On 29 December 2012 20:57, Stefan Andreatta <s.andreatta@xxxxxxxxxxx> wrote: > Now, the 2005 discussion goes into great detail on the advantages and > disadvantages of this algorithm, particularly when using small sample sizes, > and several alternatives are discussed. I do not know whether anything has > been changed after that, but I know that the very distinct problem, which I > will focus on here, still persists. It's a really hard problem to solve satisfactorily. It's a problem that has been studied in much detail. Yes, the algorithm used is still the same. See the comments within src/backend/commands/analyze.c (IBM Research Report RJ 10025 is referenced there). The general advice here is: 1) Increase default_statistics_target for the column. 2) If that doesn't help, consider using the following DDL: alter table foo alter column bar set ( n_distinct = 5.0); -- Peter Geoghegan http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training and Services -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
