Modified the compatibility layer to create JobSpec for a MR job , using the HadoopMapper in a self-read mode.
Thus the jobSpec for a MR job now has 4 operators : Mapper ---M:N-->-- Sorter(External) ---1:1-->-- Reducer ---1:1:->-- Writer
This makes the HadoopReadOperator redundant, but is not deleted as it is useful operator in othe scenarios.
The compatibility layer uses Hadoop Mapper in 'dependent mode' when it is forming a pipeling of MR jobs, where the mapper cannot get its input on its own.
git-svn-id: https://hyracks.googlecode.com/svn/trunk/hyracks@177 123451ca-8445-de46-9d55-352943316053
9 files changed