroLLM1
Reviewer
Amit Kapila5
Hou, Zhijie3
Amit Langote2
Greg Nancarrow2
Tsunakawa, Takayuki2
Antonin Houska1
Bharath Rupireddy1
Dilip Kumar1
vignesh C1
Zhihong Yu1
Tested by
Haiying1
Tang1
Reported by
Justin Pryzby1
Committer
Amit Kapila5
Thomas Munro1

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