On Fri, Oct 4, 2024 at 10:24 AM Robert Haas <robertmh...@gmail.com> wrote: > Interesting. I would not have guessed that. I wonder how it works.
ISTM that we've established a general expectation that you as a user can be fairly imprecise about which specific types you use as constants in your query, while still getting an index scan (provided all of the types involved have opclasses that are part of the same opfamily, and that the index uses one of those opclasses as its input opclass). Imagine how confusing it would be if "SELECT * FROM pgbench_accounts WHERE aid = 5" didn't get an index scan whenever the "aid" column happened to be bigint -- that would be totally unacceptable. The main reason why we have operator classes that are grouped into opfamilies is to allow the optimizer to understand the relationship between opclasses sufficient to enable this flexibility. It's concerning that there's a performance cliff with the patch whenever one of the constants is changed from (say) 2_147_483_647 to 2_147_483_648 -- who will even notice that they've actually mixed two different types of integers here? Users certainly won't see any similar problems in the simple "Var = Const" case, nor will they see problems in the mixed-type IN() list case. -- Peter Geoghegan
