You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Bryan Jacobs (JIRA)" <ji...@apache.org> on 2014/06/20 17:58:25 UTC
[jira] [Created] (MAPREDUCE-5936) MultipleInputs incorrect output
with copied Path
Bryan Jacobs created MAPREDUCE-5936:
---------------------------------------
Summary: MultipleInputs incorrect output with copied Path
Key: MAPREDUCE-5936
URL: https://issues.apache.org/jira/browse/MAPREDUCE-5936
Project: Hadoop Map/Reduce
Issue Type: Bug
Components: mrv1, mrv2
Reporter: Bryan Jacobs
The MultipleInputs class builds a Map with Path objects as keys and mapper-inputformat combinations as values.
This is not correct behavior. If MultipleInputs.addInputPath is called twice with the same Path and (for example) two different Mapper classes, the second addition will silently overwrite the first.
Expected behavior would be that the input file would be processed one time for each call to addInputPath.
This is necessary for applications which are doing join-like operations: joining a file with itself is valid, and it should not be incumbent on the application developer to recognize when the same Path is included twice to work around this bug.
MultipleInputs should be using a multimap or a map with List values.
--
This message was sent by Atlassian JIRA
(v6.2#6252)