Re: Counting propositions

From: Mikito Harakiri <mikharakiri_at_iahu.com>
Date: Tue, 22 Jun 2004 08:55:50 -0700
Message-ID: <DPYBc.18$da4.141_at_news.oracle.com>


"Paul" <paul_at_test.com> wrote in message news:CFWBc.19053$NK4.3265189_at_stones.force9.net...
> Isn't "select count(*)" a similar thing? You should explicitly name the
> data columns you want. If you change the candidate keys it may change
> the whole interpretation of the propositions. In which case it may no
> longer be true that you want to count propositions.

Count is unusual aggregate operator. Every other aggregate operator is associated with a binary operation. Indeed, the SUM is iterative application of "+", the MAX is iterative application of "max(x,y)" (or CASE ...). There are few exceptions like AVG, but they can be viewed as redundant (or at least less fundamental) operators.

What operation COUNT is application of? The increment, of course. Unlike the examples above, the increment is unary operation, so the COUNT shouldn't really have an argument!

How does COUNT(distinct x) fits in? It's just abuse of syntax by those who are reluctant to write a slightly lengthier expression with inline view. Received on Tue Jun 22 2004 - 17:55:50 CEST

Original text of this message