[jira] Updated: (HADOOP-819) LineRecordWriter should not always insert tab char between key and value

["Runping Qi (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HADOOP-819?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Runping Qi updated HADOOP-819:
------------------------------

    Attachment: patch-819.txt

> LineRecordWriter should not always insert tab char between key and value
> ------------------------------------------------------------------------
>
>                 Key: HADOOP-819
>                 URL: https://issues.apache.org/jira/browse/HADOOP-819
>             Project: Hadoop
>          Issue Type: Improvement
>          Components: mapred
>            Reporter: Runping Qi
>         Assigned To: Runping Qi
>         Attachments: patch-819.txt
>
>
> With the current implementation of LineRecordWriter in TextOutputFormat, the client cannot pass null key/or value to the write function, and a tab char is always inserted between  the key and value. This works fine most time. However, in some 
> cases, one just does not want to have the extra tab char. A common example is that, if I need to implement a utility similar 
> to the unix sort with some fields in the lines as the sort key, I can have my map to extract the sort key from each line and pass the whole line as the value. The reducer just outputs the values and ignore the keys. However, if I use TextOutputFormat, my output will have an extra tab key in each of the lines, which is annoying. 
> A simple solution is that let the write function of LineRecordWriter accept null key argument, and write out the value only if the key is null. 

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.