odict.py

# odict.py
# An Ordered Dictionary object
# Copyright (C) 2005 Nicola Larosa, Michael Foord
# E-mail: nico AT tekNico DOT net, fuzzyman AT voidspace DOT org DOT uk

# This software is licensed under the terms of the BSD license.
# http://www.voidspace.org.uk/python/license.shtml
# Basically you're free to copy, modify, distribute and relicense it,
# So long as you keep a copy of the license with it.

# Documentation at http://www.voidspace.org.uk/python/odict.html
# For information about bugfixes, updates and support, please join the
# Pythonutils mailing list:
# http://groups.google.com/group/pythonutils/
# Comments, suggestions and bug reports welcome.

"""A dict that keeps keys in insertion order"""
from __future__ import generators

__author__ = ('Nicola Larosa <nico-NoSp@m-tekNico.net>,'
    'Michael Foord <fuzzyman AT voidspace DOT org DOT uk>')

__docformat__ = "restructuredtext en"

__revision__ = '$Id: odict.py 129 2005-09-12 18:15:28Z teknico $'

__version__ = '0.2.2'

__all__ = ['OrderedDict', 'SequenceOrderedDict']

import sys
INTP_VER = sys.version_info[:2]
if INTP_VER < (2, 2):
    raise RuntimeError("Python v.2.2 or later required")

import types, warnings

class OrderedDict(dict):
    """
    A class of dictionary that keeps the insertion order of keys.
    
    All appropriate methods return keys, items, or values in an ordered way.
    
    All normal dictionary methods are available. Update and comparison is
    restricted to other OrderedDict objects.
    
    Various sequence methods are available, including the ability to explicitly
    mutate the key ordering.
    
    __contains__ tests:
    
    >>> d = OrderedDict(((1, 3),))
    >>> 1 in d
    1
    >>> 4 in d
    0
    
    __getitem__ tests:
    
    >>> OrderedDict(((1, 3), (3, 2), (2, 1)))[2]
    1
    >>> OrderedDict(((1, 3), (3, 2), (2, 1)))[4]
    Traceback (most recent call last):
    KeyError: 4
    
    __len__ tests:
    
    >>> len(OrderedDict())
    0
    >>> len(OrderedDict(((1, 3), (3, 2), (2, 1))))
    3
    
    get tests:
    
    >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
    >>> d.get(1)
    3
    >>> d.get(4) is None
    1
    >>> d.get(4, 5)
    5
    >>> d
    OrderedDict([(1, 3), (3, 2), (2, 1)])
    
    has_key tests:
    
    >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
    >>> d.has_key(1)
    1
    >>> d.has_key(4)
    0
    """

    def __init__(self, init_val=(), strict=False):
        """
        Create a new ordered dictionary. Cannot init from a normal dict,
        nor from kwargs, since items order is undefined in those cases.
        
        If the ``strict`` keyword argument is ``True`` (``False`` is the
        default) then when doing slice assignment - the ``OrderedDict`` you are
        assigning from *must not* contain any keys in the remaining dict.
        
        >>> OrderedDict()
        OrderedDict([])
        >>> OrderedDict({1: 1})
        Traceback (most recent call last):
        TypeError: undefined order, cannot get items from dict
        >>> OrderedDict({1: 1}.items())
        OrderedDict([(1, 1)])
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d
        OrderedDict([(1, 3), (3, 2), (2, 1)])
        >>> OrderedDict(d)
        OrderedDict([(1, 3), (3, 2), (2, 1)])
        """
        self.strict = strict
        dict.__init__(self)
        if isinstance(init_val, OrderedDict):
            self._sequence = init_val.keys()
            dict.update(self, init_val)
        elif isinstance(init_val, dict):
            # we lose compatibility with other ordered dict types this way
            raise TypeError('undefined order, cannot get items from dict')
        else:
            self._sequence = []
            self.update(init_val)

