Mailing-List: contact mapreduce-user-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: mapreduce-user@hadoop.apache.org
Received-SPF: pass (nike.apache.org: local policy)
From: Markus Jelsma <markus.jelsma@openindex.io>
Reply-To: markus.jelsma@openindex.io
Organization: Openindex
To: mapreduce-user@hadoop.apache.org
Subject: Renaming WritableComparable and existing SequenceFiles
Date: Wed, 30 Nov 2011 16:44:49 +0100
User-Agent: KMail/1.13.5 (Linux/2.6.32-35-generic; KDE/4.4.5; x86_64; ; )
MIME-Version: 1.0
Content-Type: Text/Plain;
  charset="us-ascii"
Content-Transfer-Encoding: 7bit
Message-Id: <201111301644.49484.markus.jelsma@openindex.io>

Hi,

I've built some SequenceFiles using a custom WritableComparable. I also 
decided to reorganize package structure and ended up renaming the whole thing. 
Since the key and value classes are embedded in the file i cannot read them 
because it either doesn't exist or cannot be cast in case the same classname 
lives in another package name.

I looked in the Job and SequenceFile and InputFormat API's but i cannot find 
something to tell the reader that the class is somewhere else now. Renaming 
using a hexeditor doesn't work as well because the package name's length is 
different now, leading to a lot of bad exceptions such as Unknown codec 
complaints.

I can still load the file by keeping the original WritableComparable and write 
back to disk using the new WritableComparable but that forces me to write some 
temporary code.

How can i tell the reader to use the new (but identical) class for 
WritableComparable when loading a SequenceFile? Or do i have to write some 
conversion code?

Thanks