DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=message-id:date:from:to:subject:cc:in-reply-to:mime-version
	:content-type:content-transfer-encoding:content-disposition
	:references;
	b=t6aPhOHBCC4RBIj3FAfnfW7D2yiVp/hyjDQMZBou6FiiVbcG+t8YPJK8CDTyQyRp7e
	9rltb5iXBsG2DM70VkW+zIhA6L8TrRvzbdE7Q7Mggm+2cLmRB3quzaPreBWz2mRIDtPO
	2le3ZzNxtLkWoAY7fxP3tmiDGYAYiSbeDAV2g=
Message-ID: <84f0acdb0808071604j73992accp1f770c4c7deec5d9@mail.gmail.com>
Date: Thu, 7 Aug 2008 18:04:58 -0500
From: "Joshua Shanks" <jjshanks@gmail.com>
To: "Gregory Stark" <stark@enterprisedb.com>
Subject: Re: query planner not using the correct index
Cc: "Tom Lane" <tgl@sss.pgh.pa.us>,
	"Craig Ringer" <craig@postnewspapers.com.au>,
	pgsql-performance@postgresql.org
In-Reply-To: <874p5wsb74.fsf@oxford.xeocode.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
References: <84f0acdb0808061435q5ebb607cjc311f1eb4da841a@mail.gmail.com>
	<489A623A.4090400@postnewspapers.com.au>
	<84f0acdb0808062014p29cf7482ied332225a0a92c63@mail.gmail.com>
	<489A6AF5.50908@postnewspapers.com.au>
	<84f0acdb0808070811l43a3da9eq240d272f8ddd105c@mail.gmail.com>
	<27696.1218143560@sss.pgh.pa.us>
	<84f0acdb0808071501n15b0182ga7eca7b1be37d705@mail.gmail.com>
	<874p5wsb74.fsf@oxford.xeocode.com>

On Thu, Aug 7, 2008 at 5:38 PM, Gregory Stark <stark@enterprisedb.com> wrote:
> Measuring n_distinct from a sample is inherently difficult and unreliable.
> When 98% of your table falls into those categories it's leaving very few
> chances for the sample to find many other distinct values.
>
> I haven't seen the whole thread, if you haven't tried already you could try
> raising the statistics target for these columns -- that's usually necessary
> anyways when you have a very skewed distribution like this.
>

I did some tweaking on default_statistics_target earlier in the thread
with no luck. I just retried it with default_statistics_target set to
500 and did the VACUUM ANALYZE on the other table this time and
started to see better results and more of the behavior I would expect.

Is there a way to set the stats target for just one column? That seems
like what we might need to do.

> Yeah, Heikki's suggested having a kind of "branch" plan node that knows how
> where the break-point is between two plans and can call the appropriate one.
> We don't have anything like that yet.
>

Is this already on a todo list or is there a bug for it?