-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/69515/#review211066
-----------------------------------------------------------




ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java
Lines 1335 (patched)
<https://reviews.apache.org/r/69515/#comment295932>

    We may bail out here if op is an instance of ReduceSinkOperator, since this 
will be a new vertex and thus we will not replicate it.
    
    This may help us having less changes in perf driver.



ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java
Lines 1337 (patched)
<https://reviews.apache.org/r/69515/#comment295933>

    What happens if there is a union with three children: first one is the 
normal operator plan, second and third are semijoin branches? I believe this 
logic would leave the third branch in the plan? Probably we do not need the 
quick bail out.


- Jesús Camacho Rodríguez


On Dec. 5, 2018, 11:33 p.m., Vineet Garg wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/69515/
> -----------------------------------------------------------
> 
> (Updated Dec. 5, 2018, 11:33 p.m.)
> 
> 
> Review request for hive, Ashutosh Chauhan and Jesús Camacho Rodríguez.
> 
> 
> Bugs: HIVE-21007
>     https://issues.apache.org/jira/browse/HIVE-21007
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> See JIRA
> 
> 
> Diffs
> -----
> 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java 02cebdc5ac 
>   ql/src/test/results/clientpositive/llap/dynamic_semijoin_reduction_3.q.out 
> cb3740d09a 
>   ql/src/test/results/clientpositive/perf/tez/constraints/query8.q.out 
> d97f9df397 
>   ql/src/test/results/clientpositive/perf/tez/query54.q.out a706e9459e 
>   ql/src/test/results/clientpositive/perf/tez/query8.q.out 9eb50396f9 
> 
> 
> Diff: https://reviews.apache.org/r/69515/diff/1/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Vineet Garg
> 
>

Reply via email to