Bugs item #2770608, was opened at 2009-04-17 14:02 Message generated for change (Comment added) made by mlkersten You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=482468&aid=2770608&group_id=56967
Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: SQL/Core Group: SQL "stable" Status: Open Resolution: None Priority: 5 Private: No Submitted By: Roberto Cornacchia (cornuz) Assigned to: Martin Kersten (mlkersten) Summary: SQL: mitosis+reorder (was 'mitosis') goes wild Initial Comment: Many of my recent experiments showed wrong/unexpected behaviours when executed with the mitosis optimizer on (default in development). However, it was always difficult to come up with a reproducible bug. This one is very easy to reproduce. It executes a primary key self-join on the system-table 'types'. The expected result is the relation itself. sql>select * from types t1, types t2 where t1.id = t2.id; 26 tuples The count is indeed correct, but there is no output. The EXPLAIN of this query is something fun to see, I won't paste it here as it is very long. ---------------------------------------------------------------------- >Comment By: Martin Kersten (mlkersten) Date: 2009-07-20 21:03 Message: Test file added BugTracker-2009/Tests/reorder.SF-2770608.sql ---------------------------------------------------------------------- Comment By: Martin Kersten (mlkersten) Date: 2009-07-20 21:03 Message: Test file added BugTracker-2009/Tests/reorder.SF-2770608.sql ---------------------------------------------------------------------- Comment By: Martin Kersten (mlkersten) Date: 2009-06-07 21:24 Message: The monetdb.conf is extended with the Mitosis-mergetable-reorder-dataflow pipeline. It aims at better locality and smaller partitions to improve processing. The results of the SQL test run shows errors in the following tests: ATIS/select-simple-join wisconsin/wisconsin-select test/marcin1 huge-expression orderby-in-function dependencies ---------------------------------------------------------------------- Comment By: Roberto Cornacchia (cornuz) Date: 2009-06-07 21:00 Message: A comment about this bug report: I do realize that the reorder optimizer is now known to be buggy and disabled. This could mean the bug report can be closed, as it addresses 'experimental' code. However, because it seems the problem isn't just in 'reorder' itself, but it is its interaction with mitosis that cause the problem, it may make sense to investigate this issue a bit further. In addition, I consider the bug report valid as the experimental code it addresses is very well reachable (without any warning to the user) also in the stable version. ---------------------------------------------------------------------- Comment By: Roberto Cornacchia (cornuz) Date: 2009-06-07 20:45 Message: It seems that my first bug report was not accurate. At that time, the 'reorder' optimizer was still on by default in the development version. Only later it was discovered to be buggy and disabled. The problem reported is still there, but it shows up only when mitosis and reorder are BOTH enabled. Disabling one of the two makes the problem disappear. sql>set optimizer='inline,remap,evaluate,costModel,coercions,emptySet,mitosis,aliases,mergetable,deadcode,constants,commonTerms,reorder,joinPath,deadcode,reduce,garbageCollector,dataflow,history,multiplex'; 0 tuples sql>select * from types t1, types t2 where t1.id = t2.id; 26 tuples ---------------------------------------------------------------------- Comment By: Stefan Manegold (stmane) Date: 2009-06-07 19:28 Message: Re-opened to remind us that we should add a test for this bug report, e.g., using the new features to set optimizers at server start-up (or at least give a reason why none can / need to be added). ---------------------------------------------------------------------- Comment By: Martin Kersten (mlkersten) Date: 2009-06-07 17:28 Message: this particular example works on the current version. Given the overhaul of Mitosis (et.al.) it is closed until a better example is given. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=482468&aid=2770608&group_id=56967 ------------------------------------------------------------------------------ Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge _______________________________________________ Monetdb-bugs mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/monetdb-bugs
