Clemens Ladisch wrote > For count(*), the database does not need the actual table rows.
I think this is not true, he has a point here: SELECT COUNT(*) WHERE <idx_key>=? needs to examine every index key prefix (excluding at least ROWID) that matches. This may mean reading in the whole index. I think b-trees can store the counts of descendant nodes for every node to solve this issue in O(log n), but I don't see anything like it in the SQLite format. -- Sent from: http://sqlite.1065341.n5.nabble.com/ _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users