Horton

Re: Removing unneeded self joins

Alexander Korotkov <aekorotkov@gmail.com>

View thread
From: Alexander Korotkov <aekorotkov@gmail.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: Andrei Lepikhov <lepihov@gmail.com>, Alena Rybakina <a.rybakina@postgrespro.ru>, Richard Guo <guofenglinux@gmail.com>, Ranier Vilela <ranier.vf@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>, Alexander Lakhin <exclusion@gmail.com>, "Gregory Stark (as CFM)" <stark.cfm@gmail.com>, Michał Kłeczek <michal@kleczek.org>, Dean Rasheed <dean.a.rasheed@gmail.com>
Date: 2025-06-27T12:55:37Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Remove GUC_NOT_IN_SAMPLE from enable_self_join_elimination

  2. Put enable_self_join_elimination into postgresql.conf.sample

  3. Get rid of ojrelid local variable in remove_rel_from_query()

  4. Implement Self-Join Elimination

  5. Revert: Remove useless self-joins

  6. Replace lateral references to removed rels in subqueries

  7. Replace relids in lateral subquery parse tree during SJE

  8. Forbid SJE with result relation

  9. Fix misuse of RelOptInfo.unique_for_rels cache by SJE

  10. Replace the relid in some missing fields during SJE

  11. Revert 56-bit relfilenode change and follow-up commits.

  12. Stabilize timetz test across DST transitions.

  13. Speed up finding EquivalenceClasses for a given set of rels

  14. Fix mark-and-restore-skipping test case to not be a self-join.

Hi, Michael!

On Fri, Jun 27, 2025 at 3:26 AM Michael Paquier <michael@paquier.xyz> wrote:
> On Thu, Jun 26, 2025 at 08:54:55AM +0200, Andrei Lepikhov wrote:
> > Before diving into the pg_hint_plan code, I wonder why you don't have
> > similar issues with the remove_useless_joins. We intentionally designed SJE
> > coupled with the left-join removal feature to avoid such type of complaints:
> >
> > CREATE TABLE test (x integer PRIMARY KEY);
> > EXPLAIN (COSTS OFF)
> > SELECT t1.* FROM test t1 LEFT JOIN test t2 ON (t1.x=t2.x);
> >
> > /*
> >      QUERY PLAN
> > ---------------------
> >  Seq Scan on test t1
> > (1 row)
> > */
> >
> > It seems that this join removal is also beyond the pg_hint_plan control ...
>
> Yeah, remove_useless_joins() has been around since the 9.0 ages as far
> as I know, and it's not bothered the module much in the tests because
> we have unlikely relied on it.
>
> For now the trick I am going to rely on is just disable
> enable_self_join_elimination in the test paths where I rely on the
> same relation and self joins to keep the plans of the regression tests
> stable.  This stuff could create separate relations, but then with
> back-patching in mind that's going to be just extra conflict noise,
> which is always annoying when dealing with plan outputs.  We've done
> that in the past with max_parallel_workers_per_gather or jit, to keep
> the expected plans stable.
>
> The point regarding the search join hook may stand, though.  Perhaps
> somebody should check if we're still OK with this change in the
> context of the self-join work.  I tend to think that we are and I
> agree that removing the joins when calling the hook can show benefits,
> but it may be surprising and users tend to be very noisy with plan
> stability, particularly if some of the FROM aliases get silenced by
> the backend without the module knowing about that.  At least there is
> the trick with SET enable_self_join_elimination available as a last
> resort method.

One thing I could additionally propose it to add hook, which gets
called before self-join elimination.  If pg_hint_plan will use this
hook, it could prevent self-join elimination or track this fact in its
data structures.

On the other hand we didn't do similar thing for
remove_useless_joins().  Not sure if it's justified now.

------
Regards,
Alexander Korotkov
Supabase