DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com;
	h=DKIM-Signature:X-Yahoo-Newman-Property:X-YMail-OSG:X-Yahoo-SMTP:Received:From:To:References:In-Reply-To:Subject:Date:Message-ID:MIME-Version:Content-Type:Content-Transfer-Encoding:X-Mailer:Thread-Index:Content-Language;
	b=jJFtvF+sAiABxQOQLOgXhWYYwpb1mO1QsZ33pu3UIdK6KnzzUsPk24lnxy7ols93rvnEocMUit26jWNU5Y1uNleCiRuBltu3HJN/oEdiUNY9GM5yhZ+/pgIvvbC8i74lIPqeAfFAFa5KV7+fen2fSkF1R8lt/gaoIId9dYDECvk=
	;
From: "David Johnston" <polobo@yahoo.com>
To: <pgsql-sql@postgresql.org>
References: 
 <CAOveQuPRqy0NoJkNFMRAhkJoWoc8jHSMz8fKdf56Xjzz5Gw27A@mail.gmail.com>
	<4ECBBFEA.4080005@gmail.com>
In-Reply-To: <4ECBBFEA.4080005@gmail.com>
Subject: Re: Change in 9.1?
Date: Tue, 22 Nov 2011 10:52:11 -0500
Message-ID: <070d01cca92e$b34ab830$19e02890$@yahoo.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
Thread-Index: AQICZe3xtdNghVArmN4STzwbL+53mAJQBEk8lTsNUYA=
Content-Language: en-us

-----Original Message-----
From: pgsql-sql-owner@postgresql.org [mailto:pgsql-sql-owner@postgresql.org]
On Behalf Of Rob Sargent
Sent: Tuesday, November 22, 2011 10:30 AM
To: pgsql-sql@postgresql.org
Subject: Re: [SQL] Change in 9.1?


On 11/22/2011 12:39 AM, Jasmin Dizdarevic wrote:
> Hi,
>
> we have a reporting tool, that sometimes uses this kind of condition.
> ...WHERE a.field = a.field
>
> To explain this: a.field can be filtered by the user. the user can 
> choose some values. if he does, this condition will be build:
> ...WHERE a.field IN (1,2,3)
>
> If the user doesn't choose any values the * = * condition is used.
>
> Since 9.1 we're experiencing problems with this construction. Have 
> there been any changes to the planner regarding this?
>
> Ty
> Regards, Jasmin

If it's a commercial product please name that reporting tool: clearly it's
to be avoided.  If it's an in-house tool clearly it's broken.

------------------------------------------------------------

While probably not optimal Dynamic SQL isn't the easiest thing to construct
and so having the occasional sub-optimal construct shouldn't be taken as
being wrong - it should at least result in a reasonably optimal, and
correct, query plan.

The bigger problem is that the OP has utterly failed to describe what kinds
of "problems" are being experienced AND has not provided enough data to
properly evaluate the situation.  Providing a self-contained test case would
at least allow people to run the query and see what is happening...and at a
minimum a query plan (with analyze ideally) - from both versions - would at
least allow some degree of analysis even without a full test-case. 

Since the data matters proving all of: a test case, AND EXPLAIN ANALYZE
results, AND a description of what is taken as being a problem, is necessary
to provide suggestions.

David J.