On Wed, Jul 24, 2024 at 6:20 AM Robert Haas <robertmh...@gmail.com> wrote:
> > I note in passing that the last time I saw a customer query with > UPPER() in the join clause was... yesterday. The problems there had > nothing to do with CTYPE, but there's no reason to suppose that it > couldn't have had such a problem. I suspect the reason we don't hear > about ctype problems now is that the collation problems are worse and > happen in similar situations. But if all the collation problems went > away, a subset of the same users would then be unhappy about ctype. I have seen and created indexes on upper() functions a number of times too, and I think this is not an uncommon pattern for case insensitive searching Before glibc 2.28, there was at least one mailing list thread where an unhappy person complained about collation problems; but for a number of years before 2.28 I guess the collation changes were uncommon so it didn’t get enough momentum to be considered a real problem until the problem became widespread a few years ago? https://www.postgresql.org/message-id/flat/BA6132ED-1F6B-4A0B-AC22-81278F5AB81E%40tripadvisor.com I myself would prefer an approach here that sets a higher bar for pg_upgrade not corrupting indexes, rather than saying it’s ok as long as it’s rare -Jeremy
