Aug 2021

2 commits
Keep stats up to date for partitioned tables
28 Aug 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14rc1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #1050 of 3 reviewers saw the final version

Applying #105 to the commit parent confirms the same two-file change, with edits that pass zero tuple counts from analyze.c and correctly enable partitioned-table entries in pgstat_initstats(). The commit credits Hosoya, Justin Pryzby, and Álvaro Herrera, describing the code as mostly traceable to Hosoya.

  • Amit Langotesaw an earlier versionLLM analysis, 88% sureIn #14, he tested the earlier reporting code and requested the restriction on inherited-stat reporting that survives in the restored functionality.
  • Masahiko Sawadasaw an earlier versionLLM analysis, 94% sureHis tests in #18 demonstrated misleading parent live/dead counts and prompted the zero-count behavior retained by this restoration.
  • Kyotaro Horiguchisaw an earlier versionLLM analysis, 90% sureIn #46, he reviewed the reused partitioned-table reporting logic, requesting consolidated zero-count handling and an explicit explanation of inherited-stat behavior.
Revert analyze support for partitioned tables
16 Aug 2021Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14rc1reverts 2 commits here
LLM analysis, 100% suremodified before commitlight changesfrom patchset #972 of 2 reviewers saw the final version

The reversal matches #97, with the development-branch adaptation described in #99: src/include/pgstat.h removes the obsolete counter and advances PGSTAT_FILE_FORMAT_ID instead of retaining an ABI placeholder. The remaining substantive removals match the posted reversal.

  • Andres Freundsaw the final versionLLM analysis, 98% sureAfter demonstrating scalability and counter-ordering problems, he explicitly proposed the adopted revert in #94, before the reversal patch was posted.
  • Tom Lanesaw the final versionLLM analysis, 100% sureHe reviewed the reversal in #98, questioning the claimed catalog-version requirement and confirming the PostgreSQL 14 ABI-preservation requirement.

Apr 2021

2 commits
Document ANALYZE storage parameters for partitioned tables
9 Apr 2021Alvaro Herreramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #792 of 2 reviewers saw the final version

The four create_table.sgml hunks come from #79, with revised introductory wording proposed in #80 and #81, plus clearer wording about disabling ANALYZE.

  • Justin Pryzbysaw the final versionLLM analysis, 100% sureHe identified the missing documentation in #78 and proposed the adopted introductory wording in #80.
  • Tom Lanesaw the final versionLLM analysis, 100% sureHis wording in #81 preserved the clarification about individual leaf partitions and was incorporated with light editing.
autovacuum: handle analyze for partitioned tables
8 Apr 2021Alvaro Herreramaster14beta1REL_19_STABLE14beta1REL_18_STABLE14beta1REL_17_STABLE14beta1REL_16_STABLE14beta1REL_15_STABLE14beta1REL_14_STABLE14beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #713 of 3 reviewers saw the final version

Applying #71 to the commit parent reproduces the implementation except for a va_cols guard, removal of the unused t_changed_tuples_reported field and its updates, and formatting changes. Although Álvaro posted this collector-side rewrite, the commit explicitly credits Hosoya as author.

  • Kyotaro Horiguchisaw the final versiontrailerHe reviewed earlier designs in #43, #46, and #50, then reviewed and tested the landed manual-ANALYZE path in #86, so final coverage includes post-commit review.
  • Tomas Vondrasaw the final versiontrailerHe drove the delta-counter redesign in #57 and #65, then examined the final design's counter and refresh behavior after landing in #74 and #76.
  • Álvaro Herrerasaw the final versiontrailerHe reviewed successive designs, rewrote and tested collector-side propagation in #71, and committed the final version as announced in #72.

Credits across 4 commits

Author
Alvaro HerreraLLM2
yuzuko1
Reviewer
Alvaro Herrera1
Kyotaro Horiguchi1
Tomas Vondra1
Reported by
Justin Pryzby1
Co author
Alvaro Herrera1
Justin Pryzby1
yuzuko1
Committer
Alvaro Herrera4

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