Re: GiST nitpicks I want to discuss (and maybe eventually fix)

On Mon, 6 Oct 2025 at 23:53, Andrey Borodin <x4mmm(at)yandex-team(dot)ru> wrote:

Thank you for review

> Makes sense. I hope one day we will add a catalog field to track index creation version. This would pave the way to get rid of invalid GiST tuples and return this flag too. We can use it for something better.

In the GIN index we have an index creation version, and it is placed
on the metapage. So does btree, if i'm not mistaken . So, the index
version is stored in data, not in catalog. I doubt we will support two
technologies for one purpose. Since GiST index has no metapage, I
doubt we will be successful here.

>
> Yes, this is Hot-standby-only record. In offline conversation you mentioned that "RelFileLocator locator;" is not needed, only "Oid dbOid;". Maybe let's save a little more bytes?

It is about ResolveRecoveryConflictWithSnapshot. This function
RelFileLocator arg and only uses database oid. Changing this is too
out-of-scope.

--
Best regards,
Kirill Reshke