It seems that one topic of discussion regarding the optimizer
is that it dosen't know how many rows are in the table untill
an analyze is done. Doing SELECT COUNT(*) seems like something
often done that also requires the number of rows in the table
to be known. Perhaps it would be prudent to keep a running
total of the number of active rows for each table cashed?
Clark
Jordan Krushen wrote:
> I know that MySQL doesn't actually hit the tables when doing a SELECT
> COUNT(*), and it returns very quickly as a result.. On large Postgres
> tables, COUNT(*) is horrendously slow. What I'm wondering is if anybody
> here knows a faster way to get the row count, be it an internal Postgres
> variable one can access, or if a COUNT(small_primary_key) would be faster.
> Any ideas?
>