On Wed, Sep 11, 2024 at 10:14:27AM -0400, Greg Sabino Mullane wrote: > I dunno - so would we smush them together and return something like: > > "ALTER_PERSISTENCE and COLUMN_REWRITE"
If multiple are set, let's just make it text[], then. > That would be a step backwards for anyone possibly using that integer > programatically to (for example) give a pretty user-facing message about > why the event was triggered. I don't know either how much people are relying on these numbers in applications. If this is like what we do in the regression tests and print it in notice messages within a PL/pgSQL function, that's not going to matter. Or just have a separate function.. Do you have a comment about mentioning the variables or the header in the docs for the stable branches? I'm aware that this is a rare practice, but so is this function's design. My argument is greppability between the code and the docs, mainly, to not miss an update of the docs if more reasons are added. That would be unlikely, but a backpatch of a reason is not impossible ABI-wise. -- Michael
signature.asc
Description: PGP signature
