[jira] [Commented] (HIVE-21322) Multiple table LEFT OUTER JOIN results are incorrect when 'is not null' used in WHERE clause.

["James Norvell (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-21322?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16786894#comment-16786894 ] 

James Norvell commented on HIVE-21322:
--------------------------------------

Confirmed that setting {{hive.merge.nway.joins=false}} appears to resolve the issue:
{code:java}
hive> select a.id, b.name from a
    > left outer join c on a.id = c.id
    > left outer join b on a.id = b.id
    > where b.name is not null;

OK
2    b
Time taken: 1.638 seconds, Fetched: 1 row(s)

hive> set hive.merge.nway.joins;
hive.merge.nway.joins=true
hive> set hive.merge.nway.joins=false;

hive> select a.id, b.name from a
    > left outer join c on a.id = c.id
    > left outer join b on a.id = b.id
    > where b.name is not null;
OK
2    b
1    a
Time taken: 9.768 seconds, Fetched: 2 row(s)

{code}

> Multiple table LEFT OUTER JOIN results are incorrect when 'is not null' used in WHERE clause.
> ---------------------------------------------------------------------------------------------
>
>                 Key: HIVE-21322
>                 URL: https://issues.apache.org/jira/browse/HIVE-21322
>             Project: Hive
>          Issue Type: Bug
>          Components: CBO
>    Affects Versions: 2.3.4
>         Environment: Hive 2.3.4 (emr-5.21.0)
>            Reporter: James Norvell
>            Priority: Minor
>         Attachments: explain-plans.txt
>
>
> Reproduction:
> Create tables:
> {code:java}
> create table a (id string); insert into a values (1),(2),(3);
> create table b (id string, name string); insert into b values (1,'a'),(2,'b'),(3,null);
> create table c (id string); insert into c values (11),(22),(33);
> {code}
> When joining a -> b -> c on id, the following query is correct:
> {code:java}
> select a.id, b.name from a 
> left outer join b on a.id = b.id 
> left outer join c on a.id = c.id 
> where b.name is not null;
> OK
> 1    a
> 2    b
> Time taken: 10.231 seconds, Fetched: 2 row(s)
> {code}
> Switching the join order from a -> c -> b results in incorrect results:
> {code:java}
> select a.id, b.name from a 
> left outer join c on a.id = c.id 
> left outer join b on a.id = b.id 
> where b.name is not null;
> OK
> 2    b
> Time taken: 10.321 seconds, Fetched: 1 row(s)
> {code}
> Disabling hive.cbo.enable or changing execution engine to mr avoids the issue:
> {code:java}
> set hive.cbo.enable=false;
> select a.id, b.name from a left outer join c on a.id = c.id left outer join b on a.id = b.id where b.name is not null;
> OK
> 1    a
> 2    b
> Time taken: 9.614 seconds, Fetched: 2 row(s)
> set hive.cbo.enable=true;
> set hive.execution.engine=mr;
> select a.id, b.name from a left outer join c on a.id = c.id left outer join b on a.id = b.id where b.name is not null;
> OK
> 1    a
> 2    b
> Time taken: 29.377 seconds, Fetched: 2 row(s)
> {code}
> Issue doesn't occur when using 'is null':
> {code:java}
> select a.id, b.name from a left outer join c on a.id = c.id left outer join b on a.id = b.id where b.name is null;
> OK
> 3    NULL
> Time taken: 9.673 seconds, Fetched: 1 row(s)
> {code}
> Explain plans for queries attached.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)