Re: apply_scanjoin_target_to_paths and partitionwise join

On Tue, Oct 1, 2024 at 3:22 AM Andrei Lepikhov <lepihov(at)gmail(dot)com> wrote:
>
> On 24/7/2024 15:22, Ashutosh Bapat wrote:
> > On Wed, Jul 24, 2024 at 9:42 AM Richard Guo <guofenglinux(at)gmail(dot)com> wrote:
> >> Is there a specific query that demonstrates benefits from this change?
> >> I'm curious about scenarios where a partitionwise join runs slower
> >> than a non-partitionwise join.
> >
> > [1] provides a testcase where a nonpartitionwise join is better than
> > partitionwise join. This testcase is derived from a bug reported by an
> > EDB customer. [2] is another bug report on psql-bugs.
> I haven't passed through the patch yet, but can this issue affect the
> decision on what to push down to foreign servers: a whole join or just a
> scan of two partitions?
> If the patch is related to the pushdown decision, I'd say it is quite an
> annoying problem for me. From time to time, I see cases where JOIN
> produces more tuples than both partitions have in total - in this case,
> it would be better to transfer tables' tuples to the main instance
> before joining them.

Sorry for replying late. I somehow didn't notice this.

A join between partitions is pushed down if only partitionwise join is
chosen and a join between partitions won't be pushed down if
partitionwise join is not chosen. Hence this bug affects pushdown as
well.

The CF entry shows as waiting for author. But that isn't the right
status. Will change it to needs review. I think we need a consensus as
to whether we want to fix this bug or not. Since this bug doesn't
affect me anymore, I will just withdraw this CF entry if there is no
interest.

--
Best Wishes,
Ashutosh Bapat