On Mon, Nov 1, 2010 at 8:16 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> This will make the min/max optimization code more visible to the rest of
> the planner in a couple of ways: aside from being called at two places
> not one, it will have some intermediate state that'll have to be kept in
> PlannerInfo, and the "useful pathkeys" logic will have to be complicit
> in letting paths that match the aggregates' requirements survive. But
> it's not real bad, and it seems a lot better than continuing with the
> fully-at-arms-length approach.
>
> Comments?
+1.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company