[jira] [Commented] (HIVE-5609) possibly revisit change in unhex() behavior

["Josh Sumali (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-5609?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13965441#comment-13965441 ] 

Josh Sumali commented on HIVE-5609:
-----------------------------------

Yes, upgrading to hive 0.12 with this change broke a Tableau query with statement CONCAT(UNHEX('E280A2'), '00000000-0000-0000-0000-000000000360 ').

> possibly revisit change in unhex() behavior
> -------------------------------------------
>
>                 Key: HIVE-5609
>                 URL: https://issues.apache.org/jira/browse/HIVE-5609
>             Project: Hive
>          Issue Type: Bug
>          Components: UDF
>            Reporter: Jason Dere
>
> Got one user case where they got bit by the change in HIVE-2482 where unhex() changed to return Binary, where it used to return Text. Looking at the MySQL unhex() function which I'm assuming Hive's unhex() is based on, MySQL actually returns a string value.   I know that we explicitly document the incompatible change in the Jira, but I'd figure that most users would be using unhex() based on how the mysql version works.  And because it's now returning binary, the resulting value from unhex() is actually much less flexible to use if it's being nested within other UDFs, because binary does not implicitly convert to other types.  
> Is anyone open to the idea of reverting unhex() back to it's original behavior, and providing a separate UDF to return the value as binary as [~appodictic] suggested in  HIVE-2482?



--
This message was sent by Atlassian JIRA
(v6.2#6252)