Jun 2025

1 commit
Rename gist stratnum support function
2 Jun 2025Peter Eisentrautmaster18beta2REL_19_STABLE18beta2REL_18_STABLE18beta2
LLM analysis, 100% surelanded as postedfrom patchset #2271 of 1 reviewer saw the final version

The rename matches #227, with only the routine catalog-version bump added; Peter explicitly confirms committing it unchanged in #229.

  • Paul A Jungwirthsaw the final versiontrailerIn #228, Paul approves the rename and questions the operator-family wording, which Peter subsequently retains.

Mar 2025

1 commit
Remove support for temporal RESTRICT foreign keys
10 Mar 2025Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #2181 of 1 reviewer saw the final version

Takes the temporal RESTRICT-removal component of #218, adding an explanatory paragraph in doc/src/sgml/ref/create_table.sgml while leaving the five feature patches uncommitted.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Feb 2025

1 commit
Cache NO ACTION foreign keys separately from RESTRICT foreign keys
9 Feb 2025Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #2101 of 2 reviewers saw the final version

Exactly takes the ri_triggers.c cache-key correction from #210, omitting its separate deliberate-failure reproducer.

  • Tom Lanesaw an earlier versionLLM analysis, 94% sureTom reports intermittent RESTRICT failures in #201 and establishes their connection to the preceding NO ACTION change in #207, prompting the cache correction.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Jan 2025

2 commits
Fix NO ACTION temporal foreign keys when the referenced endpoints change
21 Jan 2025Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1991 of 2 reviewers saw the final version

Takes the NO ACTION component of #199, preserving its coverage-query logic in ri_triggers.c with comment and test-annotation edits; whitespace changes were committed separately.

  • Sam Gabrielssonsaw an earlier versionLLM analysis, 98% sureSam supplies a reproducer and coverage-query analysis in #178 showing that the landed temporal foreign keys wrongly reject valid endpoint changes.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Improve whitespace in without_overlaps test
21 Jan 2025Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 98% surepart of a posted patchsetlight changesfrom patchset #1991 of 1 reviewer saw the final version

Extracts indentation changes in src/test/regress/sql/without_overlaps.sql and its expected output from the NO ACTION patch in #199.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Nov - Dec 2024

10 commits
Tests for logical replication with temporal keys
10 Dec 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1952 of 2 reviewers saw the final version

Takes the logical-replication test component of #195, with only Perl formatting changes in src/test/subscription/t/034_temporal.pl.

  • vignesh Csaw the final versiontrailerVignesh reviews test ordering, cleanup, copyright, and helper factoring in #194, all followed only by light revisions.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Support for GiST in get_equal_strategy_number()
10 Dec 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1952 of 2 reviewers saw the final version

The GiST equality-strategy component of #195 matches modulo whitespace; its patch header identifies Peter as author, and #192 contains his original submission.

  • Paul Jungwirthsaw the final versiontrailerPaul explicitly approves the five-patch series and verifies GiST index selection under REPLICA IDENTITY FULL in #195.
  • vignesh Csaw the final versiontrailerVignesh reviews the replication series and its coverage in #194, with only light changes before commitment.
Make the conditions in IsIndexUsableForReplicaIdentityFull() more explicit
10 Dec 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1952 of 2 reviewers saw the final version

Exactly takes the explicit replica-identity capability-check component of #195 modulo whitespace; Paul forwards Peter's patch originally posted in #192.

  • Paul Jungwirthsaw the final versiontrailerPaul approves the implementation and verifies both index-selection call paths in #195.
  • vignesh Csaw the final versiontrailerVignesh reviews IsIndexUsableForReplicaIdentityFull formatting and its commit description in #194.
Replace get_equal_strategy_number_for_am() by get_equal_strategy_number()
10 Dec 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1952 of 2 reviewers saw the final version

Matches the equality-strategy API refactoring component of #195 modulo whitespace; the attachment is Peter's work from #192, forwarded by Paul.

  • Paul Jungwirthsaw the final versiontrailerPaul reviews and tests the unified equality lookup as part of his explicit approval in #195.
  • vignesh Csaw the final versiontrailerVignesh reviews the new per-opclass equality lookup and surrounding replication checks in #194.
Improve internal logical replication error for missing equality strategy
10 Dec 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1951 of 1 reviewer saw the final version

Exactly extracts the two-line missing-equality-strategy diagnostic from #195 in execReplication.c; the commit credits Paul, who sent that source submission.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Simplify IsIndexUsableForReplicaIdentityFull()
4 Dec 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #191

Exactly takes the first component of #191, changing IsIndexUsableForReplicaIdentityFull to accept Relation instead of constructing IndexInfo.