### Special methods ###

    def __delitem__(self, key):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> del d[3]
        >>> d
        OrderedDict([(1, 3), (2, 1)])
        >>> del d[3]
        Traceback (most recent call last):
        KeyError: 3
        >>> d[3] = 2
        >>> d
        OrderedDict([(1, 3), (2, 1), (3, 2)])
        >>> del d[0:1]
        >>> d
        OrderedDict([(2, 1), (3, 2)])
        """
        if isinstance(key, types.SliceType):
            # FIXME: efficiency?
            keys = self._sequence[key]
            for entry in keys:
                dict.__delitem__(self, entry)
            del self._sequence[key]
        else:
            # do the dict.__delitem__ *first* as it raises
            # the more appropriate error
            dict.__delitem__(self, key)
            self._sequence.remove(key)

    def __eq__(self, other):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d == OrderedDict(d)
        True
        >>> d == OrderedDict(((1, 3), (2, 1), (3, 2)))
        False
        >>> d == OrderedDict(((1, 0), (3, 2), (2, 1)))
        False
        >>> d == OrderedDict(((0, 3), (3, 2), (2, 1)))
        False
        >>> d == dict(d)
        False
        >>> d == False
        False
        """
        if isinstance(other, OrderedDict):
            # FIXME: efficiency?
            #   Generate both item lists for each compare
            return (self.items() == other.items())
        else:
            return False

    def __lt__(self, other):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> c = OrderedDict(((0, 3), (3, 2), (2, 1)))
        >>> c < d
        True
        >>> d < c
        False
        >>> d < dict(c)
        Traceback (most recent call last):
        TypeError: Can only compare with other OrderedDicts
        """
        if not isinstance(other, OrderedDict):
            raise TypeError('Can only compare with other OrderedDicts')
        # FIXME: efficiency?
        #   Generate both item lists for each compare
        return (self.items() < other.items())

    def __le__(self, other):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> c = OrderedDict(((0, 3), (3, 2), (2, 1)))
        >>> e = OrderedDict(d)
        >>> c <= d
        True
        >>> d <= c
        False
        >>> d <= dict(c)
        Traceback (most recent call last):
        TypeError: Can only compare with other OrderedDicts
        >>> d <= e
        True
        """
        if not isinstance(other, OrderedDict):
            raise TypeError('Can only compare with other OrderedDicts')
        # FIXME: efficiency?
        #   Generate both item lists for each compare
        return (self.items() <= other.items())

    def __ne__(self, other):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d != OrderedDict(d)
        False
        >>> d != OrderedDict(((1, 3), (2, 1), (3, 2)))
        True
        >>> d != OrderedDict(((1, 0), (3, 2), (2, 1)))
        True
        >>> d == OrderedDict(((0, 3), (3, 2), (2, 1)))
        False
        >>> d != dict(d)
        True
        >>> d != False
        True
        """
        if isinstance(other, OrderedDict):
            # FIXME: efficiency?
            #   Generate both item lists for each compare
            return not (self.items() == other.items())
        else:
            return True

    def __gt__(self, other):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> c = OrderedDict(((0, 3), (3, 2), (2, 1)))
        >>> d > c
        True
        >>> c > d
        False
        >>> d > dict(c)
        Traceback (most recent call last):
        TypeError: Can only compare with other OrderedDicts
        """
        if not isinstance(other, OrderedDict):
            raise TypeError('Can only compare with other OrderedDicts')
        # FIXME: efficiency?
        #   Generate both item lists for each compare
        return (self.items() > other.items())

    def __ge__(self, other):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> c = OrderedDict(((0, 3), (3, 2), (2, 1)))
        >>> e = OrderedDict(d)
        >>> c >= d
        False
        >>> d >= c
        True
        >>> d >= dict(c)
        Traceback (most recent call last):
        TypeError: Can only compare with other OrderedDicts
        >>> e >= d
        True
        """
        if not isinstance(other, OrderedDict):
            raise TypeError('Can only compare with other OrderedDicts')
        # FIXME: efficiency?
        #   Generate both item lists for each compare
        return (self.items() >= other.items())

