Message-ID: <4CB8C0D2.6000000@yandex.ru>
Date: Sat, 16 Oct 2010 01:00:02 +0400
From: Nikolai Zhubr <n-a-zhubr@yandex.ru>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; ru;
	rv:1.9.1.5) Gecko/20091204 Thunderbird/3.0
MIME-Version: 1.0
To: Tom Lane <tgl@sss.pgh.pa.us>
CC: pgsql-performance@postgresql.org
Subject: Re: Index scan / Index cond limitation or ?
References: <4CB7268D.4060901@yandex.ru> <7091.1287167367@sss.pgh.pa.us>
In-Reply-To: <7091.1287167367@sss.pgh.pa.us>
Content-Type: text/plain; charset=KOI8-R; format=flowed
Content-Transfer-Encoding: 7bit

15.10.2010 22:29, Tom Lane:
> Nikolai Zhubr<n-a-zhubr@yandex.ru>  writes:
>> So, "in (1, 2)" condition is not in Index Cond anymore! Why is that? How
>> can I push it back?
>
> It thinks the indexscan condition is sufficiently selective already.
> An = ANY condition like that will force multiple index searches,
> one for each of the OR'd possibilities, so it's far from "free" to add
> it to the index condition.  The planner doesn't think it's worth it.

Aha, ok. It makes sense then. Is this specific case (=ANY in index cond) 
described somewhere with reasonable detail? I always try to RTFM first 
and most of the time I can find pretty good hints in the regular manual 
already (sufficient as a starting point at least) but this specific 
topic seems to be somewhat mysterious.

> Perhaps on your real query it is, but there's not much point in
> debating about the behavior on this toy table; without realistic
> table sizes and up-to-date stats it's impossible to say whether that
> choice is correct or not.
>
>> SELECT version();
>> PostgreSQL 8.3.1, compiled by Visual C++ build 1400
>
> You really, really, really ought to be running 8.3.something-newer.
> We didn't put out the last 11 8.3.x bugfix updates just because
> we didn't have anything better to do.

Yes, I know, and I do appreciate the efforts of postgresql devels to 
create updates for older versions too.
This server is internal-only (so it does not see any real world yet). 
Anyway, I hope to update everything to 9.0.1 soon.

Thank you!

Nikolai
>
> 			regards, tom lane
>
>