Aug 2021

1 commit
Fix regexp misbehavior with capturing parens inside "{0}".
24 Aug 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1REL_14_STABLE14rc1REL_13_STABLE13.5+4 more
LLM analysis, 100% suremodified before commitlight changesfrom patchset #271 of 1 reviewer saw the final version

The src/backend/regex/regcomp.c changes exactly match #27. The commit expands and reorders test_regex cases and adds equivalent coverage under src/test/regress, without changing the proposed implementation.

  • Mark Dilgersaw the final versionLLM analysis, 100% sureAfter testing the initial fix and questioning its new errors in #26, he tested the replacement with random patterns and confirmed its changed results were correct in #30; only test additions followed.

Aug 2021

2 commits
Let regexp_replace() make use of REG_NOSUB when feasible.
10 Aug 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #141 of 1 reviewer saw the final version

All implementation changes in src/backend/utils/adt/regexp.c, src/backend/utils/adt/varlena.c, and src/include/utils/varlena.h match #14. The commit additionally tests replacement-string escape handling in src/test/regress/sql/strings.sql and its expected output.

  • Mark Dilgersaw the final versionLLM analysis, 100% sureHe reported four hours of reviewing and testing #14 without finding new breakage in #19, before the commit added only regression coverage.
Avoid determining regexp subexpression matches, when possible.
9 Aug 2021Tom Lanemaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #100 of 2 reviewers saw the final version

Nine of ten changed files match #10; src/backend/regex/regexec.c additionally caps nmatch at the available subexpression count and moves the v->nmatch assignment. This is a local execution-path optimization beyond the reviewed attachment.

  • Andrew Dunstansaw an earlier versionLLM analysis, 95% sureHis concern about backreference numbering in #2 prompted Tom Lane to identify missing regression coverage in #3, before subsequent result-vector logic fixes.
  • Mark Dilgersaw an earlier versionLLM analysis, 98% sureHe reported the assertion and split-result regressions in #7 and #9, then passed nearly 1.5 million regex tests in #11, but the final nmatch optimization was added afterward.

Credits across 3 commits

Author
Tom LaneLLM3
Committer
Tom Lane3

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