announce: Select without replacement (paper)

Dear all,

for anyone interested, here's a discussion about the CBO clever use of the statistic formula that gives the "expected number of distinct values when selecting from a bag without replacement" for cardinality estimation:

http://www.adellera.it/investigations/select_without_replacement/index.html

That formula is used for DISTINCT/GROUP BY and for joins.

The paper focuses especially on the multi-column join case, and discusses both the "old" and the "new" algorithm (aka "multicolumn join selectivity sanity check").

For the "old" one, it is also discussed why swapping the join predicates of the SQL statement may surprisingly change the estimated cardinality.

Hope that someone will find it informing and useful :) Alberto

-- 
Alberto Dell'Era
"the more you know, the faster you go"
--
http://www.freelists.org/webpage/oracle-l