OrderedDict vs Dict in python

Question

In Tim Peter\'s answer to \"Are there any reasons not to use an ordered dictionary\", he says

OrderedDict is a subclass of dict.

It\'s not

Answer 1

I think the problem with size is due to the fact that there's no __sizeof__ method defined in Python 2.X implementation of OrderedDict, so it simply falls back to dict's __sizeof__ method.

To prove this here I've created a class A here which extends list and also added an additional method foo to check if that affects the size.

class A(list):
    def __getitem__(self, k):
        return list.__getitem__(self, k)
    def foo(self):
        print 'abcde'

>>> a = A(range(1000))
>>> b = list(range(1000))

But still same size is returned by sys.getsizeof:

>>> sys.getsizeof(a), sys.getsizeof(b)
(9120, 9120)

Of course A is going to be slow because its methods are running in Python while list's method will run in pure C.

>>> %%timeit
... for _ in xrange(1000):
...     a[_]
... 
1000 loops, best of 3: 449 µs per loop
>>> %%timeit
for _ in xrange(1000):
    b[_]
... 
10000 loops, best of 3: 52 µs per loop

And this seems to be fixed in Python 3 where there's a well defined __sizeof__ method now:

def __sizeof__(self):
    sizeof = _sys.getsizeof
    n = len(self) + 1                       # number of links including root
    size = sizeof(self.__dict__)            # instance dictionary
    size += sizeof(self.__map) * 2          # internal dict and inherited dict
    size += sizeof(self.__hardroot) * n     # link objects
    size += sizeof(self.__root) * n         # proxy objects
    return size