hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Carter Shanklin (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-7097) The Support for REGEX Column Broken in HIVE 0.13
Date Tue, 20 May 2014 15:46:39 GMT

    [ https://issues.apache.org/jira/browse/HIVE-7097?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14003492#comment-14003492
] 

Carter Shanklin commented on HIVE-7097:
---------------------------------------

Sun,

The issue is related to HIVE-6013. Harish and I debated about this change so you can put all
the blame on me.

We decided to go this path because:
1. We had a user who was trying to import about 50,000 tables from existing databases that
contained all kinds of strange characters in column names.
2. The new behavior is consistent with SQL standards
3. Most Hive users did not know about the regex feature and did not use it.
4. Other databases allow .* as part of the column name.

It's worthwhile for others to give their opinion on this. Personally I think the breaking
change is better in the long run. Can you give more detail about why you favor the old path?
Is it because of Shark compatibility? Something else?

> The Support for REGEX Column Broken in HIVE 0.13
> ------------------------------------------------
>
>                 Key: HIVE-7097
>                 URL: https://issues.apache.org/jira/browse/HIVE-7097
>             Project: Hive
>          Issue Type: Bug
>          Components: Query Processor
>    Affects Versions: 0.13.0
>            Reporter: Sun Rui
>
> The Support for REGEX Column is OK in HIVE 0.12, but is broken in HIVE 0.13.
> For example:
> {code:sql}
> select `key.*` from src limit 1;
> {code}
> will fail in HIVE 0.13 with the following error from SemanticAnalyzer:
> {noformat}
> FAILED: SemanticException [Error 10004]: Line 1:7 Invalid table alias or column reference
'key.*': (possible column names are: key, value)
> {noformat}
> This issue is related to HIVE-6037. When set "hive.support.quoted.identifiers=none",
the issue will be gone.
> I am not sure the configuration was intended to break regex column. But at least the
documentation needs to be updated: https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Select#LanguageManualSelect-REGEXColumnSpecification
> I would argue backward compatibility is more important.



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

Mime
View raw message