Mar 2026

2 commits
Short-circuit row estimation in NOT IN containing NULL consts
19 Mar 2026David Rowleymaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #182 of 3 reviewers saw the final version

The commit preserves both scalararraysel() checks from #18, with comments revised to reference var_eq_const(). As explained in #19, it removes the redundant NULL-only test, disables ANALYZE in the remaining tests, and documents their expected estimates.

  • David Geiersaw an earlier versiontrailerHe questioned memchr overhead in #2, approved the revised approach in #6, and endorsed the bitmap correction in #10, but his reviewed attachments precede the local correctness change to array_contains_nulls().
  • Zsolt Parragisaw the final versiontrailerHe demonstrated the stale-NULL-bitmap bug in #8 and approved the corrected patch with a newline nit in #12, after which only comments and tests changed.
  • David Rowleysaw the final versiontrailerHe reviewed test placement and strictness assumptions in #14, rebased the remaining patch in #17, and reviewed and committed the final version with the edits described in #19.
Move planner row-estimation tests to new planner_est.sql
18 Mar 2026David Rowleymaster19beta1REL_19_STABLE19beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #161 of 1 reviewer saw the final version

Commits only the test-relocation patch from the two-patch series in #16, leaving the optimization for later. The relocation differs only by renaming selectivity_est to planner_est and adding introductory comments, consistent with #17.

  • David Rowleysaw the final versiontrailerHe proposed the separate test relocation in #14 and confirmed reviewing, renaming, and committing that portion of #16 in #17.

Credits across 2 commits

Author
Ilia Evdokimov2
Reviewer
David Rowley2
David Geier1
Zsolt Parragi1
Committer
David Rowley2