pg.ddx.io  pgsql-bugs@postgresql.org mailing list archive  
help / color / mirror / Atom feed
Memoize in between of two JOIN nodes
3+ messages / 2 participants
[nested] [flat]

* Memoize in between of two JOIN nodes
@ 2025-03-06 11:17 Andrei Lepikhov <lepihov@gmail.com>
  2025-03-07 01:13 ` Re: Memoize in between of two JOIN nodes Richard Guo <guofenglinux@gmail.com>
  0 siblings, 1 reply; 3+ messages in thread

From: Andrei Lepikhov @ 2025-03-06 11:17 UTC (permalink / raw)
  To: pgsql-bugs@lists.postgresql.org; +Cc: Richard Guo <guofenglinux@gmail.com>

Playing with memoisation, I found the case where the Memoize is put over 
a JOIN node (see attachment).
I recall a discussion we had with Richard in which he mentioned [1] that 
this feature is still not implemented and is hard to design.
I'm not sure, but may it be a sign of a potential bug?

[1] 
https://www.postgresql.org/message-id/CAMbWs4_CQR9ikyAhjH0CwzCdp%2BFcRAeoVsi5Oohd7s%3Dc5RZABg%40mail...

-- 
regards, Andrei Lepikhov

Attachments:

  reproduction.sql (application/sql, 1.8K) download

^ permalink  raw  reply  [nested|flat] 3+ messages in thread

* Re: Memoize in between of two JOIN nodes
  2025-03-06 11:17 Memoize in between of two JOIN nodes Andrei Lepikhov <lepihov@gmail.com>
@ 2025-03-07 01:13 ` Richard Guo <guofenglinux@gmail.com>
  2025-03-10 09:22   ` Re: Memoize in between of two JOIN nodes Andrei Lepikhov <lepihov@gmail.com>
  0 siblings, 1 reply; 3+ messages in thread

From: Richard Guo @ 2025-03-07 01:13 UTC (permalink / raw)
  To: Andrei Lepikhov <lepihov@gmail.com>; +Cc: pgsql-bugs@lists.postgresql.org

On Thu, Mar 6, 2025 at 8:17 PM Andrei Lepikhov <lepihov@gmail.com> wrote:
> Playing with memoisation, I found the case where the Memoize is put over
> a JOIN node (see attachment).
> I recall a discussion we had with Richard in which he mentioned [1] that
> this feature is still not implemented and is hard to design.
> I'm not sure, but may it be a sign of a potential bug?

In your case, the Memoize node is added on top of a base relation of a
subquery RTE, not a join relation.  The final plan might be kind of
confusing because the SubqueryScan node is considered trivial and is
removed from the plan tree.

Thanks
Richard





^ permalink  raw  reply  [nested|flat] 3+ messages in thread

* Re: Memoize in between of two JOIN nodes
  2025-03-06 11:17 Memoize in between of two JOIN nodes Andrei Lepikhov <lepihov@gmail.com>
  2025-03-07 01:13 ` Re: Memoize in between of two JOIN nodes Richard Guo <guofenglinux@gmail.com>
@ 2025-03-10 09:22   ` Andrei Lepikhov <lepihov@gmail.com>
  0 siblings, 0 replies; 3+ messages in thread

From: Andrei Lepikhov @ 2025-03-10 09:22 UTC (permalink / raw)
  To: Richard Guo <guofenglinux@gmail.com>; +Cc: pgsql-bugs@lists.postgresql.org

On 7/3/2025 02:13, Richard Guo wrote:
> On Thu, Mar 6, 2025 at 8:17 PM Andrei Lepikhov <lepihov@gmail.com> wrote:
>> Playing with memoisation, I found the case where the Memoize is put over
>> a JOIN node (see attachment).
>> I recall a discussion we had with Richard in which he mentioned [1] that
>> this feature is still not implemented and is hard to design.
>> I'm not sure, but may it be a sign of a potential bug?
> 
> In your case, the Memoize node is added on top of a base relation of a
> subquery RTE, not a join relation.  The final plan might be kind of
> confusing because the SubqueryScan node is considered trivial and is
> removed from the plan tree.
Yes, I understand your point. But I kept in mind a different idea, sorry 
for my bad explanation:
In the case of another subquery, we can't use the memoize node even if 
it would help a lot:

EXPLAIN (ANALYZE, COSTS OFF, TIMING OFF, BUFFERS OFF, SUMMARY ON)
SELECT * FROM t1 LEFT JOIN (
   SELECT t2.* FROM t2 JOIN (
     VALUES (1,1), (2,2)) AS q2(x,y)
   ON (t2.y=q2.y)) q
ON (t1.x = q.x);

see the full case reproduction script in the attachment.
So, the question is: may we play with subquery flattening to let memoize 
caching a join result?
In our previous discussion, you pointed out the problem of reference 
detection. But maybe the SubqueryScan hack can make it simpler?

-- 
regards, Andrei Lepikhov

Attachments:

  demo.sql (application/sql, 1.1K) download

^ permalink  raw  reply  [nested|flat] 3+ messages in thread


end of thread, other threads:[~2025-03-10 09:22 UTC | newest]

Thread overview: 3+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2025-03-06 11:17 Memoize in between of two JOIN nodes Andrei Lepikhov <lepihov@gmail.com>
2025-03-07 01:13 ` Richard Guo <guofenglinux@gmail.com>
2025-03-10 09:22   ` Andrei Lepikhov <lepihov@gmail.com>

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox