current implementation of MapWritable and AbstractMapWritable do not
track class usage. Class name is still serialized in write() to disk
even if no instance of such class exists in stored table anymore.

I propose to add usage tracking to  AbstractMapWritable addToMap(Class
clazz) and provide function removeFromMap which will be called by
MapWritable. I can code such patch,. but i am inexperienced in hadoop
development, What hadoop branch should i target?

Hi,

inline...

2012/3/25 Radim Kolar <[EMAIL PROTECTED]>:
> current implementation of MapWritable and AbstractMapWritable do not track
> class usage. Class name is still serialized in write() to disk even if no
> instance of such class exists in stored table anymore.
>
> I propose to add usage tracking to  AbstractMapWritable addToMap(Class
> clazz) and provide function removeFromMap which will be called by
> MapWritable. I can code such patch,. but i am inexperienced in hadoop
> development, What hadoop branch should i target?

I would first search to see if a JIRA
(https://issues.apache.org/jira/secure/Dashboard.jspa) has already
been raised and if not create one. Generally you want to target trunk
with your patch. Much more information here:
http://wiki.apache.org/hadoop/HowToContribute

Cheers,
 Brock

--
Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/