On Fri, 2025-06-13 at 18:41 +0200, Daniel Verite wrote:
> > The main challenge is backwards compatibility.
> Why would it be that bad?
> FTS indexes don't get corrupted that way. You may get different
> lexems before and after the upgrade for some documents, and then
> what?
It would produce different results than if you started from scratch in
v19. It's hard for me to say whether that would be acceptable or not,
but I could see how that could be confusing to users if they notice.
Perhaps release notes are enough?
> The FTS parser had seen user-visible changes in the past, and
> regenerating tsvectors because of that were merely a suggestion.
Interesting, thank you for looking into the history here. It would
certainly be simpler to just make FTS fully collation-aware.
Regards,
Jeff Davis
