On Thu, Sep 10, 2020 at 6:52 PM Peter Eisentraut <peter.eisentr...@2ndquadrant.com> wrote: > > On 2020-09-08 21:33, Christoph Berg wrote: > > IOW, I think we should aim at simply tracking the version, and leave > > it to the admin (with the help of supplied SQL queries) to either > > rebuild indexes or waive them. > > It's certainly safer to track dependency for all indexes and then > carefully create exceptions afterwards.
Do you mean storing the collation version even when those are not relevant, and let client code (or backend command) deal with it? This would require to store a dependency per index and column (or at least if it's a column or an expression to avoid bloating the dependencies too much), as it's otherwise impossible to know if a version mismatch can be safely ignored or not. I'm also wondering how much more complexity it would add to people who want to actively monitor such mismatch using SQL queries.
