Mailing-List: contact dev-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Date: Wed, 5 Sep 2012 19:03:09 +1100 (NCT)
From: "Navis (JIRA)" <jira@apache.org>
To: hive-dev@hadoop.apache.org
Message-ID: <669178384.37906.1346832189417.JavaMail.jiratomcat@arcas>
In-Reply-To: <1303268574.4727.1346122747728.JavaMail.jiratomcat@arcas>
Subject: [jira] [Commented] (HIVE-3411) Filter predicates on outer join
 overlapped on single alias is not handled properly
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HIVE-3411?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13448563#comment-13448563 ] 

Navis commented on HIVE-3411:
-----------------------------

A boolean tag for filter result is replaced with a byte tag, limiting number of aliases for single join to max 8. 
I'm not yet decided whether to extended the limit or to throw exception at compile time.
                
> Filter predicates on outer join overlapped on single alias is not handled properly
> ----------------------------------------------------------------------------------
>
>                 Key: HIVE-3411
>                 URL: https://issues.apache.org/jira/browse/HIVE-3411
>             Project: Hive
>          Issue Type: Sub-task
>          Components: Query Processor
>    Affects Versions: 0.10.0
>         Environment: ubuntu 10.10
>            Reporter: Navis
>            Assignee: Navis
>            Priority: Minor
>         Attachments: HIVE-3411.1.patch.txt
>
>
> Currently, join predicates on outer join are evaluated in join operator (or HashSink for MapJoin) and the result value is tagged to end of each values(as a boolean), which is used for joining values. But when predicates are overlapped on single alias, all the predicates are evaluated with AND conjunction, which makes invalid result. 
> For example with table a with values,
> {noformat}
> 100 40
> 100 50
> 100 60
> {noformat}
> Query below has overlapped predicates on alias b, which is making all the values on b are tagged with true(filtered)
> {noformat}
> select * from a right outer join a b on (a.key=b.key AND a.value=50 AND b.value=50) left outer join a c on (b.key=c.key AND b.value=60 AND c.value=60);
> NULL	NULL	100	40	NULL	NULL
> NULL	NULL	100	50	NULL	NULL
> NULL	NULL	100	60	NULL	NULL
> -- Join predicate
> Join Operator
>   condition map:
>        Right Outer Join0 to 1
>        Left Outer Join1 to 2
>   condition expressions:
>     0 {VALUE._col0} {VALUE._col1}
>     1 {VALUE._col0} {VALUE._col1}
>     2 {VALUE._col0} {VALUE._col1}
>   filter predicates:
>     0 
>     1 {(VALUE._col1 = 50)} {(VALUE._col1 = 60)}
>     2 
> {noformat}
> but this should be 
> {noformat}
> NULL	NULL	100	40	NULL	NULL
> 100	50	100	50	NULL	NULL
> NULL	NULL	100	60	100	60
> {noformat}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira