[
https://issues.apache.org/jira/browse/HIVE-19202?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16466817#comment-16466817
]
zhuwei commented on HIVE-19202:
-------------------------------
[~dvoros] The query which failing is in our production environment and a little
bit complicate. According to the code logic, there are some keyword related:
count/sum/group by/join
> CBO failed due to NullPointerException in HiveAggregate.isBucketedInput()
> -------------------------------------------------------------------------
>
> Key: HIVE-19202
> URL: https://issues.apache.org/jira/browse/HIVE-19202
> Project: Hive
> Issue Type: Bug
> Components: CBO
> Affects Versions: 2.1.1
> Reporter: zhuwei
> Assignee: zhuwei
> Priority: Critical
> Fix For: 3.1.0
>
> Attachments: HIVE-19202.1.patch, HIVE-19202.2.patch
>
>
> I ran a query with join and group by with below settings, COB failed due to
> NullPointerException in HiveAggregate.isBucketedInput()
> set hive.execution.engine=tez;
> set hive.cbo.costmodel.extended=true;
>
> In class HiveRelMdDistribution, we implemented below functions:
> public RelDistribution distribution(HiveAggregate aggregate, RelMetadataQuery
> mq)
> public RelDistribution distribution(HiveJoin join, RelMetadataQuery mq)
>
> But in HiveAggregate.isBucketedInput, the argument passed to distribution is
> "this.getInput()"
> , obviously it's not right here. The right argument needed is "this"
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
