Am Dienstag, 18. März 2008 schrieb Tom Lane:
> Or we could apply Peter's patch more or less as-is, but I don't like
> that. I don't think it solves the stated problem: if you know that CASE
> branches 3 and 5 don't match, that still doesn't help you in a monster
> query with lots of CASEs. I think we can and must do better.
Yeah, that and the other reason I sort of gave up on this approach is that it
is nearly impossible to find some good terminology that works for all callers
of select_common_type() (VALUES, UNION, JOIN, IN, CASE, ARRAY, COALESCE,
GREATEST, according to my notes). A pointer into the statement would
certainly be much nicer.
