Horton

Re: Ordered Partitioned Table Scans

Simon Riggs <simon@2ndquadrant.com>

View thread View on postgr.es
From: Simon Riggs <simon@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: David Rowley <david.rowley@2ndquadrant.com>, Julien Rouhaud <rjuju123@gmail.com>, Antonin Houska <ah@cybertec.at>, Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2019-03-22T15:48:02Z
Lists: pgsql-hackers 
On Fri, 22 Mar 2019 at 11:39, Tom Lane <tgl@sss.pgh.pa.us> wrote:

> Simon Riggs <simon@2ndquadrant.com> writes:
> > I agree that the issue of mixing sorts at various points will make
> nonsense
> > of the startup cost/total cost results.
>
> Right.
>
> > I don't see LIMIT costing being broken as a reason to restrict this
> > optimization. I would ask that we allow improvements to the important use
> > case of ORDER BY/LIMIT, then spend time on making LIMIT work correctly.
>
> There's not time to reinvent LIMIT costing for v12.  I'd be happy to
> see some work done on that in the future, and when it does get done,
> I'd be happy to see Append planning extended to allow this case.
> I just don't think it's wise to ship one without the other.
>

I was hoping to motivate you to look at this personally, and soon. LIMIT is
so broken that any improvements count as bug fixes in my book.

-- 
Simon Riggs                http://www.2ndQuadrant.com/
<http://www.2ndquadrant.com/>
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Commits

  1. Use Append rather than MergeAppend for scanning ordered partitions.