It may not be immediately obvious how to do this sort of local command-line analysis, so let's look at an example.
With the UFO datafile on the local filesystem, execute the following command:
$ cat ufo.tsv | shapetimemapper.rb | sort| shapetimereducer.rb
With a single Unixcommand line, we produced output identical to our previous full MapReduce job. If you look at what the command line does, this makes sense.
Firstly, the input file is sent—a line at a time—to the mapper. The output of this is passed through the Unix sort utility and this sorted output is passed a line at a time to the reducer. This is of course a very simplified representation of our general MapReduce job workflow.
Then the obvious question is why should we bother with Hadoop if we can do equivalent analysis at the command line. The answer of course is our old friend, scale. This simple approach works fine for a file such as the...