Hi, all, I'm new to Hadoop, and I'm finding myself having a hard time creating highly configurable Mapper and Reducer instances, due to the fact that Hadoop seems to require that Mapper and Reducer instances be instantiated through reflection, so I'm sort of wondering whether I'm doing things wrong.
In my application I have a number of composable classes that can be used to generate mapper and reducer instances by the end-user of the library. Such composition can be done either in code or at runtime by interpretation of a script written in a custom DSL, and I'd like to avoid having separate mapper and reducer classes for the different types of construction; this doesn't seem like it should be something that's the responsibility of the map/reduce part of the library. What I'm wondering is, is there any way to simply serialize a Mapper or Reducer object, and have the serialized instance copied, passed around and used everywhere instead of always having the Mapper and Reducer instantiated by reflection? This would greatly simplify library design in my case. Thanks, Kris
