[
https://issues.apache.org/jira/browse/HIVE-5056?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13739106#comment-13739106
]
Navis commented on HIVE-5056:
-----------------------------
{noformat}
RS := ReduceSinkOperator, JOIN := JoinOperator, RR := RowResolver
If there are 0:T1(a1,a2) and 1:T2(b1,b2) and JOIN has value expression like
{0:[T1.a1,T2.a2], 1:[T2.b2,T2.b1]}, output of JOIN is like this.
output : expression
_col0 : T1.a1
_col1 : T1.a2
_col2 : T2.b2 (not b1)
_col3 : T2.b1 (not b2)
{noformat}
MapJoinProcessor should remove RSs, which means column
expressions(ExprNodeDescs) are not valid and should be rebased on parent of RSs
(because expressions should be based on RR of parent). Previously, hive found
columns by referencing RR of parents of RSs. In this case, the order is
something like [T1.a1-T2.a2-T2.b1-T2.b2], which results invalid column mapping
like this.
{noformat}
output : expression
_col0 : T1.a1
_col1 : T1.a2
_col2 : T2.b1
_col3 : T2.b2
{noformat}
> MapJoinProcessor ignores order of values in removing RS
> -------------------------------------------------------
>
> Key: HIVE-5056
> URL: https://issues.apache.org/jira/browse/HIVE-5056
> Project: Hive
> Issue Type: Bug
> Components: Query Processor
> Reporter: Navis
> Assignee: Navis
> Attachments: HIVE-5056.D12147.1.patch, HIVE-5056.D12147.2.patch
>
>
> http://www.mail-archive.com/user@hive.apache.org/msg09073.html
--
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
