RE: AW: question about index cost estimates

Tom Lane wrote:
>
> Zeugswetter Andreas SB <ZeugswetterA@wien.spardat.at> writes:
> > Another metric that would be interesing is a number that represents
> > the divergence of the heap order from the index sort order.
>
> Oh, absolutely!  Got any ideas about how to get that number in a
> reasonable amount of time (not too much more than VACUUM takes now)?
> I've been drawing a blank :-(
>

It seems hard to find the practical theory.
So how about a kind of sample scanning in a separate ANALYZE
command ?

Regards.

Hiroshi Inoue
Inoue@tpf.co.jp