hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jesus Camacho Rodriguez (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-12508) HiveAggregateJoinTransposeRule places a heavy load on the metadata system
Date Tue, 24 Nov 2015 22:44:11 GMT

    [ https://issues.apache.org/jira/browse/HIVE-12508?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15025601#comment-15025601
] 

Jesus Camacho Rodriguez commented on HIVE-12508:
------------------------------------------------

The issue was (if I recall well) that you end up with cycles in the planning graph (between
equivalent sets of expressions) and then a metadata provider can fire up indefinitely.

But I guess that as currently we execute this rule in isolation in Hive, and we know this
rule will not produce cycles, we could close it as the metadata provider will never fire up
indefinitely.

> HiveAggregateJoinTransposeRule places a heavy load on the metadata system
> -------------------------------------------------------------------------
>
>                 Key: HIVE-12508
>                 URL: https://issues.apache.org/jira/browse/HIVE-12508
>             Project: Hive
>          Issue Type: Bug
>          Components: CBO
>    Affects Versions: 2.0.0
>            Reporter: Jesus Camacho Rodriguez
>            Assignee: Jesus Camacho Rodriguez
>         Attachments: HIVE-12508.patch
>
>
> Finding out whether the input is already unique requires a call to areColumnsUnique that
currently (until CALCITE-794 is fixed) places a heavy load on the metadata system. This can
lead to long CBO planning.
> This is a temporary fix that avoid the call to the method till then.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message