You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@hive.apache.org by "Pengcheng Xiong (JIRA)" <ji...@apache.org> on 2018/11/05 18:27:00 UTC

[jira] [Commented] (HIVE-20867) Rewrite INTERSECT into LEFT SEMI JOIN instead of UNION + Group by

    [ https://issues.apache.org/jira/browse/HIVE-20867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16675550#comment-16675550 ] 

Pengcheng Xiong commented on HIVE-20867:
----------------------------------------

I have some questions about this jira. Could you share your design document on this? I assumed that we compared several candidates when we made the decision, and lefts semi join was one of them. We chose union-based one because a) a similar approach can be applied to except(all) as well, thus we have better code reuse. b) when we have more then 2 branches as the inputs of intersect, we assume that in the future those branches can be executed in parallel. Comparing with left-semi join one, we need to do the join one by one.

> Rewrite INTERSECT into LEFT SEMI JOIN instead of UNION + Group by
> -----------------------------------------------------------------
>
>                 Key: HIVE-20867
>                 URL: https://issues.apache.org/jira/browse/HIVE-20867
>             Project: Hive
>          Issue Type: Improvement
>          Components: Query Planning
>    Affects Versions: 4.0.0
>            Reporter: Vineet Garg
>            Assignee: Vineet Garg
>            Priority: Major
>




--
This message was sent by Atlassian JIRA
(v7.6.3#76005)