Can I change the base class value in a python subclass?

When subclassing a base type, like float, is it possible to "recalculate" or "reassign" the original value? If I have the following class definiti开发者_JS百科on,

import collections

class MovingAverage(float):
    def __new__(self, initial_value, window):
        self.d = collections.deque([initial_value], window)
        return float.__new__(self, initial_value)
    def add(self, new_value):
        self.d.append(new_value)
        print sum(self.d) / len(self.d)
        # here, I want to _set_ the value of MovingAverage to
        #     sum(self.d) / len(self.d)

When I start with

>>> ma = MovingAverage(10, 3)
>>> ma
10.0

but

>>> ma.add(3)
6.5
>>> ma
10.0

The other class definition I've tried is this:

import collections

class MovingAverage(float):
    def __new__(self, iterable, window):
        self.d = collections.deque(iterable, window)
        initial_value = sum(iterable) / len(iterable)
        return float.__new__(self, initial_value)
    def add(self, new_value):
        self.d.append(new_value)
        return MovingAverage(self.d, self.d.maxlen)

This time, when I start with

>>> ma = MovingAverage([10], 3)
>>> ma
10.0

and

>>> ma.add(3)
6.5
>>> ma
10.0
>>> ma = ma.add(3)
>>> ma
5.333333333333333

However, I think (I haven't tested to find out) it makes this significantly slower. So, can it be done? Can I somehow set it so that the return from ma is the value that I'm looking for? Or do I need to define a value attribute, change the base class to object, and abandon my pretense that I have a chance of controlling the return value of the class?

---

No. Since these types are immutable, you should be using encapsulation, not inheritance.

---

I don't think that Ignacios answer above needs any improvements, but since I had this class laying around I just thought I should share. It avoids doing big sum operations many times and also avoids rounding errors that might occur if you use a (more) naïve algorithm:

class MovingAverage:
    def __init__(self):
        self.sum = None
        self.num = 0

    def add(self, val):
        if self.sum is None:
            self.sum = val
        else:
            self.sum += val
        self.num += 1
        return self.val

    @property
    def val(self):
        if self.sum is None:
            return None
        else:
            return self.sum/self.num

if __name__ == "__main__":
    print("Running small test suite")
    fail = False

    m = MovingAverage()
    try:
        assert m.val is None, "A new MovingAverage should be None"
        assert m.add(10) == 10, "New average should be returned"
        assert m.val == 10, "The average should update for each add"
        assert m.add(20) == 15, "New average should be returned"
        assert m.val == 15, "The average should update for each add"
        assert m.add(0) == 10, "Adding zero should work"
        assert m.add(-10) == 5, "Adding a negative number should work"
        assert m.add(-1) == 19/5, "Result should be able to be a fraction"
    except AssertionError as e:
        print("Fail: %s" % e.args[0])
        fail = True

    if not fail: print("Pass")