Re: constant time count(*) ? - Mailing list pgsql-general

From Dann Corbit
Subject Re: constant time count(*) ?
Date
Msg-id D90A5A6C612A39408103E6ECDD77B8294CE208@voyager.corporate.connx.com
Whole thread Raw
In response to constant time count(*) ?  (Mark Harrison <mh@pixar.com>)
List pgsql-general
This should definitely be a FAQ.

The semantics of MVCC (multi-version concurrency control) means that you
can't just store a number somewhere in the header of the table like some
other database systems do.

Try a count(*) on Oracle and you will see similar behavior.  They use
MVCC also.

> -----Original Message-----
> From: Mark Harrison [mailto:mh@pixar.com]
> Sent: Wednesday, October 15, 2003 11:00 AM
> To: pgsql-general@postgresql.org
> Subject: [GENERAL] constant time count(*) ?
>
>
> We're looking into moving some data from mysql to postgresql,
> and notice that count(*) does not seem to be a constant-time
> function as it seems to be in mysql.
>
> planb=# explain select count(*) from assets;
>                             QUERY PLAN
> ----------------------------------------------------------------
>   Aggregate  (cost=22.50..22.50 rows=1 width=0)
>     ->  Seq Scan on assets  (cost=0.00..20.00 rows=1000
> width=0) (2 rows)
>
> Is there a way to optimize count(*) such that it does not
> have to do a sequential scan?  We use this on some big tables
> and it is slowing down processing quite a lot.
>
> Thanks!
> Mark
>
> --
> Mark Harrison
> Pixar Animation Studios
>
>
> ---------------------------(end of
> broadcast)---------------------------
> TIP 2: you can get off all lists at once with the unregister command
>     (send "unregister YourEmailAddressHere" to
> majordomo@postgresql.org)
>

pgsql-general by date:

Previous
From: Peter Eisentraut
Date:
Subject: Re: constant time count(*) ?
Next
From: Andrew Sullivan
Date:
Subject: Re: constant time count(*) ?