hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From 闫昆 (JIRA) <j...@apache.org>
Subject [jira] [Created] (HIVE-6803) Hive mutil group by results are not normal
Date Tue, 01 Apr 2014 03:50:16 GMT
闫昆 created HIVE-6803:
------------------------

             Summary: Hive mutil group by results are not normal
                 Key: HIVE-6803
                 URL: https://issues.apache.org/jira/browse/HIVE-6803
             Project: Hive
          Issue Type: Bug
          Components: Query Processor
    Affects Versions: 0.9.0
            Reporter: 闫昆


Hi all 
I'm learning hive,I am by looking at the following two SQL syntax tree , there is a problem。

predicate:
                  expr: ((qq rlike '.*(baidu).*') or (qq rlike '.*(sina).*'))
                  type: boolean

explain from telno_qq
insert overwrite local directory '/tmp/s' select telno,count(1)  where qq RLIKE '.*(baidu).*'
group by telno
insert overwrite local directory '/tmp/d' select telno,count(1)  where qq RLIKE '.*(sina).*'
group by telno;

predicate:
                  expr: (qq rlike '.*(baidu).*')
                  type: boolean
predicate:
                  expr: (qq rlike '.*(sina).*')
                  type: boolean

explain from telno_qq
insert overwrite local directory '/tmp/s' select telno  where qq RLIKE '.*(baidu).*' 
insert overwrite local directory '/tmp/d' select telno  where qq RLIKE '.*(sina).*' ;

The first query plan will satisfy the conditions of the two select data into / tmp / s and
/ tmp / d them, but actually I just want to / tmp / s data exist only baidu , / tmp / d exists
only sina data . But the result is to meet or sina baidu records are placed in / tmp / s and
/ tmp / d both of them went to the temporary directory .

The second query plan is right , and the query result is correct



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

Mime
View raw message