You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by Vikram Dixit Kumaraswamy <vi...@gmail.com> on 2013/08/05 19:57:13 UTC

Re: Review Request 12827: HIVE-4611 - SMB joins fail based on bigtable selection policy.

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/12827/
-----------------------------------------------------------

(Updated Aug. 5, 2013, 5:57 p.m.)


Review request for hive, Ashutosh Chauhan, Brock Noland, and Gunther Hagleitner.


Changes
-------

Addressed Gunther's comments.


Bugs: HIVE-4611
    https://issues.apache.org/jira/browse/HIVE-4611


Repository: hive-git


Description
-------

SMB joins fail based on bigtable selection policy. The default setting for hive.auto.convert.sortmerge.join.bigtable.selection.policy will choose the big table as the one with largest average partition size. However, this can result in a query failing because this policy conflicts with the big table candidates chosen for outer joins. This policy should just be a tie breaker and not have the ultimate say in the choice of tables.


Diffs (updated)
-----

  ql/src/java/org/apache/hadoop/hive/ql/ErrorMsg.java 12e9334 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/AbstractSMBJoinProc.java cc9de54 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/AvgPartitionSizeBasedBigTableSelectorForAutoSMJ.java 5320143 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/BigTableSelectorForAutoSMJ.java db5ff0f 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/LeftmostBigTableSelectorForAutoSMJ.java db3c9e7 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/MapJoinProcessor.java cd1b4ad 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/TableSizeBasedBigTableSelectorForAutoSMJ.java b882f87 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/correlation/CorrelationOptimizer.java 3071713 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CommonJoinTaskDispatcher.java e214807 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SortMergeJoinTaskDispatcher.java da5115b 
  ql/src/test/queries/clientnegative/auto_sortmerge_join_1.q c858254 
  ql/src/test/queries/clientpositive/auto_sortmerge_join_15.q PRE-CREATION 
  ql/src/test/results/clientnegative/auto_sortmerge_join_1.q.out 0eddb69 
  ql/src/test/results/clientnegative/smb_bucketmapjoin.q.out 7a5b8c1 
  ql/src/test/results/clientpositive/auto_sortmerge_join_15.q.out PRE-CREATION 

Diff: https://reviews.apache.org/r/12827/diff/


Testing
-------

All tests pass on hadoop 1.


Thanks,

Vikram Dixit Kumaraswamy


Re: Review Request 12827: HIVE-4611 - SMB joins fail based on bigtable selection policy.

Posted by Vikram Dixit Kumaraswamy <vi...@gmail.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/12827/
-----------------------------------------------------------

(Updated Aug. 14, 2013, 7:21 p.m.)


Review request for hive, Ashutosh Chauhan, Brock Noland, and Gunther Hagleitner.


Changes
-------

Addressed Ashutosh's comments.


Bugs: HIVE-4611
    https://issues.apache.org/jira/browse/HIVE-4611


Repository: hive-git


Description
-------

SMB joins fail based on bigtable selection policy. The default setting for hive.auto.convert.sortmerge.join.bigtable.selection.policy will choose the big table as the one with largest average partition size. However, this can result in a query failing because this policy conflicts with the big table candidates chosen for outer joins. This policy should just be a tie breaker and not have the ultimate say in the choice of tables.


Diffs (updated)
-----

  ql/src/java/org/apache/hadoop/hive/ql/ErrorMsg.java 12e9334 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/AbstractSMBJoinProc.java fda2f84 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/AvgPartitionSizeBasedBigTableSelectorForAutoSMJ.java 1bed28f 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/BigTableSelectorForAutoSMJ.java db5ff0f 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/LeftmostBigTableSelectorForAutoSMJ.java db3c9e7 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/MapJoinProcessor.java cd1b4ad 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/TableSizeBasedBigTableSelectorForAutoSMJ.java d33ea91 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/correlation/CorrelationOptimizer.java 3071713 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/CommonJoinTaskDispatcher.java e214807 
  ql/src/java/org/apache/hadoop/hive/ql/optimizer/physical/SortMergeJoinTaskDispatcher.java da5115b 
  ql/src/test/queries/clientnegative/auto_sortmerge_join_1.q c858254 
  ql/src/test/queries/clientpositive/auto_sortmerge_join_15.q PRE-CREATION 
  ql/src/test/results/clientnegative/auto_sortmerge_join_1.q.out 0eddb69 
  ql/src/test/results/clientnegative/smb_bucketmapjoin.q.out 7a5b8c1 
  ql/src/test/results/clientpositive/auto_sortmerge_join_15.q.out PRE-CREATION 

Diff: https://reviews.apache.org/r/12827/diff/


Testing
-------

All tests pass on hadoop 1.


Thanks,

Vikram Dixit Kumaraswamy