[ 
https://issues.apache.org/jira/browse/HIVE-28082?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17838155#comment-17838155
 ] 

Shohei Okumiya commented on HIVE-28082:
---------------------------------------

It seems to be correct, and those behaviors look intentional as we explicitly 
handle the exceptions.
 * 
[https://github.com/apache/hive/blob/rel/release-4.0.0/ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDAFSum.java#L447-L454]
 * 
[https://github.com/apache/hive/blob/rel/release-4.0.0/ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDAFAverage.java#L551-L556]

> HiveAggregateReduceFunctionsRule could generate an inconsistent result
> ----------------------------------------------------------------------
>
>                 Key: HIVE-28082
>                 URL: https://issues.apache.org/jira/browse/HIVE-28082
>             Project: Hive
>          Issue Type: Bug
>          Components: CBO
>    Affects Versions: 4.0.0-beta-1
>            Reporter: Shohei Okumiya
>            Assignee: Shohei Okumiya
>            Priority: Major
>
> HiveAggregateReduceFunctionsRule translates AVG, STDDEV_POP, STDDEV_SAMP, 
> VAR_POP, and VAR_SAMP. Those UDFs accept string types and try to decode them 
> as floating point values. It is possible that undecodable values exist.
> We found that it could cause inconsistent behaviors with or without CBO.
> {code:java}
> 0: jdbc:hive2://hive-hiveserver2:10000/defaul> SELECT AVG('text');
> ...
> +------+
> | _c0  |
> +------+
> | 0.0  |
> +------+
> 1 row selected (18.229 seconds)
> 0: jdbc:hive2://hive-hiveserver2:10000/defaul> set hive.cbo.enable=false;
> No rows affected (0.013 seconds)
> 0: jdbc:hive2://hive-hiveserver2:10000/defaul> SELECT AVG('text');
> ...
> +-------+
> |  _c0  |
> +-------+
> | NULL  |
> +-------+ {code}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to