hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Deepak Jaiswal <djais...@hortonworks.com>
Subject Re: Review Request 63427: HIVE-17396
Date Thu, 02 Nov 2017 20:40:05 GMT


> On Nov. 1, 2017, 3:48 a.m., Lefty Leverenz wrote:
> > common/src/java/org/apache/hadoop/hive/conf/HiveConf.java
> > Lines 3017-3018 (patched)
> > <https://reviews.apache.org/r/63427/diff/2/?file=1873282#file1873282line3017>
> >
> >     Please spell out the acronyms TS, DPP, and nDVs in the parameter description.
> >     
> >     Also, the description isn't clear to me -- "to control if ..." what?  Something
about reduction, or feeding?  (Forgive my ignorance, I'm just looking at the grammatical structure
of the sentence along with the parameter name.)
> 
> Deepak Jaiswal wrote:
>     Thanks for the feedback. How about this below?
>     
>     The factor to decide if semijoin branch feeding into a TableScan which has an outgoing
Dynamic Partition pruning(DPP) branch based on number of distinct values.
> 
> Lefty Leverenz wrote:
>     Is "feeding" what gets decided?  (In other words, should it be "feeds"?)  If not,
what's the verb that "decide if" refers to?
>     
>     Thanks for spelling out the acronyms.  Keeping DPP in parentheses is good, although
you need a space before the opening parenthesis.  And maybe "pruning" should be "Pruning"
just for consistency.

Thanks. Let me update the patch.


- Deepak


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/63427/#review189778
-----------------------------------------------------------


On Oct. 31, 2017, 1:19 a.m., Deepak Jaiswal wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/63427/
> -----------------------------------------------------------
> 
> (Updated Oct. 31, 2017, 1:19 a.m.)
> 
> 
> Review request for hive, Ashutosh Chauhan and Jason Dere.
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> Dynamic Semijoin Reduction : markSemiJoinForDPP marks unwanted semijoin branches
> 
> In method markSemiJoinForDPP (HIVE-17399), the nDVs comparison should not have equality
as there is a chance that the values are same on both sides and the branch is still marked
as good when it shouldn't be.
> Add a configurable factor to see how useful this is if nDVs on smaller side are only
slightly less than that on TS side.
> 
> 
> Diffs
> -----
> 
>   common/src/java/org/apache/hadoop/hive/conf/HiveConf.java 6631a6e45d 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/TezCompiler.java da30c3b642 
>   ql/src/test/queries/clientpositive/dynamic_semijoin_reduction.q 6cc0a7f7a9 
>   ql/src/test/results/clientpositive/llap/dynamic_semijoin_reduction.q.out 1a1a4d9b2d

> 
> 
> Diff: https://reviews.apache.org/r/63427/diff/2/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Deepak Jaiswal
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message