[jira] [Commented] (CALCITE-760) Aggregate recommender blows up if row count estimate is too high

["Vladimir Sitnikov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CALCITE-760?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16623291#comment-16623291 ] 

Vladimir Sitnikov commented on CALCITE-760:
-------------------------------------------

[~julianhyde], OOM happens since org.pentaho.aggdes.algorithm.impl.MonteCarloLatticeImpl#chooseAggregate => org.pentaho.aggdes.algorithm.impl.MonteCarloLatticeImpl#costQuery => org.pentaho.aggdes.algorithm.impl.LatticeImpl#getParents is trying "infinite" amount of various aggregates.

org.pentaho.aggdes.algorithm.Algorithm.ParameterEnum#aggregateLimit is basicaly ignored.

Should a priority queue be used there?
Should there be a limit on the number of attempts?

> Aggregate recommender blows up if row count estimate is too high
> ----------------------------------------------------------------
>
>                 Key: CALCITE-760
>                 URL: https://issues.apache.org/jira/browse/CALCITE-760
>             Project: Calcite
>          Issue Type: Bug
>            Reporter: Julian Hyde
>            Assignee: Julian Hyde
>            Priority: Major
>
> If you run the aggregate recommendation algorithm with a value of rowCountEstimate that is wrong and large, the algorithm runs for a long time and eventually fails with "OutOfMemoryError: GC overhead limit exceeded".
> I have added LatticeTest.testLatticeWithBadRowCountEstimate as a test case.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)