[
https://issues.apache.org/jira/browse/MAPREDUCE-6215?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ahmed Radwan updated MAPREDUCE-6215:
------------------------------------
Attachment: apache-sortedMapWritable.patch
This small attached patch should solve this issue where the map is cleared at
the beginning of readFields() which guarantees older entries from previous
calls are removed before new entries are added to the map.
> Map is not cleared in SortedMapWritable.readFields(DataInput in)
> ----------------------------------------------------------------
>
> Key: MAPREDUCE-6215
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-6215
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mrv1, mrv2
> Affects Versions: trunk
> Reporter: Ahmed Radwan
> Attachments: apache-sortedMapWritable.patch
>
>
> The readFields(DataInput in) in SortedMapWritable doesn't reset the
> internally used SortedMap. If this Writable object is reused then entries
> will get accumulated in the map with every call of readFields().
> I have seen this behavior when using SequenceFileInputFormat. See
> SequenceFileRecordReader.java and SequenceFile.Reader.
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
