Flatten nested Python dictionaries, compressing keys. From [StackOverflow](http://stackoverflow.com/questions/6027558/flatten-nested-python-dictionaries-compressing-keys)

Demonstration.md

>>> x = {
        'a':1,
        'b':2,
        'c':{
            'aa':11,
            'bb':22,
            'cc':{
                'aaa':111
            }
        }
    }
>>> flattenDict(x)
{('c', 'aa'): 11, ('c', 'bb'): 22, ('b',): 2, ('a',): 1, ('c', 'cc', 'aaa'): 111}

>>> {'_'.join(k):v for k,v in flattenDict(x).items()}
{'a': 1, 'b': 2, 'c_aa': 11, 'c_bb': 22, 'c_cc_aaa': 111}

>>> flattenDict(x, keyReducer=lambda a,b:hash((a,b))%10000)
{9775: 22, 980: 2, 5717: 1, 2539: 111, 6229: 11}

flatten_dict.py

from collections import *
from itertools import *

same = lambda x:x  # identity function
add = lambda a,b:a+b
_tuple = lambda x:(x,)  # python actually has coercion, avoid it like so

def flattenDict(dictionary, keyReducer=add, keyLift=_tuple, init=()):

    # semi-lazy: goes through all dicts but lazy over all keys
    # reduction is done in a fold-left manner, i.e. final key will be
    #     r((...r((r((r((init,k1)),k2)),k3))...kn))

    def _flattenIter(pairs, _keyAccum=init):
        atoms = ((k,v) for k,v in pairs if not isinstance(v, Mapping))
        submaps = ((k,v) for k,v in pairs if isinstance(v, Mapping))
        def compress(k):
            return keyReducer(_keyAccum, keyLift(k))
        return chain(
            (
                (compress(k),v) for k,v in atoms
            ),
            *[
                _flattenIter(submap.items(), compress(k))
                for k,submap in submaps
            ]
        )
    return dict(_flattenIter(dictionary.items()))