Fix ALTER TABLE / REPLICA IDENTITY for temporal tables
21 Nov 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1872 of 2 reviewers saw the final version

Extracts publisher-side replica-identity validation and regression tests from #187, leaving subscriber lookup and TAP tests for later; the expected index description has a small formatting adjustment.

  • Matthias van de Meentsaw the final versionLLM analysis, 81% sureMatthias explains the amcanunique versus indisunique inconsistency behind the proposed GiST replica-identity fix in #185.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
doc: Update pg_constraint.conexclop docs for WITHOUT OVERLAPS
13 Nov 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1772 of 2 reviewers saw the final version

Exactly takes the conexclop documentation component of #177 in doc/src/sgml/catalogs.sgml.

  • jian hesaw the final versionLLM analysis, 99% sureJian explicitly reports that conexclop documentation omits temporal primary-key and unique constraints in #169.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
doc: Add PERIOD to ALTER TABLE reference docs
13 Nov 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1771 of 1 reviewer saw the final version

Extracts the PERIOD syntax addition from the combined ALTER TABLE documentation patch in #177, correcting the referenced-column placeholder to refcolumn.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch
doc: Add WITHOUT OVERLAPS to ALTER TABLE reference docs
13 Nov 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1772 of 2 reviewers saw the final version

Exactly extracts the WITHOUT OVERLAPS syntax additions from the combined ALTER TABLE documentation patch in #177.

  • jian hesaw the final versionLLM analysis, 99% sureJian explicitly identifies the missing WITHOUT OVERLAPS syntax in alter_table.sgml in #169.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Sep 2024

3 commits
Add temporal FOREIGN KEY contraints
17 Sep 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1682 of 2 reviewers saw the final version

Extracts the temporal foreign-key component of #168, with matching changed lines across its 16 files, including tablecmds.c, ri_triggers.c, and regression tests; the remaining temporal features stay separate.

  • Peter Eisentrautsaw the final versiontrailerPeter explicitly reviews and approves the first four components in #170 and confirms their commitment in #173.
  • jian hesaw the final versiontrailerJian reviews the earlier foreign-key implementation, approves multirange support in #117, and tests collation behavior in #119, followed only by light changes to this component before its reintroduction.
Add temporal PRIMARY KEY and UNIQUE constraints
17 Sep 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1682 of 2 reviewers saw the final version

Combines the temporal primary-key/unique and empty-value rejection components of #168. Applying both to the commit's parent reproduces the committed files apart from comment wrapping, declaration formatting, and the catalog-version bump.

  • Peter Eisentrautsaw the final versiontrailerPeter explicitly approves the primary-key and empty-check components in #170 and reviews their combined final form as committer.
  • jian hesaw the final versiontrailerJian reviews empty-value validation and index handling in #166 and #167, then checks the revised implementation and supplies test cleanup in #169.
Add stratnum GiST support function
17 Sep 2024Peter Eisentrautmaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1682 of 2 reviewers saw the final version

Extracts the GiST strategy-number support component of #168; its changed lines match across all 16 submitted files, with only src/include/catalog/catversion.h added.

  • Peter Eisentrautsaw the final versiontrailerPeter reviews support-function placement, naming, build integration, and validation in #86, then explicitly approves the reintroduced component in #170 before committing it.
  • jian hesaw the final versiontrailerJian reviews the strategy translation implementation and identifies its inaccurate equals-only comment in #87, with only light changes to this component before reintroduction.

May 2024

3 commits
Revert temporal primary keys and foreign keys
16 May 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1reverts 9 commits here
LLM analysis, 100% surelanded as postedfrom patchset #146

Applying all ten rollback attachments from #146 to the commit's parent reproduces the committed files exactly except src/include/catalog/catversion.h; the series was squashed as proposed.

Fix ON CONFLICT DO NOTHING/UPDATE for temporal indexes
10 May 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1322 of 2 reviewers saw the final version

Takes the ON CONFLICT component of #132, removing the unused ii_HasWithoutOverlaps field and assignment while retaining its checks and tests.

