This is a basic Hadoop word count example refactored to make unit testing easy. The mapper and reducer classes define write() methods which take key, value, and context arguments. The unit test harness classes override these methods, passing in null for context and storing the key, value pairs. This allows the algorithm to be tested outside of a running MapReduce program.