Hadoop with arbitrary, unsplittable, preexisting binary file

Question

I'm trying to run a MapReduce job over a large set of preexisting binary files. The files are already there and I can't change their format.

Should I write my own InputFormat for this? How can I make a simple InputFormat that simply returns an InputStream so that I can process the file?

Answer 1

I do not think we have a inbuilt InputFormat which ignores splits and feeds the mapper a entire file.

You will need to write your own custom InputFormat. The details for which you can find here