On 27 March 2018 at 13:26, Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:
> David Rowley wrote:
>
>> Anyway, the options are not zero for anyone who is strongly affected
>> with no other workaround. They just need to disable parallel query,
>> which to me seems fairly similar to the 8.4 release note's "the
>> previous behavior can be restored by disabling enable_hashagg"
>
> synchronized_seqscans is another piece of precedent in the area, FWIW.
This is true. I guess the order of aggregation could be made more
certain if we remove the cost based optimiser completely, and just
rely on a syntax based optimiser. Unfortunately, even with that
UPDATE will still reuse free space created by DELETE+VACUUM, so we'd
need to get rid of VACUUM too. Of course, I'm not being serious, but I
was just trying to show there's a lot more to be done if we want to
make the aggregate input order deterministic without an ORDER BY in
the aggregate function.
--
David Rowley http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
