| From: | He Kuang <hekuang(dot)git(at)gmail(dot)com> |
|---|---|
| To: | pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Cc: | Richard Guo <guofenglinux(at)gmail(dot)com>, he kuang <hekuang(dot)git(at)gmail(dot)com> |
| Subject: | [PATCH] Fix hash join costing when the inner side is provably unique |
| Date: | 2026-11-13 12:51:54 |
| Message-ID: | CAC5HyryDtq_CaL5v4k+t3vGMBYeJVUVoB5sC9vp4oceK0xfWgQ@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi hackers,
Attached is a standalone fix for hash join costing when the inner side
is provably unique. compute_semi_anti_join_factors() asks
clauselist_selectivity() with the join's own SpecialJoinInfo, so
eqjoinsel() returns ordinary equality selectivity instead of the
semi-join selectivity; separately, an inner-unique plain inner join
should charge hashjointuples on the joinrel row estimate, not
outer_matched_rows. It follows db2d9932 (Richard Guo) but has no
dependency on any other patch, applies cleanly to master 13eb5640
(2026-11-13), and passes make check 243/243.
Comments welcome.
| Attachment | Content-Type | Size |
|---|---|---|
| fix-hash-join-costing-inner-unique.patch | text/plain | 9.6 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Rafia Sabih | 2026-11-13 13:03:48 | Semi-join related bug in postgres_fdw |
| Previous Message | João Rito | 2026-11-13 12:17:22 | Re: [PATCH] Add recovery boundary WAL record for database and tablespace commands |