[
https://issues.apache.org/jira/browse/FLINK-29657?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17618782#comment-17618782
]
luoyuxia commented on FLINK-29657:
----------------------------------
When support hive dialect, we follow Hive 2.3's behavior since it's wide-used
and stable , so we may miss some thing in other versions.
But for your problem, it's quick to fix. We just need to adjust the logic in
`NumExprProcessor#process`
> Flink hive parser considers literal floating point number differently than
> Hive SQL
> -----------------------------------------------------------------------------------
>
> Key: FLINK-29657
> URL: https://issues.apache.org/jira/browse/FLINK-29657
> Project: Flink
> Issue Type: Sub-task
> Reporter: Runkang He
> Priority: Major
>
> Hive SQL consider literal floating number(such as 1.1) as double, but Flink
> hive parser consider this as decimal, so it causes that some hive udf that
> accepts double arg, will not pass type check in hive parser.
> Hive SQL's behavior:
> hive> explain select 1.1 + false;
> 2022-10-17 16:37:14,286 FAILED: SemanticException [Error 10014]: Line 1:15
> Wrong arguments 'false': No matching method for class
> org.apache.hadoop.hive.ql.udf.generic.GenericUDFOPNumericPlus with
> ({*}double{*}, boolean)
> Flink hive parser's behavior:
> in NumExprProcessor#process, it process non-postfix number as decimal by
> default.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
