Re: New "single-call SRF" APIs are very confusingly named

Hi,

On 2022-10-14 10:28:34 +0900, Michael Paquier wrote:
> On Thu, Oct 13, 2022 at 12:48:20PM -0700, Andres Freund wrote:
> > Maybe something like InitMaterializedSRF() w/
> > MAT_SRF_(USE_EXPECTED_DESC|BLESS)
>
> Or just SetMaterializedFuncCall()?

I think starting any function that's not a setter with Set* is very likely to
be misunderstood (SetReturning* is clearer, but long). This just reads like
you're setting the materialized function call on something.

> Do we always have to mention the SRF part of it once we tell about the
> materialization part?

Yes. The SRF is the important part.

> The latter sort implies the former once a function returns multiple tuples.

There's lot of other other things that can be materialized.

> I don't mind doing some renaming of all that even post-release, though
> comes the question of keeping some compabitility macros for
> compilation in case one uses these routines?

Agreed that we'd need compat. I think it'd need to be compatibility function,
not just renaming via macro, so we keep ABI compatibility.

Greetings,

Andres Freund