    def __repr__(self):
        """
        Used for __repr__ and __str__
        
        >>> r1 = repr(OrderedDict((('a', 'b'), ('c', 'd'), ('e', 'f'))))
        >>> r1
        "OrderedDict([('a', 'b'), ('c', 'd'), ('e', 'f')])"
        >>> r2 = repr(OrderedDict((('a', 'b'), ('e', 'f'), ('c', 'd'))))
        >>> r2
        "OrderedDict([('a', 'b'), ('e', 'f'), ('c', 'd')])"
        >>> r1 == str(OrderedDict((('a', 'b'), ('c', 'd'), ('e', 'f'))))
        True
        >>> r2 == str(OrderedDict((('a', 'b'), ('e', 'f'), ('c', 'd'))))
        True
        """
        return '%s([%s])' % (self.__class__.__name__, ', '.join(
            ['(%r, %r)' % (key, self[key]) for key in self._sequence]))

    def __setitem__(self, key, val):
        """
        Allows slice assignment, so long as the slice is an OrderedDict
        >>> d = OrderedDict()
        >>> d['a'] = 'b'
        >>> d['b'] = 'a'
        >>> d[3] = 12
        >>> d
        OrderedDict([('a', 'b'), ('b', 'a'), (3, 12)])
        >>> d[:] = OrderedDict(((1, 2), (2, 3), (3, 4)))
        >>> d
        OrderedDict([(1, 2), (2, 3), (3, 4)])
        >>> d[::2] = OrderedDict(((7, 8), (9, 10)))
        >>> d
        OrderedDict([(7, 8), (2, 3), (9, 10)])
        >>> d = OrderedDict(((0, 1), (1, 2), (2, 3), (3, 4)))
        >>> d[1:3] = OrderedDict(((1, 2), (5, 6), (7, 8)))
        >>> d
        OrderedDict([(0, 1), (1, 2), (5, 6), (7, 8), (3, 4)])
        >>> d = OrderedDict(((0, 1), (1, 2), (2, 3), (3, 4)), strict=True)
        >>> d[1:3] = OrderedDict(((1, 2), (5, 6), (7, 8)))
        >>> d
        OrderedDict([(0, 1), (1, 2), (5, 6), (7, 8), (3, 4)])
        
        >>> a = OrderedDict(((0, 1), (1, 2), (2, 3)), strict=True)
        >>> a[3] = 4
        >>> a
        OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a[::1] = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a
        OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a[:2] = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4), (4, 5)])
        Traceback (most recent call last):
        ValueError: slice assignment must be from unique keys
        >>> a = OrderedDict(((0, 1), (1, 2), (2, 3)))
        >>> a[3] = 4
        >>> a
        OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a[::1] = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a
        OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a[:2] = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a
        OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a[::-1] = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> a
        OrderedDict([(3, 4), (2, 3), (1, 2), (0, 1)])
        
        >>> d = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> d[:1] = 3
        Traceback (most recent call last):
        TypeError: slice assignment requires an OrderedDict
        
