"David G. Johnston" <david.g.johnston@gmail.com> writes:
> I'm more keen on the idea of having the system understand when an ORDER BY
> is missing - that seems like what users are more likely to actually do.
That side of it could perhaps be useful, but not if it's an unintelligent
analysis. If someone has a perfectly safe query written according to
the old-school method:
SELECT string_agg(...) FROM (SELECT ... ORDER BY ...) ss;
they are not going to be too pleased with a nanny-ish warning (much
less an error) saying that the aggregate's input ordering is
underspecified.
I also wonder whether we'd accept any ORDER BY whatsoever, or try
to require one that produces a sufficiently-unique input ordering.
regards, tom lane