[
https://issues.apache.org/jira/browse/DRILL-1102?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14051625#comment-14051625
]
Jinfeng Ni commented on DRILL-1102:
-----------------------------------
Yes, I think so. I run the smoking test (TPCH on SF1) and the automation
test. No regression found.
On Thu, Jul 3, 2014 at 5:43 AM, Aditya Kishore (JIRA) <[email protected]>
> Use same set of keys to distribute rows when planner generates exchange for a
> join operator.
> ---------------------------------------------------------------------------------------------
>
> Key: DRILL-1102
> URL: https://issues.apache.org/jira/browse/DRILL-1102
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Jinfeng Ni
> Assignee: Jinfeng Ni
>
> Today, planner might use a different key sets when generate hash-to-random
> exchange operator for a join's left and right side.
> For instance, if we have join condition :
> L1 = R1 and L2 = R2,
> Planner may generate a plan where the join's left side is hash distributed
> using L1, while the right side is hash distributed using (R1, R2). Such plan
> is not valid, since it would produce wrong query result.
> To do that, we need make sure the planner will use the same set of keys when
> ask for hash distribution for a join's left and right side.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
