[
https://issues.apache.org/jira/browse/HADOOP-3261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12590579#action_12590579
]
Dennis Kubes commented on HADOOP-3261:
--------------------------------------
For one, the next generation of nutch would benefit from this. If we were to
remove the plugin architecture we could use dependency injection in its place
which would make it easier to change out components and to test various
component pieces.
Right now most of the objects are created through ReflectionUtils. Should be
easy to change that out with some type of factory class which allow object to
be created via Reflection, Spring, something else. It should work for
partitioners and the others as well.
> Allow Mapper and Reducer classes to be created a Spring Beans
> -------------------------------------------------------------
>
> Key: HADOOP-3261
> URL: https://issues.apache.org/jira/browse/HADOOP-3261
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Environment: All
> Reporter: Dennis Kubes
> Assignee: Dennis Kubes
> Attachments: HADOOP-3261-1-20080415.patch, spring.jar
>
>
> This improvement abstracts the creation of Mappers and Reducers in factory
> classes and adds the ability to define a spring beans.xml file in the
> classpath and have the Mapper and Reducer classes be created as spring beans.
> This is transparent to the rest of Hadoop but the benefit is that is allows
> Mapper and Reducer classes to have their dependencies injected versus having
> them hardcoded. This will help in the ability to test various mapreduce jobs.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
