PyMR is a small python wrapper for map-reduce programming in hadoop streaming mode. It emulates many operations as Java MR framework. One can simply override map() or reduce() method to make code running in streaming mode.
https://github.com/kn45/PyMR

Never bother with input, group by key, which have been handled by PyMR. Just override such method:

• Mapper:
  _setup(self), _map(self, inl), _cleanup(self)
• Reducer:
  _setup(self), _reduce(self, key, values), _cleanup(self)
  self.delim, self.key_fields

cat test_in | python test.py m | sort | python test.py r

Lots. Add more functions such as counter, env_getter etc.