        >>> d = OrderedDict([(0, 1), (1, 2), (2, 3), (3, 4)])
        >>> d[:1] = OrderedDict([(9, 8)])
        >>> d
        OrderedDict([(9, 8), (1, 2), (2, 3), (3, 4)])
        """
        if isinstance(key, types.SliceType):
            if not isinstance(val, OrderedDict):
                # FIXME: allow a list of tuples?
                raise TypeError('slice assignment requires an OrderedDict')
            keys = self._sequence[key]
            # NOTE: Could use ``range(*key.indices(len(self._sequence)))``
            indexes = range(len(self._sequence))[key]
            if key.step is None:
                # NOTE: new slice may not be the same size as the one being
                #   overwritten !
                # NOTE: What is the algorithm for an impossible slice?
                #   e.g. d[5:3]
                pos = key.start or 0
                del self[key]
                newkeys = val.keys()
                for k in newkeys:
                    if k in self:
                        if self.strict:
                            raise ValueError('slice assignment must be from '
                                'unique keys')
                        else:
                            # NOTE: This removes duplicate keys *first*
                            #   so start position might have changed?
                            del self[k]
                self._sequence = (self._sequence[:pos] + newkeys +
                    self._sequence[pos:])
                dict.update(self, val)
            else:
                # extended slice - length of new slice must be the same
                # as the one being replaced
                if len(keys) != len(val):
                    raise ValueError('attempt to assign sequence of size %s '
                        'to extended slice of size %s' % (len(val), len(keys)))
                # FIXME: efficiency?
                del self[key]
                item_list = zip(indexes, val.items())
                # smallest indexes first - higher indexes not guaranteed to
                # exist
                item_list.sort()
                for pos, (newkey, newval) in item_list:
                    if self.strict and newkey in self:
                        raise ValueError('slice assignment must be from unique'
                            ' keys')
                    self.insert(pos, newkey, newval)
        else:
            if key not in self:
                self._sequence.append(key)
            dict.__setitem__(self, key, val)

    def __getitem__(self, key):
        """
        Allows slicing. Returns an OrderedDict if you slice.
        >>> b = OrderedDict([(7, 0), (6, 1), (5, 2), (4, 3), (3, 4), (2, 5), (1, 6)])
        >>> b[::-1]
        OrderedDict([(1, 6), (2, 5), (3, 4), (4, 3), (5, 2), (6, 1), (7, 0)])
        >>> b[2:5]
        OrderedDict([(5, 2), (4, 3), (3, 4)])
        >>> type(b[2:4])
        <class '__main__.OrderedDict'>
        """
        if isinstance(key, types.SliceType):
            # FIXME: does this raise the error we want?
            keys = self._sequence[key]
            # FIXME: efficiency?
            return OrderedDict([(entry, self[entry]) for entry in keys])
        else:
            return dict.__getitem__(self, key)

    __str__ = __repr__

    def __setattr__(self, name, value):
        """
        Implemented so that accesses to ``sequence`` raise a warning and are
        diverted to the new ``setkeys`` method.
        """
        if name == 'sequence':
            warnings.warn('Use of the sequence attribute is deprecated.'
                ' Use the keys method instead.', DeprecationWarning)
            # NOTE: doesn't return anything
            self.setkeys(value)
        else:
            # FIXME: do we want to allow arbitrary setting of attributes?
            #   Or do we want to manage it?
            object.__setattr__(self, name, value)

    def __getattr__(self, name):
        """
        Implemented so that access to ``sequence`` raises a warning.
        
        >>> d = OrderedDict()
        >>> d.sequence
        []
        """
        if name == 'sequence':
            warnings.warn('Use of the sequence attribute is deprecated.'
                ' Use the keys method instead.', DeprecationWarning)
            # NOTE: Still (currently) returns a direct reference. Need to
            #   because code that uses sequence will expect to be able to
            #   mutate it in place.
            return self._sequence
        else:
            # raise the appropriate error
            raise AttributeError("OrderedDict has no '%s' attribute" % name)

    def __deepcopy__(self, memo):
        """
        To allow deepcopy to work with OrderedDict.
        