  • jian hesaw the final versionLLM analysis, 99% sureJian explains in #124 that existing exclusion metadata makes ii_HasWithoutOverlaps unnecessary, and the committed patch removes it.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Add test for REPLICA IDENTITY with a temporal key
8 May 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1281 of 1 reviewer saw the final version

Exactly takes the replica-identity regression-test component of #128; #131 confirms its separate commitment.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Mar 2024

4 commits
Add temporal FOREIGN KEY contraints
24 Mar 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1202 of 2 reviewers saw the final version

Squashes the range and multirange foreign-key components of #120, leaving FOR PORTION OF, cascading actions, and PERIOD declarations out; #121 confirms the two-component landing.

  • Peter Eisentrautsaw the final versiontrailerPeter drives the shared-RI and range_agg simplification in #101, checks comparison and collation behavior, and reviews the final version as committer.
  • jian hesaw the final versiontrailerJian supplies deferred-constraint tests, approves the multirange component in #117, and investigates collations in #119, followed only by light changes.
Use half-open interval notation in without_overlaps tests
19 Mar 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1081 of 1 reviewer saw the final version

Extracts existing-test literal changes from the foreign-key component of #108, making without_overlaps inputs use the same half-open notation as expected output.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Use daterange and YMD in without_overlaps tests instead of tsrange.
19 Mar 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1082 of 2 reviewers saw the final version

Exactly takes the first component of #108, replacing tsrange and verbose dates with daterange and YMD literals in without_overlaps tests.

  • jian hesaw the final versionLLM analysis, 99% sureJian proposes using daterange to shorten and clarify the temporal tests in #103, which the subsequent test-only patch implements.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Rename pg_constraint.conwithoutoverlaps to conperiod
5 Mar 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1001 of 1 reviewer saw the final version

Takes the conperiod rename component of #100, additionally updating the catalog documentation, overlooked comments, and catalog version.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Jan 2024

3 commits
Fix comment on gist_stratnum_btree
25 Jan 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #912 of 2 reviewers saw the final version

Exactly takes the first component of #91, correcting the gist_stratnum_btree comment to describe all supported comparison strategies.

  • jian hesaw the final versionLLM analysis, 98% sureJian points out the inaccurate equals-only description in #87, and this commit supplies the focused comment correction.
  • Peter Eisentrautsaw the final versioncommittercommitted the patch
Add temporal PRIMARY KEY and UNIQUE constraints
24 Jan 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #882 of 2 reviewers saw the final version

Takes the primary-key/unique component of #88, polishing documentation and GetOperatorFromWellKnownStrategy error messages for translation as explained in #90.

  • Peter Eisentrautsaw the final versiontrailerPeter supplies parser and naming cleanup in #62, reviews the generalized implementation in #86, and performs final edits as committer in #90.
  • jian hesaw the final versiontrailerJian repeatedly tests constraint behavior and reviews the near-final primary-key component in #87, after which only light changes affect this component.
Add stratnum GiST support function
19 Jan 2024Peter Eisentrautmaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #882 of 2 reviewers saw the final version

Takes the GiST stratnum component of #88, with documentation/comment polishing and a catalog-version bump; the temporal feature components remain separate.

  • Peter Eisentrautsaw the final versiontrailerPeter reviews support-function placement, naming, meson integration, and strategy validation in #86 and reviews the final version as committer.
  • jian hesaw the final versiontrailerJian reviews the stratnum implementation and identifies its inaccurate equals-only comment in #87, with only light changes thereafter.

Jan 2022

1 commit
Rename functions to avoid future conflicts
10 Jan 2022Peter Eisentrautmaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #141 of 1 reviewer saw the final version

Extracts the eleven unchanged brin_minmax_multi.c hunks from #14, renaming the private serialization functions to avoid conflicts with rangetypes.h; Peter requests the split in #15.

  • Peter Eisentrautsaw the final versioncommittercommitted the patch

Credits across 29 commits

Author
Paul A Jungwirth1 LLM23
Peter EisentrautLLM6
Reviewer
Peter Eisentrautcommitter23
jian he6
Paul A Jungwirth4
vignesh C4
Reported by
Sam Gabrielsson1
Committer
Peter Eisentraut29

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