From: Harald Fuchs <nospam@sap.com>
Subject: Re: Diffcult query
Date: 22 Mar 2003 10:52:19 +0100
Organization: Linux Private Site
Lines: 38
Message-ID: <pusmtfohik.fsf@srv.protecting.net>
References: <d9Iea.117330$JE5.48428@news2.central.cox.net>
Reply-To: hf303@protecting.net
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.0808 (Gnus v5.8.8) Emacs/20.7
To: pgsql-sql@postgresql.org

In article <d9Iea.117330$JE5.48428@news2.central.cox.net>,
"Dan Winslow" <d.winslow@cox.net> writes:

> Hi folks. This is giving me fits. I feel like it oughta be simple, but
> apparantly its not...I think.

> Given a table :

> create table tablea (
>   code1 varchar(32),
>   code2 varchar(32),
>   cost    int
> );

> and the rows

> code1    code2    cost
> ----------------------------------
> "aaa"        "bbb"    2
> "ddd"        "eee"    3
> "bbb"        "aaa"    6
> "ggg"         "hhh"    4

> I need a ( preferably single ) query that will sum the costs for all unique
> pairs of codes, without regard to column order. That is, for summing
> purposes, rows 1 and 3 are identical, and should produce the sum of 8 for
> the unordered pair("aaa","bbb"). It should also, of course, prevent the case
> where ("bbb","aaa") is considered a seperate pair. Any ideas would be much
> appreciated.

How about this one?

  SELECT CASE WHEN code1 < code2 THEN code1 ELSE code2 END ||
         '\0' ||
         CASE WHEN code1 < code2 THEN code2 ELSE code1 END AS codes,
         SUM(cost) AS costs
  FROM tablea
  GROUP BY codes;