-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/38493/#review99815
-----------------------------------------------------------



ql/src/java/org/apache/hadoop/hive/ql/optimizer/correlation/ReduceSinkDeDuplication.java
 (line 180)
<https://reviews.apache.org/r/38493/#comment156786>

    For the hash partitions to match, the partition cols have to exactly match 
between the two RS instances.
    
    For the keys, the rules involve prefix matches. 
    
    One of the RS key set has to be an exact prefix to the other RS key set - 
the ordering is relevant, since sorting by (a,b) cannot merge into sorting by 
(b,a).



ql/src/java/org/apache/hadoop/hive/ql/optimizer/correlation/ReduceSinkDeDuplication.java
 (line 187)
<https://reviews.apache.org/r/38493/#comment156787>

    This won't necessarily enforce order of the keys (a,b) != (b,a)



ql/src/java/org/apache/hadoop/hive/ql/optimizer/correlation/ReduceSinkDeDuplication.java
 (line 190)
<https://reviews.apache.org/r/38493/#comment156788>

    Here, also the ordering is relevant - but they are mutable.
    
    (a,b) != (b,a) ... but one of the mismatched RS entries can be rewritten if 
they're both hash partitioned.


- Gopal V


On Sept. 18, 2015, 3:01 p.m., Ashutosh Chauhan wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/38493/
> -----------------------------------------------------------
> 
> (Updated Sept. 18, 2015, 3:01 p.m.)
> 
> 
> Review request for hive and Gopal V.
> 
> 
> Bugs: HIVE-11132
>     https://issues.apache.org/jira/browse/HIVE-11132
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> Queries using join and group by produce incorrect output when 
> hive.auto.convert.join=false and hive.optimize.reducededuplication=true
> 
> 
> Diffs
> -----
> 
>   
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/correlation/ReduceSinkDeDuplication.java
>  56334ed 
>   ql/src/test/queries/clientpositive/join_grp_diff_keys.q PRE-CREATION 
>   ql/src/test/results/clientpositive/join_grp_diff_keys.q.out PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/38493/diff/
> 
> 
> Testing
> -------
> 
> New q test and regression suite.
> 
> 
> Thanks,
> 
> Ashutosh Chauhan
> 
>

Reply via email to