Horton

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

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Melanie Plageman <melanieplageman@gmail.com>
Cc: Andres Freund <andres@anarazel.de>, pgsql-hackers@postgresql.org, Michael Paquier <michael.paquier@gmail.com>
Date: 2022-10-15T02:41:08Z
Lists: pgsql-hackers

Attachments

On Fri, Oct 14, 2022 at 05:09:46PM -0400, Melanie Plageman wrote:
> To summarize, I am in support of renaming SetSingleFuncCall() ->
> InitMaterializedSRF() and SRF_SINGLE_USE_EXPECTED ->
> MAT_SRF_USE_EXPECTED_TUPLE_DESC (or just DESC) as suggested elsewhere in
> this thread. And I think we should eventually consider renaming the
> multi* function names and consider if ExprSingleResult is a good name.

As already mentioned, these have been around for years, so the impact
would be bigger.  Attached is a patch for HEAD and REL_15_STABLE to
switch this stuff with new names, with what's needed for ABI
compatibility.  My plan would be to keep the compatibility parts only
in 15, and drop them from HEAD.
--
Michael

Commits

  1. Remove compatibility declarations for InitMaterializedSRF()

  2. Rename SetSingleFuncCall() to InitMaterializedSRF()

  3. Create routine able to set single-call SRFs for Materialize mode