[jira] [Created] (HADOOP-9262) Allow jobs to override the input Key/Value read from a sequence file's headers

["David Parks (JIRA)" <ji...@apache.org>]

David Parks created HADOOP-9262:
-----------------------------------

             Summary: Allow jobs to override the input Key/Value read from a sequence file's headers
                 Key: HADOOP-9262
                 URL: https://issues.apache.org/jira/browse/HADOOP-9262
             Project: Hadoop Common
          Issue Type: Improvement
          Components: io
    Affects Versions: 1.0.3
            Reporter: David Parks
            Priority: Minor


There's no clean way to upgrade a sequence file when the model objects in an existing sequence file change in the development process.

If we could override the Key/Value class types read from the sequence file headers we could write jobs that read in the old version of a model object using a different name (MyModel_old for example) make the necessary updates, and write out the new version of the object (MyModel for example).

The problem we experience now is that we have to hack up the code to match the Key/Value class types written to the sequence file, or manually change the headers of each sequence file.

Versioning model objects every time they change isn't a good approach to development because it introduces the likelyhood of less maintained code using an incorrect, old version of the model object.

--
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