Re: BUG #1252: Optimization of SELECT for NOT NULL case

From: "Alexander Kirpa" <postgres(at)bilteks(dot)com>
To: Bruno Wolff III <bruno(at)wolff(dot)to>
Cc: pgsql-bugs(at)postgresql(dot)org
Subject: Re: BUG #1252: Optimization of SELECT for NOT NULL case
Date: 2004-09-15 20:29:43
Message-ID: 4148D067.21269.72BB1B5B@localhost
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

On 15 Sep 2004, at 12:21, Bruno Wolff III wrote:

> On Tue, Sep 14, 2004 at 00:57:07 +0100,
> PostgreSQL Bugs List <pgsql-bugs(at)postgresql(dot)org> wrote:
> >
> > Description: Optimization of SELECT for NOT NULL case
> >
> > Details:
> >
> > CREATE TABLE x (id int4 NOT NULL);
> > Populate table, for instance, 10M rows;
> > SELECT count(*) from x WHERE x IS NULL;
> > Optimizator should rewrite "x IS NULL" to simple "false"
>
> Based on responses to other optimization requests I have seen, I think
> the answer to this one is going to be that it isn't worth paying the
> cost for every query to check for this case, since no one is going to
> write a query like this except by mistake.
>

Regarding your point of view possible exist reason for remove
optimization for case like below
SELECT count(*) from x where id>1 AND id<0
I basically don't see any significant difference in optimization
for "id IS NULL" and "id>1 AND id<0"

From other point of view not always well know NOT NULL a case
for some columns, especially in case JOIN and complex subselects
or dynamic SQL. Main reason to include this optimization, don't
need any a scan of table in optimized case as result well know
before any work with data from tables.

From last point of view optimization in DB servers need for make
work for application developer (not always with serious knowledge
in mathematics and relational databases) more easy and for
customer more swiftly.

And latest point of view: DB server developer have way for
enhance for instance EXPLAIN statement by displaying hint for
static SQL.

In any case thank you for postgreSQL 8, particularly for
background disk writer and improving using of shared buffers.
Good DB engine have more knowledge about information in
shared buffers and using her, that any disk cache system of OS,
plus avoiding or minimize using cache remove don't need
"moving" between cache and buffers.

Best regards,
Alexander Kirpa

In response to

Responses

Browse pgsql-bugs by date

  From Date Subject
Next Message Bender, Cheryl 2004-09-15 20:53:53 Re: Windows 8.0 beta on Server 2003 Installer problem
Previous Message Tom Lane 2004-09-15 18:07:10 Re: BUG #1252: Optimization of SELECT for NOT NULL case