Re: join to view over custom aggregate seems like it should be faster

"Merlin Moncure" <mmoncure@gmail.com> writes:
> right, i see that it's actually the 'group by' that does it:
> select a, b from foo join (select a, b from bar group by a,b) q using (a,b);
> is enough to keep it from using the index on a,b from bar.  thats too bad...

Some day it'd be nice to be able to reorder grouping/aggregation steps
relative to joins, the way we can now reorder outer joins.  Don't hold
your breath though ... I think it'll take some pretty major surgery on
the planner.

            regards, tom lane