Mar 2021

7 commits
BRIN minmax-multi indexes
26 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1Also in: WIP: BRIN bloom indexes
LLM analysis, 99% surepart of a posted patchsetmoderate changesfrom patchset #1660 of 4 reviewers saw the final version

Commits the core minmax-multi component of #166, excluding its other components; brin_minmax_multi.c matches, but pg_amproc.dat drops cross-type records and corrects timestamp distance-procedure references. This source is later than the sole patchset in thread #37595.

  • Alvaro Herrerasaw an earlier versiontrailerRequested tests, readable summaries and retained in-memory state, and discussed bsearch_arg placement, before subsequent builder and compaction changes.
  • Alexander Korotkovsaw an earlier versiontrailerReviewed distance support and option design and tested the series before recommending dedicated summary types, preceding the hybrid-builder rework.
  • Sokolov Yurano review foundtrailerThe commit credits this reviewer, but no review activity by this person appears in the supplied thread.
  • John Naylorsaw an earlier versiontrailerBenchmarked and profiled construction, reviewed the hybrid builder and caught an assertion error, but compaction logic and distance catalog corrections followed his testing of #160.
BRIN bloom indexes
26 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1Also in: WIP: BRIN bloom indexes
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #1661 of 5 reviewers saw the final version

Commits the core Bloom component of #166, with only catalog-version rebasing and a citation-character correction; its contrib conversion is excluded. The matching submission is newer than the Bloom prototype in thread #37595.

  • Alvaro Herrerasaw an earlier versiontrailerRequested regression coverage and readable summary output before the hashing and summary-representation changes.
  • Alexander Korotkovsaw an earlier versiontrailerReviewed parameter placement and tested the series while proposing dedicated summary types, before the final Bloom algorithm was selected.
  • Sokolov Yurano review foundtrailerHis earlier hashing suggestion is referenced in #98, but no message from him reviewing a patch appears in this thread.
  • Nico Williamsno review foundtrailerThe commit credits this reviewer, but the supplied thread contains no review messages from him.
  • John Naylorsaw the final versiontrailerReviewed hashing, filter sizing and documentation, supplied an experimental alternative, and tested #160, after which the selected Bloom implementation received light cleanup.
Support the old signature of BRIN consistent function
26 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 97% surewritten from the discussion, no matching patch0 of 4 reviewers saw the final version

No indexed patchset contains this post-landing restoration; #171 announces reinstating legacy consistent-function support after reconsidering extension compatibility. The dispatch approach originated in Nikita Glukhov's #40, and the commit credits him alongside Tomas Vondra.

  • Mark Dilgersaw an earlier versiontrailerTested the original scan-key infrastructure and identified its allocation-type issue before the later compatibility implementation and restoration.
  • Alexander Korotkovsaw an earlier versiontrailerReviewed legacy-signature compatibility and explained extension-upgrade requirements, but did not review an attached version of the final restoration.
  • Masahiko Sawadasaw an earlier versiontrailerTested the earlier series containing dual-signature support, before its removal and post-landing restoration.
  • John Naylorsaw an earlier versiontrailerReviewed and tested the interface revisions before Tomas subsequently restored the old signature and reverted existing-opclass conversion.
Remove unnecessary pg_amproc BRIN minmax entries
26 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surewritten from the discussion, no matching patch1 of 1 reviewer saw the final version

#168 proposes and attaches the pg_amproc cleanup using an unindexed .patchx attachment, and #170 confirms its landing. The commit removes redundant cross-type support-procedure records from pg_amproc.dat.

  • Alvaro Herrerasaw the final versiontrailerReviewed the unindexed cleanup in #169, distinguishing required cross-type operators from unnecessary support-procedure records.
Optimize allocations in bringetbitmap
22 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1641 of 5 reviewers saw the final version

The src/backend/access/brin/brin.c diff exactly matches the allocation-optimization component of #164; the other seven components are outside this commit.

  • Alvaro Herrerano review foundtrailerHis discussion reviews other BRIN changes, but does not show review or testing of this allocation optimization.
  • Mark Dilgerno review foundtrailerHis allocation-related review predates this single-chunk optimization and concerns the original scan-key patch.
  • Alexander Korotkovno review foundtrailerHis documented reviews precede introduction of this allocation optimization in #96.
  • Masahiko Sawadano review foundtrailerHis testing of #61 predates introduction of the allocation optimization.
  • John Naylorsaw the final versiontrailerTested the later complete series, including this prerequisite, and reported the first six patches ready in #161.
Move IS [NOT] NULL handling from BRIN support functions
22 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1641 of 5 reviewers saw the final version

Exactly matches the NULL-handling component of #164, centralizing processing in brin.c and adding oi_regular_nulls. Nikita Glukhov supplied the additional insertion/union NULL-processing work in #40 and is explicitly co-credited.

  • Nikita Glukhovsaw an earlier versiontrailerReviewed the initial NULL cleanup and supplied the broader oi_regular_nulls implementation in #40, before later consolidation and fixes.
  • Mark Dilgersaw an earlier versiontrailerTested the early series containing NULL-key extraction before its broader centralization and subsequent fixes.
  • Alexander Korotkovsaw an earlier versiontrailerReviewed NULL-handling compatibility and explicitly raised forwarding of NULL keys for custom opclasses in #57.
  • Masahiko Sawadasaw an earlier versiontrailerTested the series before the NULL patches were combined and early-rejection logic was corrected.
  • John Naylorsaw the final versiontrailerTested the late prerequisite series in #161, with no substantive subsequent change to this component.
Pass all scan keys to BRIN consistent function at once
22 Mar 2021Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1641 of 5 reviewers saw the final version

Matches the all-scan-keys component of #164 across brin.c, existing opclasses, validation, documentation and catalogs, with only the catalog-version date changed.

  • Alvaro Herrerasaw an earlier versiontrailerReviewed series structure and supported processing keys together in #149, before legacy-signature support was removed.
  • Mark Dilgersaw an earlier versiontrailerRan regression tests and identified the ScanKey allocation-type issue in #9, before substantial interface revisions.
  • Alexander Korotkovsaw an earlier versiontrailerReviewed compatibility, catalog signatures and extensibility documentation before the final decision to require the new signature.
  • John Naylorsaw the final versiontrailerReviewed the interface decision and tested #160 after legacy-signature support was removed.
  • Nikita Glukhovsaw an earlier versiontrailerCorrected validation and catalog signatures and introduced compatibility dispatch in #40, which was subsequently removed before this landing.

Nov 2020

1 commit
Properly detoast data in brin_form_tuple
6 Nov 2020Tomas Vondramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1REL_13_STABLE13.1+5 moreAlso in: Fix brin_form_tuple to properly detoast data (part of a posted patchset)

Credits across 8 commits

Author
Tomas Vondra8
Nikita Glukhov2
Reviewer
Alexander Korotkov6
Alvaro Herrera6
John Naylor6
Mark Dilger4
Masahiko Sawada3
Nikita Glukhov2
Yura Sokolov2
Nico Williams1
Committer
Tomas Vondra8