[ 
https://issues.apache.org/jira/browse/HIVE-18140?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16264276#comment-16264276
 ] 

Zoltan Haindrich commented on HIVE-18140:
-----------------------------------------

HIVE-18108 also collaborates to mess-up this thing...I'll attach the refactored 
fix here - which is designed to not break many case ; it started fixing this 
issue because its what naturally does ; I started it as HIVE-18015 ... but it 
looks like it will be safer to separate the refactor and fixing all the issues.

> Partitioned tables statistics can go wrong in basic stats mixed case
> --------------------------------------------------------------------
>
>                 Key: HIVE-18140
>                 URL: https://issues.apache.org/jira/browse/HIVE-18140
>             Project: Hive
>          Issue Type: Sub-task
>            Reporter: Zoltan Haindrich
>
> suppose the following scenario:
> * part1 has basic stats {{RC=10,DS=1K}}
> * all other partition has no basic stats (and a bunch of rows)
> then 
> [this|https://github.com/apache/hive/blob/d9924ab3e285536f7e2cc15ecbea36a78c59c66d/ql/src/java/org/apache/hadoop/hive/ql/stats/StatsUtils.java#L378]
>  condition would be false; which in turn produces estimations for the whole 
> partitioned table: {{RC=10,DS=1K}}



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to