On 24 Aug 2011, at 11:18pm, GB wrote: > Well, that is exactly what the sqlite_stat2 table is meant for. It's > information is supposed to make the query planner able to decide upon > the usefulness of an index. Unfortunately, histogram information is not > collected for the implicit rowid index by the ANALYZE command, so the > planner has to rely on some rule of thumb which in turn seems to favour > the rowid index.
Had you thought of creating an explicit index on the rowid column, then running ANALYZE again ? Simon. _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
