On Tuesday 15 November 2005 10:12, Tom Lane wrote:
> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> > Tom Lane wrote:
> >> CLUSTER says "order the table according to the order of the
> >> entries in this index". A partial index doesn't define an
> >> ordering for the whole table, only the rows that have entries in
> >> that index. So it doesn't seem to me that you are asking for
> >> something that has a well defined meaning.
> >
> > I assume it would cluster the part of the table covered by the
> > partial index, and the rest of the table would be in any order.
> > It seems like reasonable behavior, though this is the first
> > request I can remember.
>
> But what is the point? You might as well cluster by a full index.
>
> This is *not* trivial to implement, btw, so one request with no
> justification should not be enough to get it on the TODO list.
Not trivial? Seems to me more like impossible to implement for the
general case which would require you to resolve the situation where
someone requests multiple, overlapping, clustered partial indexes
where the ordering requirements are in conflict.
Cheers,
Steve
