Re: limits of max, min optimization

Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> On 2022-Jul-18, Pavel Stehule wrote:
>> I am trying to fix one slow query, and found that optimization of min, max
>> functions is possible only when there is no JOIN in the query.

> See preprocess_minmax_aggregates() in
> src/backend/optimizer/plan/planagg.c
> Maybe it is possible to hack that code so that this case can be handled
> better.

The comments show this was already thought about:

     * We also restrict the query to reference exactly one table, since join
     * conditions can't be handled reasonably.  (We could perhaps handle a
     * query containing cartesian-product joins, but it hardly seems worth the
     * trouble.)

                        regards, tom lane