Is it possible for the partitioner's constructor to accept the
hadoop0.20 context object?
My keys can be complex and would (if possible) like to partition in
say the i'th element of the key( when the key is a tuple) but the
partition method only takes the key and value and there is no access
to the context.
A work around is to dynamically change a class field( before
specifying my class as the partioner class) - is this possible in