Phil Endecott <spam_from_postgresql_general@chezphil.org> writes:
> I presume that PostgreSQL doesn't try to push the limit clause into the
> subqueries of a UNION ALL in this way. I believe it is safe, isn't it?
Hmm. You don't actually want to push the LIMIT as such into the subplan
--- that would create an extra level of plan node that would simply
waste time at runtime. What you want is for the subquery to be planned
on the assumption that only a small number of tuples will be fetched,
so that "fast start" plans are preferred.
We have all the mechanism for this, but prepunion.c isn't exploiting it.
I'll see what I can do.
regards, tom lane