[TAJO-972] Broadcast join with left outer join returns duplicated rows.

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.9.0
• Assignee: Hyoungjun Kim
• Reporter: Hyoungjun Kim
• Created: 22/Jul/14 10:16
• Updated: 13/Aug/14 08:44
• Resolved: 13/Aug/14 08:44

Description

If LEFT OUTER JOIN has broadcast table and broadcast target table is left side, every tasks run join operation with all rows in broadcast table. So some tasks match and other tasks doesn't match.
For example:

default>select * from small
id
-----------------
1
2
3

default>select * from large
1
4    <-- Block1 in HDFS
5
...
2    <-- Block2 in HDFS
6

default> select a.id, b.id from small a left outer join large b on a.id = b.id
a.id    b.id
---------------------------
1  1
2  null
3  null
1  null
2  2
3  null