-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/25179/#review52079
-----------------------------------------------------------



ql/src/java/org/apache/hadoop/hive/ql/optimizer/optiq/stats/HiveRelMdRowCount.java
<https://reviews.apache.org/r/25179/#comment90814>

    My concern is that the computation of selectivity doesn't take into account 
differences in NDV which can result in under estimation.



ql/src/java/org/apache/hadoop/hive/ql/optimizer/optiq/stats/HiveRelMdUniqueKeys.java
<https://reviews.apache.org/r/25179/#comment90816>

    Currently NDV in the stats is not accurate.
    As you mentioned max(col) - min(col) is more accurate, hence we are less 
likely to see this code in effect due to inaccuracies in NDV.
    Is it possible to add some logging statements for : Table name, column 
name, row count, NDV , min and max ?


- Mostafa Mokhtar


On Sept. 2, 2014, 8:23 p.m., Harish Butani wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/25179/
> -----------------------------------------------------------
> 
> (Updated Sept. 2, 2014, 8:23 p.m.)
> 
> 
> Review request for hive, Gunther Hagleitner, John Pullokkaran, and Mostafa 
> Mokhtar.
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> CBO: more cost model changes
> 
> 
> Diffs
> -----
> 
>   
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/optiq/HiveDefaultRelMetadataProvider.java
>  2c08772 
>   
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/optiq/stats/HiveRelMdRowCount.java
>  PRE-CREATION 
>   
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/optiq/stats/HiveRelMdSelectivity.java
>  df70de2 
>   
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/optiq/stats/HiveRelMdUniqueKeys.java
>  PRE-CREATION 
>   ql/src/java/org/apache/hadoop/hive/ql/parse/SemanticAnalyzer.java 6c293c8 
> 
> Diff: https://reviews.apache.org/r/25179/diff/
> 
> 
> Testing
> -------
> 
> existing tests.
> 
> 
> Thanks,
> 
> Harish Butani
> 
>

Reply via email to