Horton

Re: Hybrid Hash/Nested Loop joins and caching results from subplans

David Rowley <dgrowleyml@gmail.com>

View thread View on postgr.es
From: David Rowley <dgrowleyml@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>, Andres Freund <andres@anarazel.de>
Date: 2020-08-19T21:59:38Z
Lists: pgsql-hackers 
On Wed, 19 Aug 2020 at 16:23, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> David Rowley <dgrowleyml@gmail.com> writes:
> > I don't object to making the change. I just object to making it only
> > to put it back again later when someone else speaks up that they'd
> > prefer to keep nodes modular and not overload them in obscure ways.
>
> > So other input is welcome.  Is it too weird to overload SubPlan and
> > Nested Loop this way?  Or okay to do that if it squeezes out a dozen
> > or so nanoseconds per tuple?
>
> If you need somebody to blame it on, blame it on me - but I agree
> that that is an absolutely horrid abuse of NestLoop.  We might as
> well reduce explain.c to a one-liner that prints "Here Be Dragons",
> because no one will understand what this display is telling them.

Thanks for chiming in. I'm relieved it's not me vs everyone else anymore.

> I'm also quite skeptical that adding overhead to nodeNestloop.c
> to support this would actually be a net win once you account for
> what happens in plans where the caching is of no value.

Agreed.

David

Commits

  1. Add Result Cache executor node (take 2)

  2. Add Result Cache executor node

  3. Allow estimate_num_groups() to pass back further details about the estimation

  4. Allow users of simplehash.h to perform direct deletions

  5. Cache if PathTarget and RestrictInfos contain volatile functions

  6. Fix pull_varnos' miscomputation of relids set for a PlaceHolderVar.