I want to re-implement defaultdict myself. In principle, this feels like a really simple question, and one I might give as a “homework assignment”, but the more I dug in, the more complicated it got. My first attempt looked something like this

class DefaultDict(dict):
    def __init__(self, factory):
        self._factory = factory
        super().__init__()

    def __getitem__(self, key):
        missing = object()
        super_value = super().get(key, missing)
        if super_value is missing:
            return self._factory()

        return super_value

However, this does not pass even the most trivial test suite of

x = DefaultDict(lambda: 'no value :(')
y = collections.defaultdict(lambda: 'no value :(')
x['a'] = 'hello'
y['a'] = 'hello'
print(x['a'] == y['a']) # True
print(x.get('a') == y.get('a')) # True
print(x['b'] == y['b']) # True
print(x.get('b') == y.get('b')) # False; x.get('b') is None

After some research, I found __missing__, but an implementation using it fails the above test suite in exactly the same way

class DefaultDict(dict):
    def __init__(self, factory):
        self._factory = factory
        super().__init__()

    def __missing__(self, key):
        return self._factory()

It seems to me that the only way to do this completely is to re-implement dict methods like get and pop, which I don’t think should be necessary. I glanced at the C implementation of defaultdict, and as far as I can tell, it only implements __missing__ as well, so I highly doubt this is required.

What would be the most “pythonic” way to re-implement defaultdict?