hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dave Beech (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (MAPREDUCE-5374) CombineFileRecordReader does not set "map.input.*" configuration parameters for first file read
Date Thu, 04 Jul 2013 13:09:47 GMT

     [ https://issues.apache.org/jira/browse/MAPREDUCE-5374?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Dave Beech updated MAPREDUCE-5374:

    Status: Patch Available  (was: Open)

I've uploaded a patch against branch-1. Just to confirm, this bug affects the newer API org.apache.hadoop.mapreduce.lib.input.CombineFileRecordReader.

A quick glance at the mapred version leads me to believe it wouldn't be affected, but I haven't
checked to be certain. I will do this and amend the patch if necessary. 
> CombineFileRecordReader does not set "map.input.*" configuration parameters for first
file read
> -----------------------------------------------------------------------------------------------
>                 Key: MAPREDUCE-5374
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5374
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>    Affects Versions: 1.2.0
>            Reporter: Dave Beech
>            Assignee: Dave Beech
>         Attachments: MAPREDUCE-5374.patch
> The CombineFileRecordReader operates on splits consisting of multiple files. Each time
a new record reader is initialised for a "chunk", certain parameters are supposed to be set
on the configuration object (map.input.file, map.input.start and map.input.length)
> However, the first reader is initialised in a different way to subsequent ones (i.e.
initialize is called by the MapTask directly rather than from inside the record reader class).
Because of this, these config parameters are not set properly and are returned as null when
you access them from inside a mapper. 

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

View raw message