"Simon Riggs" <simon@2ndquadrant.com> writes:
> Select count(*) could be evaluated against any available index
> sub-tables, since all that is required is to count the rows. That would
> be significantly faster than a full file scan and accurate too.
PostgreSQL stores MVCC information in heap tuples only, so index-only
plans such as you're suggesting can't be used (i.e. we need to check
the heap tuple to see if a particular index entry is visible to the
current transaction).
-Neil