Transparent wrapper class for data structure in Python

Question

I needed a deque who's maxlen can be set again after it has been initialized. So, since new-style classes in Python are treated as types, I decided to create a new class borrowing from deque to accomplish this. My approach is to have an internal deque as an attribute and when maxlen is set, it replaces the internal deque with a new one, initialized to the old contents and the new maxlen.

I tried subclassing deque but it's too crude, creating a deque that's useless on top with the useful one internally. So I chose to simply have all of the wrapper class's attributes (except special ones) point to the internal deque's. However, due to how new-style classes work, the special attributes have to be handled manually. Since all I want is to override maxlen's setter, all of this seems inelegant and I'm wondering if there's a cleaner way of accomplishing this.

From what I've read here, it seems I could subclass this class in __new__ to skip overriding the special attributes, but that seems even more hairy than what I already wrote.

This is the result, stripped of extraneous comments (complete code here if you want something runnable, with tests):

# -*- coding: utf-8 -*-

from __future__ import print_function

from collections import deque


class ResizableDeque(object):
    def __init__(self, *args, **kwargs):
        self.internal = deque(*args, **kwargs)

        skip_list = [
            'maxlen'
        ] + [attr for attr in dir(deque) if attr.startswith('__') and
             attr.endswith('__')]

        for attr in dir(deque):
            if attr not in skip_list:
                setattr(self, attr, getattr(self.internal, attr))

    @property
    def maxlen(self):
        return self.internal.maxlen

    @maxlen.setter
    def maxlen(self, value):
        templist = list(self.internal)
        self.internal = deque(templist, value)

    def __str__(self):
        return self.internal.__str__()

    def __repr__(self):
        return self.internal.__repr__()

    def __getitem__(self, value):
        return self.internal.__getitem__(value)

    def __setitem__(self, index, value):
        return self.internal.__setitem__(index, value)

    # these have not been tested
    def __copy__(self):
        return self.internal.__copy__()

    def __delitem__(self, index):
        return self.internal.__delitem__(index)

    def __iadd__(self, other):
        return self.internal.__iadd__(other)

    def __len__(self):
        return self.internal.__len__()

    # not sure if overriding __sizeof__ is wise this way
    def __sizeof__(self):
        return self.__sizeof__() + self.internal.__sizeof__()

    # pretty sure this is ok
    def __format__(self, spec):
        return self.internal.__format__(spec)

Answer 1

Correctness?

I copied your code, and added the following testcase.

if __name__ == "__main__":
    k = ResizableDeque(maxlen=10)
    k.append(12)
    print(k)
    k.maxlen = 10
    k.append(13)
    k += [14]
    print(k)

Expected result:

deque([12], maxlen=10)
deque([12, 13, 14], maxlen=10)

But I got

deque([12], maxlen=10)
deque([12, 14], maxlen=10)

The reason is that you copy over a bunch of items in __init__, but not in the maxlen.setter.

Whitelisting vs blacklisting

In __init__ you copy several items over using a skip_list. Why not explicitly define the items you want to copy over instead?

Actually, why copy at all? Performance would be a proper reason, but then I'd suggest copying just the items that have proven to be necessary to copy (after profiling).