19
\$\begingroup\$

I want to gain experience creating data structures that look and feel like Python builtin types. As a first exercise, I've written a WraparoundList class meant to be identical to the builtin list, except that accessing out-of-bounds elements "wraps around".

Goals:

  • The only behavior that differs from that of a list is when explicitly indexed with [].
  • Should look and feel like the Python builtin, i.e., wouldn't look too out of place in the collections module.
  • Should be compatible with both Python 2.7.x and 3.x (though I only have tested on 2.7.13).

The complete source code with doctests follows:

#!/usr/bin/env python

from sys import maxint as MAXINT

class WraparoundList(list):
    """A list whose index wraps around when out of bounds.

    A `WraparoundList` is the same as an ordinary `list`, 
    except that out-of-bounds indexing causes the index 
    value to wrap around. The wrapping behavior is as if
    after reaching the last element, one returned to the 
    other end of the list and continued counting.

    >>> x = WraparoundList('abcd')
    >>> x
    ['a', 'b', 'c', 'd']
    >>> x[3]
    'd'
    >>> x[4] # wraps to x[0]
    'a'
    >>> x[-6] = 'Q' # wraps to x[-2]
    >>> x
    ['a', 'b', 'Q', 'd']
    >>> del x[7] # wraps to x[3]
    >>> x 
    ['a', 'b', 'Q']

    Indices used in out-of-range slices also wrap around.
    If the slice's `start` or `stop` is out-of-bounds, it 
    gets wrapped around.

    >>> x = WraparoundList('abcd')
    >>> x
    ['a', 'b', 'c', 'd']
    >>> x[:10] # wraps to x[:2]
    ['a', 'b']
    >>> x[-7:3] # wraps to x[-3:3]
    ['b', 'c']

    The one way in which slicing a `WraparoundList` differs 
    from slicing an ordinary `list` is the case of using the
    list length as the upper limit.

    >>> x = WraparoundList('abcd')
    >>> x
    ['a', 'b', 'c', 'd']
    >>> x[2:]
    ['c', 'd']
    >>> x[2:4] # wraps to x[2:0]
    []

    Initializing a `WraparoundList` with a nested iterable
    does not cause inner indices to wrap. To have a multi-
    dimensional `WraparoundList`, all the elements of the 
    outer `WraparoundList` must also be `WraparoundList`s.

    >>> x = WraparoundList([list('abc'), list('def')])
    >>> x
    [['a', 'b', 'c'], ['d', 'e', 'f']]
    >>> x[3]
    ['d', 'e', 'f']
    >>> x[3][5]
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    IndexError: list index out of range
    >>> y = WraparoundList([WraparoundList(i) for i in x])   
    >>> y[3][5]
    'f'
    """
    def __getitem__(self, i):
        """x.__getitem__(i) <=> x[i]"""
        if isinstance(i, slice):
            return list.__getitem__(self, self._wrap_slice(i))
        else:
            return list.__getitem__(self, self._wrap_index(i))

    def __getslice__(self, i, j):
        """x.__getslice__(i, j) <=> x[i:j]"""
        return self.__getitem__(slice(i, j, None))

    def __setitem__(self, i, y):
        """x.__setitem__(i, y) <=> x[i] = y"""
        if isinstance(i, slice):
            list.__setitem__(self, self._wrap_slice(i), y)
        else:
            list.__setitem__(self, self._wrap_index(i), y)

    def __setslice__(self, i, j):
        """x.__setslice__(i, j) <=> x[i:j] = y"""
        self.__setitem__(slice(i, j, None))

    def __delitem__(self, i):
        """x.__delitem__(i, y) <=> del x[i]"""
        if isinstance(i, slice):
            list.__delitem__(self, self._wrap_slice(i))
        else:
            list.__delitem__(self, self._wrap_index(i))

    def __delslice__(self, i, j):
        """x.__delslice__(i, j) <=> del x[i:j]"""
        self.__delitem__(slice(i, j, None))

    def _wrap_index(self, i):
        _len = len(self)
        if i >= _len:
            return i % _len
        elif i < -_len:
            return i % (-_len)
        else:
            return i

    def _wrap_slice(self, slc):
        if slc.start is None:
            start = None
        else:
            start = self._wrap_index(slc.start) 
        if slc.stop is None:
            stop = None
        elif slc.stop == MAXINT:
            # __*slice__ methods treat absent upper bounds as sys.maxint, which would
            # wrap around to a system-dependent (and probably unexpected) value. Setting 
            # to `None` in this case forces the slice to run to the end of the list.
            stop = None
        else:
            stop = self._wrap_index(slc.stop)
        step = slc.step
        return slice(start, stop, step)

