Message-ID: <4C8F1FED.3050902@enterprisedb.com>
Date: Tue, 14 Sep 2010 10:10:37 +0300
From: Heikki Linnakangas <heikki.linnakangas@enterprisedb.com>
User-Agent: Mozilla/5.0 (X11; U; Linux i686; fi-FI;
	rv:1.9.1.10) Gecko/20100619 Iceowl/1.0b1 Icedove/3.0.5
MIME-Version: 1.0
To: Tom Lane <tgl@sss.pgh.pa.us>
CC: Gaetano Mendola <mendola@gmail.com>,
 pgsql-performance@postgreSQL.org
Subject: Re: Useless sort by
References: <i6ks04$2jet$1@news.hub.org> <12103.1284389088@sss.pgh.pa.us>
	<4C8E4982.70004@gmail.com> <14518.1284396487@sss.pgh.pa.us>
In-Reply-To: <14518.1284396487@sss.pgh.pa.us>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 13/09/10 19:48, Tom Lane wrote:
> Gaetano Mendola<mendola@gmail.com>  writes:
>> Of course I'm not suggesting to take away the "sort by" and give the user
>> an unsorted result, I'm asking why the the optimizer in cases like:
>
>>     select unique(a) from v_table_with_order_by;
>
>> doesn't takes away the "order by" inside the view and puts it back "rewriting the
>> query like this:
>
>>     select unique(a) from v_table_without_order_by
>>     order by a;
>
> That changes the order in which the rows are fed to unique(a).  The
> principal real-world use for a non-top-level ORDER BY is exactly to
> determine the order in which rows are fed to a function, so we will
> have a revolt on our hands if we break that.

You could check for volatile functions. I think this could be done 
safely. However, it doesn't seem worthwhile, it would be a fair amount 
of code, and it's not usually a good idea to put an ORDER BY in a view 
or subquery anyway unless you also have volatile functions in there, or 
you want to coerce the optimizer to choose a certain plan.

-- 
   Heikki Linnakangas
   EnterpriseDB   http://www.enterprisedb.com