Re: functions returning sets

Joe Conway <mail(at)joeconway(dot)com> writes:
> Any comment on the patch?

It seems like a bad idea to change the contents of the SRF_XXX() macros
for 7.4.1; if we do that, any existing already-compiled-for-7.4 user
SRFs will be broken, and there's no easy way to catch the problem.
We don't normally require people to recompile user-defined functions for
dot releases anyway.

I would suggest leaving end_MultiFuncCall() with its existing API,
and adding a separate shutdown callback function that is registered
during init_MultiFuncCall and deregistered by end_MultiFuncCall.
(This should be workable without API change since init_MultiFuncCall
can get to the econtext via fcinfo->resultinfo.)

You may not even need to add any fields to FuncCallContext --- consider
passing the fcinfo pointer to the callback, rather than passing the
FuncCallContext pointer.

regards, tom lane