[ 
https://issues.apache.org/jira/browse/HIVE-6013?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13882559#comment-13882559
 ] 

Lefty Leverenz commented on HIVE-6013:
--------------------------------------

I guess I misunderstood.  So with hive.support.quoted.identifiers=column, you 
can use backticks to allow Unicode characters in column names but backticks are 
not required if the names are restricted to alphanumeric plus underscore.  Is 
that right, for both DDL and queries?

And if you omit the backticks, you can use * as a wildcard for "all columns" in 
queries.  But you can't have regex queries unless 
hive.support.quoted.identifiers=none, in which case backticks surround the 
regular expression.

What happens if hive.support.quoted.identifiers=none (in Hive 0.13) and you use 
backticks for a regex, then escape some Unicode characters?  That's probably 
not a realistic use case, but I'm wondering if it's a way to select Unicode 
column names when you've got the wrong configuration setting.

> Supporting Quoted Identifiers in Column Names
> ---------------------------------------------
>
>                 Key: HIVE-6013
>                 URL: https://issues.apache.org/jira/browse/HIVE-6013
>             Project: Hive
>          Issue Type: Improvement
>          Components: Query Processor
>            Reporter: Harish Butani
>            Assignee: Harish Butani
>             Fix For: 0.13.0
>
>         Attachments: HIVE-6013.1.patch, HIVE-6013.2.patch, HIVE-6013.3.patch, 
> HIVE-6013.4.patch, HIVE-6013.5.patch, HIVE-6013.6.patch, HIVE-6013.7.patch, 
> QuotedIdentifier.html
>
>
> Hive's current behavior on Quoted Identifiers is different from the normal 
> interpretation. Quoted Identifier (using backticks) has a special 
> interpretation for Select expressions(as Regular Expressions). Have 
> documented current behavior and proposed a solution in attached doc.
> Summary of solution is:
> - Introduce 'standard' quoted identifiers for columns only. 
> - At the langauage level this is turned on by a flag.
> - At the metadata level we relax the constraint on column names.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Reply via email to