Content-Type: text/plain;
  charset="iso-8859-1"
From: Richard Huxton <dev@archonet.com>
Organization: Archonet Ltd
To: Gary Stainburn <gary.stainburn@ringways.co.uk>,
	Tomasz Myrta <jasiek@klaster.net>
Subject: Re: efficient count/join query
Date: Fri, 7 Feb 2003 13:36:40 +0000
User-Agent: KMail/1.4.3
Cc: pgsql-sql@postgresql.org
References: <200302071007.28697.gary.stainburn@ringways.co.uk>
	<200302071209.55330.gary.stainburn@ringways.co.uk>
	<200302071257.48028.gary.stainburn@ringways.co.uk>
In-Reply-To: <200302071257.48028.gary.stainburn@ringways.co.uk>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Message-Id: <200302071336.40600.dev@archonet.com>

On Friday 07 Feb 2003 12:57 pm, Gary Stainburn wrote:
>
> Thinking about it, I'm not wanting to perform a join as such, but a merge
> of the two selects below, then some form of group by to sum() the two
> counts.
>
>     select rosid as sid, rojid as jid, count(*) as count
>                   from roster group by sid, jid order by sid, jid;
>     select hsid as sid, hjid as jid, hcount as count
>                   from history order by sid, jid;
>
> so that
>
> 1	2	1
> 1	3	2
>
> and
>
> 1	3	1
> 1	4	2
>
> becomes
>
> 1	2	1
> 1	3	3
> 1	4	2

You want a UNION, something like (untested):

CREATE VIEW all_counts AS
SELECT rosid as sid, rojid as jid, count(*) as all_count FROM roster GROUP =
BY=20
sid,jid
UNION ALL
SELECT hsid,hjid,hcount FROM history;

SELECT sid,jid,SUM(all_count) FROM all_counts GROUP BY sid,jid;

--=20
  Richard Huxton