Emitter emits key/value pairs

JSONProtocol parse input/output values as JSON strings

KeyValue is the primary type for interacting with Hadoop.

MapReduceJob is the interface expected by the job runner

StreamProtocol is a set of routines for marshaling and unmarshaling key/value pairs from the input stream. Map Reduce jobs can define their own protoc

TSVProtocol outputs keys as tab-separated lines

As example, just to show we can write our own custom protocols

data sink -- useful for benchmarking