Alvaro Herrera <alvherre@commandprompt.com> writes:
> I'm not sure I follow. Are you suggesting adding a new function,
> similar to pg_class_ownercheck, which additionally checks for temp-ness?
No, I was just suggesting adding the check for temp-ness in cluster()
and cluster_rel() where we do pg_class_ownercheck. We already have the
rel open there and so it's cheap to do the temp-ness check.
I guess it's a question of which path you are more concerned about
making cheap, of course. Your proposal was to filter before putting the
rel into the list at all, which certainly saves cycles when we reject a
remote temp table; but it adds cycles for all other tables. I argue
that the remote-temp-table case is uncommon (else we'd have had many
more trouble reports) and therefore optimizing it at the cost of the
normal case isn't a win.
regards, tom lane