diff options
| author | John Naylor <john.naylor@postgresql.org> | 2025-02-18 11:04:55 +0700 |
|---|---|---|
| committer | John Naylor <john.naylor@postgresql.org> | 2025-02-18 11:04:55 +0700 |
| commit | 53d3daa491be458e543dd5bf24d40595e588e4e7 (patch) | |
| tree | 2dd02a2a3cefa844c90dd047bb95dae80b9f9ede /src/backend/executor/nodeAgg.c | |
| parent | 164bac92f08ccddd6701d44a5338d72c22f7b5c2 (diff) | |
| download | postgresql-53d3daa491be458e543dd5bf24d40595e588e4e7.tar.gz postgresql-53d3daa491be458e543dd5bf24d40595e588e4e7.zip | |
Specialize intarray sorting
There is at least one report in the field of storing millions of
integers in arrays, so it seems like a good time to specialize
intarray's qsort function. In doing so, streamline the comparators:
Previously there were three, two for each direction for sorting
and one passed to qunique_arg. To preserve the early exit in the
case of descending input, pass the direction as an argument to
the comparator. This requires giving up duplicate detection, which
previously allowed skipping the qunique_arg() call. Testing showed
no regressions this way.
In passing, get rid of nearby checks that the input has at least
two elements, since preserving them would make some macros less
readable. These are not necessary for correctness, and seem like
premature optimizations.
Author: Andrey M. Borodin <x4mmm@yandex-team.ru>
Discussion: https://postgr.es/m/098A3E67-E4A6-4086-9C66-B1EAEB1DFE1C@yandex-team.ru
Diffstat (limited to 'src/backend/executor/nodeAgg.c')
0 files changed, 0 insertions, 0 deletions