Subject: Re: Wrong plan for subSELECT with GROUP BY
From: Simon Riggs <simon@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-performance@postgresql.org
In-Reply-To: <23992.1147442716@sss.pgh.pa.us>
References: <446482E4.6050906@ritek.hu>  <23992.1147442716@sss.pgh.pa.us>
Content-Type: text/plain
Date: Mon, 15 May 2006 09:48:33 +0100
Message-Id: <1147682913.3465.232.camel@localhost.localdomain>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit

On Fri, 2006-05-12 at 10:05 -0400, Tom Lane wrote:
> Antal Attila <atesz@ritek.hu> writes:
> > If this is a deficiency of the planner, I'd like to suggest this feature 
> > into the planner.
> 
> This really falls into the category of "you've got to be kidding".

Agreed

> There's no way that it'd be reasonable for the planner to expend cycles
> on every query to look for corner cases like this.  

OT: Should we have a way of telling the optimizer how much time and
effort we would like it to go to? Some of the new optimizations and many
yet to come cover smaller and smaller sub-cases. 

At least internally, we could mark the cost-of-optimization as we go, so
we can play with the external interface later.

-- 
  Simon Riggs             
  EnterpriseDB   http://www.enterprisedb.com