[
https://issues.apache.org/jira/browse/HIVE-2116?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13020701#comment-13020701
]
Ron Bodkin commented on HIVE-2116:
----------------------------------
Table b was an HBase table, rather than a traditional HDFS file, if that is
relevant to the issue.
> Optimize map-side scans for right-side of join
> ----------------------------------------------
>
> Key: HIVE-2116
> URL: https://issues.apache.org/jira/browse/HIVE-2116
> Project: Hive
> Issue Type: Improvement
> Reporter: Ron Bodkin
>
> I had a large query like select * from a join b on a.key=b.key where...;
> Table b was too large, so I attempted to optimize by adding constraints on b
> to the where clause, e.g.,
> where b.size>=mn and b.size<=mx and ...;
> However, the Hive 0.8.0 optimizer pushed the constraint on b into the reduce
> phase (defeating its purpose).
> I was able to force Hive to run the optimization map-side by this workaround:
> join (select * from b where size>=mn and size<=mx) b on a.key=b.key where ...;
> But it would be nice for Hive to pull filters on joined records into the map
> phase where possible.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
