Message-ID: <49D3B56D.9050802@zg.htnet.hr>
Date: Wed, 01 Apr 2009 20:41:49 +0200
From: Rikard Pavelic <rikard.pavelic@zg.htnet.hr>
User-Agent: Thunderbird 2.0.0.21 (Windows/20090302)
MIME-Version: 1.0
To: Tom Lane <tgl@sss.pgh.pa.us>
CC: pgsql-performance@postgresql.org
Subject: Re: self join revisited
References: <49D3969F.4030701@zg.htnet.hr> <7884.1238606350@sss.pgh.pa.us>
In-Reply-To: <7884.1238606350@sss.pgh.pa.us>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

Tom Lane wrote:
> Rikard Pavelic <rikard.pavelic@zg.htnet.hr> writes:
>   
>> It would be great if Postgres could rewrite this query
>>     
>
> AFAICS those queries are not going to produce the same results,
> so Postgres would be 100% incorrect to rewrite it like that for you.
>
> (If they do produce the same results, it would depend on a bunch
> of assumptions you have not stated.)
>
> 			regards, tom lane
>   

Can I try again? :)

How hard would it be to teach the planner about preserving uniqueness of
relations in subqueries?
And using that information to remove unnecessary self joins on unique sets?

I can try to rewrite some queries to test it on real data for how much
gain it would provide.

Regards,
Rikard