Mailing-List: contact mapreduce-dev-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: mapreduce-dev@hadoop.apache.org
Date: Thu, 8 Nov 2012 19:56:13 +0000 (UTC)
From: "Mark Fuhs (JIRA)" <jira@apache.org>
To: mapreduce-dev@hadoop.apache.org
Message-ID: <1972688295.88518.1352404573062.JavaMail.jiratomcat@arcas>
Subject: [jira] [Created] (MAPREDUCE-4782) NLineInputFormat skips first line
 of last InputSplit
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

Mark Fuhs created MAPREDUCE-4782:
------------------------------------

             Summary: NLineInputFormat skips first line of last InputSplit
                 Key: MAPREDUCE-4782
                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-4782
             Project: Hadoop Map/Reduce
          Issue Type: Bug
          Components: client
    Affects Versions: 0.22.0, trunk
         Environment: job.setMapperClass(Mapper.class);  // just pass text lines through to output
job.setInputFormatClass(NLineInputFormat.class);
NLineInputFormat.setNumLinesPerSplit(job, 100);
NLineInputFormat.setInputPaths(job, "/path/to/a_file_with_many_lines.txt");
            Reporter: Mark Fuhs


NLineInputFormat creates FileSplits that are then used by LineRecordReader to generate Text values. To deal with an idiosyncrasy of LineRecordReader, the begin and length fields of the FileSplit are constructed differently for the first FileSplit vs. the rest.

After looping through all lines of a file, the final FileSplit is created, but the creation does not respect the difference of how the first vs. the rest of the FileSplits are created.

This results in the first line of the final InputSplit being skipped. I've created a patch to NLineInputFormat, and this fixes the problem.

--
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