On Fri, 16 Jul 2004 12:04:54 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> select sum(x), sum(y), sum(x)-sum(y) from ...
>
>At least since 7.4, the system will notice the duplicate aggregates
>and run only two summations to compute the above, followed by a single
>subtraction at the end. The apparently more intelligent way suggested
>by Jean will have to run three summations, and thus end up being a net
>loss.
Also note that Jean-Luc's
select sum( x), sum( y), sum(x-y) from whatever group by z;
gives a different result in the presence of NULLs.
Servus
Manfred