        >>> from copy import deepcopy
        >>> a = OrderedDict([(1, 1), (2, 2), (3, 3)])
        >>> a['test'] = {}
        >>> b = deepcopy(a)
        >>> b == a
        True
        >>> b is a
        False
        >>> a['test'] is b['test']
        False
        """
        from copy import deepcopy
        return self.__class__(deepcopy(self.items(), memo), self.strict)


### Read-only methods ###

    def copy(self):
        """
        >>> OrderedDict(((1, 3), (3, 2), (2, 1))).copy()
        OrderedDict([(1, 3), (3, 2), (2, 1)])
        """
        return OrderedDict(self)

    def items(self):
        """
        ``items`` returns a list of tuples representing all the 
        ``(key, value)`` pairs in the dictionary.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.items()
        [(1, 3), (3, 2), (2, 1)]
        >>> d.clear()
        >>> d.items()
        []
        """
        return zip(self._sequence, self.values())

    def keys(self):
        """
        Return a list of keys in the ``OrderedDict``.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.keys()
        [1, 3, 2]
        """
        return self._sequence[:]

    def values(self, values=None):
        """
        Return a list of all the values in the OrderedDict.
        
        Optionally you can pass in a list of values, which will replace the
        current list. The value list must be the same len as the OrderedDict.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.values()
        [3, 2, 1]
        """
        return [self[key] for key in self._sequence]

    def iteritems(self):
        """
        >>> ii = OrderedDict(((1, 3), (3, 2), (2, 1))).iteritems()
        >>> ii.next()
        (1, 3)
        >>> ii.next()
        (3, 2)
        >>> ii.next()
        (2, 1)
        >>> ii.next()
        Traceback (most recent call last):
        StopIteration
        """
        def make_iter(self=self):
            keys = self.iterkeys()
            while True:
                key = keys.next()
                yield (key, self[key])
        return make_iter()

    def iterkeys(self):
        """
        >>> ii = OrderedDict(((1, 3), (3, 2), (2, 1))).iterkeys()
        >>> ii.next()
        1
        >>> ii.next()
        3
        >>> ii.next()
        2
        >>> ii.next()
        Traceback (most recent call last):
        StopIteration
        """
        return iter(self._sequence)

    __iter__ = iterkeys

    def itervalues(self):
        """
        >>> iv = OrderedDict(((1, 3), (3, 2), (2, 1))).itervalues()
        >>> iv.next()
        3
        >>> iv.next()
        2
        >>> iv.next()
        1
        >>> iv.next()
        Traceback (most recent call last):
        StopIteration
        """
        def make_iter(self=self):
            keys = self.iterkeys()
            while True:
                yield self[keys.next()]
        return make_iter()

### Read-write methods ###

    def clear(self):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.clear()
        >>> d
        OrderedDict([])
        """
        dict.clear(self)
        self._sequence = []

    def pop(self, key, *args):
        """
        No dict.pop in Python 2.2, gotta reimplement it
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.pop(3)
        2
        >>> d
        OrderedDict([(1, 3), (2, 1)])
        >>> d.pop(4)
        Traceback (most recent call last):
        KeyError: 4
        >>> d.pop(4, 0)
        0
        >>> d.pop(4, 0, 1)
        Traceback (most recent call last):
        TypeError: pop expected at most 2 arguments, got 3
        """
        if len(args) > 1:
            raise TypeError, ('pop expected at most 2 arguments, got %s' %
                (len(args) + 1))
        if key in self:
            val = self[key]
            del self[key]
        else:
            try:
                val = args[0]
            except IndexError:
                raise KeyError(key)
        return val

    def popitem(self, i=-1):
        """
        Delete and return an item specified by index, not a random one as in
        dict. The index is -1 by default (the last item).
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.popitem()
        (2, 1)
        >>> d
        OrderedDict([(1, 3), (3, 2)])
        >>> d.popitem(0)
        (1, 3)
        >>> OrderedDict().popitem()
        Traceback (most recent call last):
        KeyError: 'popitem(): dictionary is empty'
        >>> d.popitem(2)
        Traceback (most recent call last):
        IndexError: popitem(): index 2 not valid
        """
        if not self._sequence:
            raise KeyError('popitem(): dictionary is empty')
        try:
            key = self._sequence[i]
        except IndexError:
            raise IndexError('popitem(): index %s not valid' % i)
        return (key, self.pop(key))

    def setdefault(self, key, defval = None):
        """
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.setdefault(1)
        3
        >>> d.setdefault(4) is None
        True
        >>> d
        OrderedDict([(1, 3), (3, 2), (2, 1), (4, None)])
        >>> d.setdefault(5, 0)
        0
        >>> d
        OrderedDict([(1, 3), (3, 2), (2, 1), (4, None), (5, 0)])
        """
        if key in self:
            return self[key]
        else:
            self[key] = defval
            return defval

    def update(self, from_od):
        """
        Update from another OrderedDict or sequence of (key, value) pairs
        
