Michael Fork wrote on 29.12.2009 18:08:
> I have an index scan on a custom function that is returning a wildly
> incorrect row estimate that is throwing off the rest of the query
> planning. The result of the function is roughly unique - there are a
> handful with multiple entries - but the planner is estimating 227,745
> rows. I re-ran ANALYZE on the table and the results did not change.
> Any suggestions on how to get more accurate planner result?
You can add the ROWS nnn option to your create statement to give the planner a hint about the number of rows:
http://www.postgresql.org/docs/current/static/sql-createfunction.html
Check out the /ROWS result_rows/ part.
Thomas