Re: Pushdown target list below gather node (WAS Re: WIP: Upper planner pathification)

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Robert Haas <robertmhaas(at)gmail(dot)com>
Cc: Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, David Rowley <david(dot)rowley(at)2ndquadrant(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Pushdown target list below gather node (WAS Re: WIP: Upper planner pathification)
Date: 2016-03-16 16:57:48
Message-ID: 20081.1458147468@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Wed, Mar 16, 2016 at 12:36 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> And there is a larger problem with this: I'm not sure that it's
>> appropriate for apply_projection_to_path to assume that the subpath is not
>> shared with any other purposes. If it is shared, and we update the
>> subpath's target in-place, we just broke the other path chains.

> That's true. I don't see an obvious hazard here, because the Gather's
> child came from the rel's partial_pathlist, and the only way it gets
> used from there is to stick the Gather on top of it. So it really
> can't show up anywhere else. I think.

The key question I think is could there ever be more than one Gather
sharing the same subpath?

> (To some lesser extent, apply_projection_to_path is always
> scary like that.)

Right, that's why there's also create_projection_path for when you
aren't sure.

> Mmmph. That seems like a 2-bit solution, but I guess it would work.
> What if we taught create_projection_plan() to elide the Result node in
> that case?

Yeah, I was thinking about the same thing. The comment block above
where you're looking would need some adjustment.

regards, tom lane

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Pavel Stehule 2016-03-16 17:02:35 Re: Idle In Transaction Session Timeout, revived
Previous Message Robert Haas 2016-03-16 16:55:51 Re: Choosing parallel_degree