        >>> d = OrderedDict(((1, 0), (0, 1)))
        >>> d.update(OrderedDict(((1, 3), (3, 2), (2, 1))))
        >>> d
        OrderedDict([(1, 3), (0, 1), (3, 2), (2, 1)])
        >>> d.update({4: 4})
        Traceback (most recent call last):
        TypeError: undefined order, cannot get items from dict
        >>> d.update((4, 4))
        Traceback (most recent call last):
        TypeError: cannot convert dictionary update sequence element "4" to a 2-item sequence
        """
        if isinstance(from_od, OrderedDict):
            for key, val in from_od.items():
                self[key] = val
        elif isinstance(from_od, dict):
            # we lose compatibility with other ordered dict types this way
            raise TypeError('undefined order, cannot get items from dict')
        else:
            # FIXME: efficiency?
            # sequence of 2-item sequences, or error
            for item in from_od:
                try:
                    key, val = item
                except TypeError:
                    raise TypeError('cannot convert dictionary update'
                        ' sequence element "%s" to a 2-item sequence' % item)
                self[key] = val

    def rename(self, old_key, new_key):
        """
        Rename the key for a given value, without modifying sequence order.
        
        For the case where new_key already exists this raise an exception,
        since if new_key exists, it is ambiguous as to what happens to the
        associated values, and the position of new_key in the sequence.
        
        >>> od = OrderedDict()
        >>> od['a'] = 1
        >>> od['b'] = 2
        >>> od.items()
        [('a', 1), ('b', 2)]
        >>> od.rename('b', 'c')
        >>> od.items()
        [('a', 1), ('c', 2)]
        >>> od.rename('c', 'a')
        Traceback (most recent call last):
        ValueError: New key already exists: 'a'
        >>> od.rename('d', 'b')
        Traceback (most recent call last):
        KeyError: 'd'
        """
        if new_key == old_key:
            # no-op
            return
        if new_key in self:
            raise ValueError("New key already exists: %r" % new_key)
        # rename sequence entry
        value = self[old_key] 
        old_idx = self._sequence.index(old_key)
        self._sequence[old_idx] = new_key
        # rename internal dict entry
        dict.__delitem__(self, old_key)
        dict.__setitem__(self, new_key, value)

    def setitems(self, items):
        """
        This method allows you to set the items in the dict.
        
        It takes a list of tuples - of the same sort returned by the ``items``
        method.
        
        >>> d = OrderedDict()
        >>> d.setitems(((3, 1), (2, 3), (1, 2)))
        >>> d
        OrderedDict([(3, 1), (2, 3), (1, 2)])
        """
        self.clear()
        # FIXME: this allows you to pass in an OrderedDict as well :-)
        self.update(items)

    def setkeys(self, keys):
        """
        ``setkeys`` all ows you to pass in a new list of keys which will
        replace the current set. This must contain the same set of keys, but
        need not be in the same order.
        
