Row sampling..

Hi All,

I'm trying to gain a good understanding of how PostgreSQL determines
what to sample when doing a stats analysis on a table. Using PostgreSQL
7.4's pg_stats table I can get a good overall understanding of
variations in the table, but I need to know how PostgreSQL makes it's
choices on what rows to sample. The other thing I also noted, is that I
can change the stats of a column to be as high as 1000, but PostgreSQL
still may not sample all 1000 elements..

can someone help me gain a good understanding of that area of Postgres
so that I can make better choices on optimizing?

--
Chris Bowlby <chris(at)pgsql(dot)com>
PostgreSQL Inc.