[
https://issues.apache.org/jira/browse/HIVE-16298?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15942197#comment-15942197
]
Lefty Leverenz commented on HIVE-16298:
---------------------------------------
Doc note: This adds *hive.stats.correlated.multi.key.joins* to HiveConf.java,
so it needs to be documented in the wiki.
* [Configuration Properties -- Statistics |
https://cwiki.apache.org/confluence/display/Hive/Configuration+Properties#ConfigurationProperties-Statistics]
Added a TODOC3.0 label (which might be changed to 2.2 or 2.3 later).
> Add config to specify multi-column joins have correlated columns
> ----------------------------------------------------------------
>
> Key: HIVE-16298
> URL: https://issues.apache.org/jira/browse/HIVE-16298
> Project: Hive
> Issue Type: Bug
> Reporter: Jason Dere
> Assignee: Siddharth Seth
> Labels: TODOC3.0
> Fix For: 2.2.0
>
> Attachments: HIVE-16298.1.patch, HIVE-16298.2.patch
>
>
> The default row estimation for multi-key joins divides the row estimate by
> the product of the NDVs for each join column, which can cause the row
> estimate to be low. Try adding a config to assume the columns are correlated,
> where we only divide the row estimate by the largest NDV.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
