Re: Implement targetlist SRFs using ROWS FROM() (was Changed SRF in targetlist handling)

Andres Freund <andres(at)anarazel(dot)de> writes:
> Attached is a significantly updated patch series (see the mail one up
> for details about what this is, I don't want to quote it in its
> entirety).

I've reviewed the portions of 0005 that have to do with making the parser
mark queries with hasTargetSRF. The code as you had it was wrong because
it would set the flag as a consequence of SRFs in function RTEs, which
we don't want. It seemed to me that the best way to fix that was to rely
on the parser's p_expr_kind mechanism to tell which part of the query
we're in, whereupon we might as well make the parser act more like it does
for aggregates and window functions and give a suitable error at parse
time for misplaced SRFs. The attached isn't perfect, in that it doesn't
know about nesting restrictions (ie that SRFs must be at top level of a
function RTE), but we could improve that later if we wanted, and anyway
it's definitely a good bit nicer than before.

This also incorporates the part of 0002 that I agree with, namely
disallowing SRFs in UPDATE, since check_srf_call_placement() naturally
would do that.

I also renamed the flag to hasTargetSRFs, which is more parallel to
hasAggs and hasWindowFuncs, and made some effort to use it in place
of expression_returns_set() searches.

I'd like to go ahead and push this, since it's a necessary prerequisite
for either approach we might adopt for the rest of the patch series,
and the improved error reporting and avoidance of expensive
expression_returns_set searches make it a win IMO even if we were not
planning to do anything more with SRFs.

regards, tom lane