Am 20.11.2006 um 06:35 schrieb Gurjeet Singh:
> But there is another bug: if there are recommendations like idx
> (a,b,c), idx(a,b) and idx(a) it would be a good idea to create just
> idx(a). I will add this to pg_advise as an optional feature.
>
> I'd say it's a new feature request and not a bug :) But I don't
> understand why would you want to not build idx(a,b,c) in such a
> situation? idx(a,b,c) can be useful in places where idx(a,b) or idx
> (a) is required, but the same can't be said about idx(a) or idx
> (a,b) being useful where idx(a,b,c) is required!
>
You are right - that's what I actually meant...
Kai