import itertools

from itertools import groupby
from operator import itemgetter

testdata = [
        {
            'Elements' : [
                { 'Type': 't', 'Id': 1 },
                { 'Type': 'u', 'Id': 2 },
                { 'Type': 't', 'Id': 3 },
                { 'Type': 'u', 'Id': 1 },
            ]
        },
        {
            'Elements' : [
                { 'Type': 'u', 'Id': 1 },
                { 'Type': 't', 'Id': 2 },
                { 'Type': 'u', 'Id': 3 },
                { 'Type': 't', 'Id': 1 },
                { 'Type': 'u', 'Id': 2 },

            ]
        },
    ]

flat = sum((f['Elements'] for f in testdata), [])
sflat = sorted(flat, key=lambda e: (e['Type'], e['Id']))
gflat = [list(g) for k, g in groupby(sflat, key=lambda e: (e['Type'], e['Id']))]

#print("Grouped:")
result = [ list(g) for kf in (itemgetter('Type', 'Id'),)
        for _, g in groupby(sorted(sum((f['Elements'] for f in testdata), []), key=kf), key=kf) ]
assert result == gflat
#print(result)

def _extractor_func(extractor, kwargs):
    def _ident(x):
        return x

    if extractor:
        if len(extractor) > 1:
            raise TypeError('Too many positional arguments. Only 1 allowed.')
        return extractor[0]

    if not kwargs:
        return _ident

    if len(kwargs) != 1:
        raise TypeError('Too many named parameters. Only 1 allowed.')

    for var, code in kwargs.items():
        return eval('lambda {}: {}'.format(var, code))

def select_many(seq, *extractor, **kw):

    extract = _extractor_func(extractor, kw)

    for item in seq:
        for val in extract(item):
            yield val

def group_by(seq, *extractor, **kw):
    extract = _extractor_func(extractor, kw)

    for key, group in itertools.groupby(sorted(seq, key=extract), key=extract):
        yield list(group)

flat2 = list(select_many(testdata, f = "f['Elements']"))
assert flat2 == flat
result2 = group_by(flat2, e = "e['Type'], e['Id']")
assert result2 == result2