Jul 2026

6 commits
Remove btree_gist's useless logic for encoding-aware truncation.
3 Jul 2026Tom Lanemasterunreleased
LLM analysis, 100% suremodified before commitlight changesfrom patchset #81 of 1 reviewer saw the final version

Matches the encoding-aware truncation cleanup in #8, with one additional removal of the unused mb/pg_wchar.h include from contrib/btree_gist/btree_utils_var.c, suggested in #9.

  • Ayush Tiwarisaw the final versiontrailerIn #9, he identified the now-unused include, and the committed version incorporates that light cleanup.
Tighten up btree_gist's handling of truncated bounds.
3 Jul 2026Tom Lanemasterunreleased
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #80 of 1 reviewer saw the final version

Extends the lower-bound prefix-check removal in #8 by dropping unnecessary prefix checks in less-than strategies and refactoring both gbt_var_consistent() and gbt_num_consistent() with bound-test macros and reversed operands. The extension was discussed in #10 and tested in #11, but the final broader refactoring was not attached.

  • Ayush Tiwarisaw an earlier versiontrailerHe supported flipping comparisons in #9 and reported passing regression tests for removing the less-than prefix checks in #11, before the final broader refactoring of both consistent functions.
Sync signatures of gbt_var_consistent() and gbt_num_consistent().
3 Jul 2026Tom Lanemasterunreleased
LLM analysis, 99% surewritten from the discussion, no matching patch

No attachment changes gbt_num_consistent() to accept StrategyNumber by value or makes gbt_var_consistent()'s key argument const. This is Tom Lane's additional API cleanup accompanying the cosmetic work reported as pushed in #13, with the commit's Discussion trailer linking it to this thread.

Fix btree_gist's NotEqual strategy on internal index pages.
3 Jul 2026Tom LanemasterunreleasedREL_19_STABLE19beta2REL_18_STABLE18.6REL_17_STABLE17.11REL_16_STABLE16.15REL_15_STABLE15.19REL_14_STABLE14.24
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #81 of 1 reviewer saw the final version

Derives from the NotEqual patch in #8, whose Author trailer credits Ayush Tiwari despite Tom Lane sending the series. The commit adds pruning when untruncated bounds both equal the query, proposed inline in #8, and aligns the fixed-size NotEqual branch in contrib/btree_gist/btree_utils_num.c.

  • Tom Lanesaw the final versiontrailerHe challenged the initial analysis in #4, accepted the callback explanation in #6, revised the patch and proposed pruning in #8, and reviewed the final version as committer.
Reverse-engineer some documentation for btree_gist's varlena modules.
3 Jul 2026Tom Lanemasterunreleased
LLM analysis, 98% suremodified before commitlight changesfrom patchset #81 of 1 reviewer saw the final version

Matches the six-file documentation patch in #8 except for omitting the temporary comment questioning byteacmp in gbt_bit_ssup_cmp(). That question was already resolved by 558c4ea9a4, which landed first.

  • Ayush Tiwarisaw the final versiontrailerHe explicitly said the documentation patch looked good in #9, and only an obsolete comment was omitted afterward.
Use the proper comparator in gbt_bit_ssup_cmp.
3 Jul 2026Tom LanemasterunreleasedREL_19_STABLE19beta2REL_18_STABLE18.6
LLM analysis, 100% surelanded as postedfrom patchset #81 of 1 reviewer saw the final version

The comparator patch in #8 makes the same byteacmp-to-bitcmp replacement in contrib/btree_gist/btree_bit.c. Landing it before the documentation patch only removes the need to delete that patch's temporary XXX comment.

  • Ayush Tiwarisaw the final versiontrailerIn #9, he confirmed that leaf entries are genuine bit/varbit values and therefore require bitcmp.

Credits across 6 commits

Author
Tom Lane5
Ayush Tiwari1
Reviewer
Ayush Tiwari4
Tom Lane1
Reported by
王跃林1
Committer
Tom Lane6