Re: COUNT(*) and index-only scans

Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I think what Robert is complaining about is that we won't
> currently consider an index that matches neither any WHERE clauses
> nor ORDER BY, ie, count(*) over the whole table won't get
> considered for an index-only scan, regardless of cost estimates.
I guess the trick would be to get it to consider such plans only
under some conditions, to avoid explosive growth in planning time
for some types of queries.  Some statistics bucket for the number of
non-frozen tuples in the relation, maybe?
-Kevin