Re: index v. seqscan for certain values

On Mon, Apr 12, 2004 at 15:05:02 -0400,
  Jeremy Dunn <jdunn@autorevenue.com> wrote:
>
> Agreed.  However, given that count(*) is a question that can be answered
> _solely_ using the index (without reference to the actual data blocks),
> I'd expect that the break-even point would be considerably higher than
> the < 3% (~38,000 / ~1.3M) I'm currently getting.  Does PG not use
> solely the index in this situation??

That isn't true. In order to check visibility you need to look at the
data blocks.