Re: Idempotence and "Replication Insensitivity" are equivalent ?

Chris Smith wrote:
[..]
> In the case of variance and standard deviation, there's actually a
> better way than to use an accumulating aggregate function.

What specific algorithm do you have in mind ? What and why is it "better" in comparison to ?

>I'd be
> willing to bet that Oracle actually does it differently; by scanning the
> data twice, once to determine the mean, and again to determine the
> variance.

You'd lose you bet because Oracle computes variance in one pass.

>That's *not* evaluating it as an aggregate function, whatever
> Oracle calls it.
>

[...]
> --
> Chris Smith
Received on Tue Sep 19 2006 - 20:34:31 CEST