Re: Idempotence and "Replication Insensitivity" are equivalent ?

From: Aloha Kakuikanu <aloha.kakuikanu_at_yahoo.com>
Date: 24 Sep 2006 07:54:15 -0700
Message-ID: <1159109655.870412.287400_at_d34g2000cwd.googlegroups.com>


vc wrote:
> Aloha Kakuikanu wrote:
> > vc wrote:
> > > According to ANSI SQL, nulls are ignored in the aggregate functions
> > > (except count(*)). It may not make sense to you, but it makes sense
> > > to a lot of other folks.
> >
> > Or yeah, those folks defined null as a neutral element for sum
> > aggregate too. Apparently they are unaware of identity:
> >
> > 0 + 0 + ... = 0
>
> The above does not make any obvious sense. What's a "neutral element"
> ?

Ugh,

http://www.google.com/search?hl=en&q=neutral+element+wikipedia&btnG=Google+Search

There are actually 2 issues:
1. How sum is defined on empty set. It is certainly 0, not null. 2. How the sum is defined when there are unknown elements. Unknown elements are interpreted as 0s.

> > What I'm saying is that you have to define different neutral elements
> > for different aggregates in order to maintain elementary consistency
>
> Could you, like, you know, rephrase that ?

OK, I should retract this statement back. What consistency one can hope for when

null != null

And you don't have to go far for the concrete examples, e.g. summing null and x explicitly gives null, while doing this via aggregation gives x! Received on Sun Sep 24 2006 - 16:54:15 CEST

Original text of this message