hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vineet Garg (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-17020) Aggressive RS dedup can incorrectly remove OP tree branch
Date Wed, 23 Jan 2019 01:37:00 GMT

    [ https://issues.apache.org/jira/browse/HIVE-17020?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16749373#comment-16749373
] 

Vineet Garg commented on HIVE-17020:
------------------------------------

[~gopalv] [~lirui] Can you take a look?

Latest patch has following changes:
# Earlier fix provided by Gopal to skip RS-SEL(multiple children)-RS pattern
# Fix in RS dedup to remove a specific child instead of removing all children while merging
RS-XX-RS
# Fix in SPDO to better reconcile sort columns added by the optimization (partition cols +
bucketing cols) and original columns (e.g. target table defined with SORT BY).

> Aggressive RS dedup can incorrectly remove OP tree branch
> ---------------------------------------------------------
>
>                 Key: HIVE-17020
>                 URL: https://issues.apache.org/jira/browse/HIVE-17020
>             Project: Hive
>          Issue Type: Bug
>            Reporter: Rui Li
>            Assignee: Vineet Garg
>            Priority: Major
>             Fix For: 4.0.0
>
>         Attachments: HIVE-17020.1.patch, HIVE-17020.2.patch, HIVE-17020.3.patch, HIVE-17020.4.patch,
HIVE-17020.5.patch, HIVE-17020.6.patch, HIVE-17020.7.patch, HIVE-17020.8.patch, HIVE-17020.9.patch
>
>
> Suppose we have an OP tree like this:
> {noformat}
>      ...
>       |
>      RS[1]
>       |
>     SEL[2]
>     /    \
> SEL[3]   SEL[4]
>   |         |
> RS[5]     FS[6]
>   |
>  ... 
> {noformat}
> When doing aggressive RS dedup, we'll remove all the operators between RS5 and RS1, and
thus the branch containing FS6 is lost.



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

Mime
View raw message