A beta-tester has found another bug in the reason query optimizer enhancements!
One of the changes in version 3.23.0 is that the push-down optimization [1] was made more aggressive. In other words, the conditions on which the push-down optimization depended were relaxed somewhat so that it applies in more situations. In once case, the optimization went too far and generated a query that did not produce the same answer. The problem has now been fixed on trunk [2] and a new pre-release snapshot has been uploaded [3]. We are continuing to study the problem to try to think of new ways to break this optimization. So, there have now been two cases during this release cycle where beta-testers have found obscure cases where an over-eager query optimization resulted in an incorrect answer. Thank you testers! We are currently in "pencils down" for the 3.23.0 release cycle. No new features or enhancements - only bug fixes from here on out. We will start release testing soon. So if you have not yet downloaded and tested a pre-release snapshot, please do so sooner rather than later. This is especially true if you run complex queries that use lots of views, subqueries, and LEFT JOINs. Report any problems or concerns to this mailing list, or directly to me. [1] https://sqlite.org/draft/optoverview.html#pushdown [2] https://sqlite.org/src/info/f08c1731b0b1dddc [3] https://sqlite.org/download.html -- D. Richard Hipp [email protected] _______________________________________________ sqlite-users mailing list [email protected] http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
