Re: BUG #15291: Lateral join has different/suprising semantics (with impure function)

=?utf-8?q?PG_Bug_reporting_form?= <noreply(at)postgresql(dot)org> writes:
> [ LATERAL is ignored when sub-select has no actual lateral references ]

> I can understand why it's happening what is happening, yet I still find it
> to be problematic. In my opinion those two selects should yield same
> results. Hence for impure functions lateral semantics should be preserved
> and not optimized away.

I believe this behavior is effectively required by the SQL spec.
In any case, removing it would pessimize many queries in which writing
an unnecessary LATERAL has no bad side-effects today, so I'm disinclined
to change a behavior that's stood for half a dozen releases now.

regards, tom lane