I know that it is contrary to normal Hadoop operation, but how can I
configure my M/R job to send one complete file to each mapper task? This
is intended to be used on many files in the 1.5 MB range as the first
step in a chain of processes.