Skip site navigation (1) Skip section navigation (2)

Re: Statistics on views (execute a plan from within analyze)

From: Martijn van Oosterhout <kleptog(at)svana(dot)org>
To: Oscar Täckström <cotackst(at)uwaterloo(dot)ca>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: Statistics on views (execute a plan from within analyze)
Date: 2007-04-02 14:54:14
Message-ID: 20070402145414.GC9547@svana.org (view raw or flat)
Thread:
Lists: pgsql-hackers
On Fri, Mar 30, 2007 at 12:01:33PM -0400, Oscar Täckström wrote:
> I am working on a course project on implementing collection of statistics
> on views in pgsql. The statistics will be used in conjunction with view
> matching in the optimizer, to improve selectivity estimates.

Interesting idea...

> For this to be possible, I need to be able to execute view definitions
> from within the analyze code (so I can then sample the result set and
> collect the statistics on the sampled tuples). Is there a preferred way
> to do this? A simple solution is to use the SPI, but is this really
> recommended for use from within this code? I don't want to materialize
> the result set, but calculate the statistics directly in memory and just
> throw away the tuples that's not part of the sample.

I beleive you can use SPI to only retreive tuples as they are
calculated, i.e. it doesn't materialise the dataset. However, if this
is a concern, you could always use SPI to create a cursor to iterate
over your set, That will give complete control...

Hope this helps,
-- 
Martijn van Oosterhout   <kleptog(at)svana(dot)org>   http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.

In response to

pgsql-hackers by date

Next:From: Peter EisentrautDate: 2007-04-02 15:40:22
Subject: Re: Implicit casts to text
Previous:From: Simon RiggsDate: 2007-04-02 14:17:32
Subject: One-time plans

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group