def main():
    pass

if __name__ == '__main__':
    main()
\$\endgroup\$
  • 4
    \$\begingroup\$ I'd recommend calling it a circular list \$\endgroup\$ – theonlygusti Jan 29 '17 at 18:05
  • 1
    \$\begingroup\$ @theonlygusti Considered that, but thought that it may be too suggestive of an infinite iterable a la itertools.cycle. \$\endgroup\$ – Endulum Jan 29 '17 at 21:15
  • \$\begingroup\$ Interestingly, itertools.cycle does not appear to have real Python source code, apart from being plugged into the Python object API. The source is here in CPython: github.com/python/cpython/blob/master/Modules/… \$\endgroup\$ – Vasili Syrakis Jan 30 '17 at 14:07
  • \$\begingroup\$ In case you didn't know, negative indices already wrap around in regular Python lists and other sequence types. \$\endgroup\$ – mkrieger1 Jan 30 '17 at 22:57
  • \$\begingroup\$ @mkrieger1 Not sure what you mean. x=range(4); x[-10] raises IndexError, at least in 2.7.13. Do you mean "wrap" in a different sense? \$\endgroup\$ – Endulum Jan 31 '17 at 4:34
14
\$\begingroup\$
  1. This is well documented, well commented code.

  2. The docstring says:

    The one way in which slicing a WraparoundList differs from slicing an ordinary list is the case of using the list length as the upper limit.

    but this isn't quite the whole story — an ordinary list can also be sliced using a value greater than the list length, and in that case WraparoundList also has a different behaviour:

    >>> x = [1, 2, 3]
    >>> x[:10]
    [1, 2, 3]
    >>> x = WraparoundList(x)
    >>> x[:10]
    [1]
    
  3. The code is not portable to Python 3, because there's no sys.maxint (all integers in Python 3 are "long"). I suggest something like this:

    try:
        # In Python 2.7, when __*slice__ methods are called with no "stop"
        # value, sys.maxint is passed instead.
        from sys import maxint as NO_STOP
    except ImportError:
        # Python 3 does not have sys.maxint or use the __*slice__ methods.
        NO_STOP = object()
    

    I prefer a name like NO_STOP because it communicates the intention rather than the implementation.

  4. _wrap_index raises ZeroDivisionError if the list is empty:

    >>> w = WraparoundList([])
    >>> w[0]
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
      File "cr153920.py", line 79, in __getitem__
        return list.__getitem__(self, self._wrap_index(i))
      File "cr153920.py", line 110, in _wrap_index
        return i % _len
    ZeroDivisionError: integer division or modulo by zero
    

    Raising an exception is the right thing to do in this case, but I would expect to get an IndexError instead.

  5. The code calls list.__getitem__ directly, rather than via the super function. But this has the unsatisfactory consequence that if someone has another class C also inheriting from list and overriding the __getitem__ method, and combines WraparoundList and C via inheritance, like this:

    class D(WraparoundList, C):
        pass
    

    Then D()[0] calls WraparoundList.__getitem__, which calls list.__getitem__, but C.__getitem__ is never called, contrary to what one would expect. If you want to support subclassing of WraparoundList, then you need to write:

    return super(WraparoundList, self).__getitem__(self._wrap_slice(i))
    

    and so on.

  6. With a little refactoring, you could avoid some of the repetition. In particular, if you had a method like this:

    def _wrap_arg(self, i):
        if isinstance(i, slice):
            return self._wrap_slice(i)
        else:
            return self._wrap_index(i)
    

    Then you'd be able to write:

    def __getitem__(self, i):
        """x.__getitem__(i) <=> x[i]"""
        return super(WraparoundList, self).__getitem__(self._wrap_arg(i))
    

    and so on.

  7. Once you've done the refactoring above, you'll see that _wrap_slice is only called from one place, so it could be inlined at its point of use.

  8. There is no need to include an empty main function or an if __name__ == '__main__': section — if there's nothing to do, then there's no need to write code to do it.

\$\endgroup\$
  • \$\begingroup\$ Wonderful! One clarification: Do I understand correctly that the __*slice__ methods are redundant if I'm handling slices in __*item__, even in 2.7.X? \$\endgroup\$ – Endulum Jan 29 '17 at 21:08
  • \$\begingroup\$ See the documentation: "built-in types in CPython currently still implement __getslice__(). Therefore, you have to override it in derived classes when implementing slicing." \$\endgroup\$ – Gareth Rees Jan 29 '17 at 21:16
0
\$\begingroup\$

On top of Gareth Ree's excellent answer: because the WraparoundList and the usual list behave differently (for instance when sliced using a value greater than the list length), your class does not respect Liskov substitution principle (see also less formal explanation).

In a nutshell: because some code using a list would behave differently if it was to use a WraparoundList, then WraparoundList should not inherit from list even if the "WraparoundList is a list" relationship is respected.

A way to change this could be to stop inheriting from list but instead to use a list internally to store data (Composition over inheritance).

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.