        If you pass in new keys that don't match, a ``KeyError`` will be
        raised.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.keys()
        [1, 3, 2]
        >>> d.setkeys((1, 2, 3))
        >>> d
        OrderedDict([(1, 3), (2, 1), (3, 2)])
        >>> d.setkeys(['a', 'b', 'c'])
        Traceback (most recent call last):
        KeyError: 'Keylist is not the same as current keylist.'
        """
        # FIXME: Efficiency? (use set for Python 2.4 :-)
        # NOTE: list(keys) rather than keys[:] because keys[:] returns
        #   a tuple, if keys is a tuple.
        kcopy = list(keys)
        kcopy.sort()
        self._sequence.sort()
        if kcopy != self._sequence:
            raise KeyError('Keylist is not the same as current keylist.')
        # NOTE: This makes the _sequence attribute a new object, instead
        #       of changing it in place.
        # FIXME: efficiency?
        self._sequence = list(keys)

    def setvalues(self, values):
        """
        You can pass in a list of values, which will replace the
        current list. The value list must be the same len as the OrderedDict.
        
        (Or a ``ValueError`` is raised.)
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.setvalues((1, 2, 3))
        >>> d
        OrderedDict([(1, 1), (3, 2), (2, 3)])
        >>> d.setvalues([6])
        Traceback (most recent call last):
        ValueError: Value list is not the same length as the OrderedDict.
        """
        if len(values) != len(self):
            # FIXME: correct error to raise?
            raise ValueError('Value list is not the same length as the '
                'OrderedDict.')
        self.update(zip(self, values))

### Sequence Methods ###

    def index(self, key):
        """
        Return the position of the specified key in the OrderedDict.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.index(3)
        1
        >>> d.index(4)
        Traceback (most recent call last):
        ValueError: list.index(x): x not in list
        """
        return self._sequence.index(key)

    def insert(self, index, key, value):
        """
        Takes ``index``, ``key``, and ``value`` as arguments.
        
        Sets ``key`` to ``value``, so that ``key`` is at position ``index`` in
        the OrderedDict.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.insert(0, 4, 0)
        >>> d
        OrderedDict([(4, 0), (1, 3), (3, 2), (2, 1)])
        >>> d.insert(0, 2, 1)
        >>> d
        OrderedDict([(2, 1), (4, 0), (1, 3), (3, 2)])
        >>> d.insert(8, 8, 1)
        >>> d
        OrderedDict([(2, 1), (4, 0), (1, 3), (3, 2), (8, 1)])
        """
        if key in self:
            # FIXME: efficiency?
            del self[key]
        self._sequence.insert(index, key)
        dict.__setitem__(self, key, value)

    def reverse(self):
        """
        Reverse the order of the OrderedDict.
        
        >>> d = OrderedDict(((1, 3), (3, 2), (2, 1)))
        >>> d.reverse()
        >>> d
        OrderedDict([(2, 1), (3, 2), (1, 3)])
        """
        self._sequence.reverse()

    def sort(self, *args, **kwargs):
        """
        Sort the key order in the OrderedDict.
        
        This method takes the same arguments as the ``list.sort`` method on
        your version of Python.
        
        >>> d = OrderedDict(((4, 1), (2, 2), (3, 3), (1, 4)))
        >>> d.sort()
        >>> d
        OrderedDict([(1, 4), (2, 2), (3, 3), (4, 1)])
        """
        self._sequence.sort(*args, **kwargs)

class Keys(object):
    # FIXME: should this object be a subclass of list?
    """
    Custom object for accessing the keys of an OrderedDict.
    
    Can be called like the normal ``OrderedDict.keys`` method, but also
    supports indexing and sequence methods.
    """

    def __init__(self, main):
        self._main = main

    def __call__(self):
        """Pretend to be the keys method."""
        return self._main._keys()

    def __getitem__(self, index):
        """Fetch the key at position i."""
        # NOTE: this automatically supports slicing :-)
        return self._main._sequence[index]

    def __setitem__(self, index, name):
        """
        You cannot assign to keys, but you can do slice assignment to re-order
        them.
        
        You can only do slice assignment if the new set of keys is a reordering
        of the original set.
        """
        if isinstance(index, types.SliceType):
            # FIXME: efficiency?
            # check length is the same
            indexes = range(len(self._main._sequence))[index]
            if len(indexes) != len(name):
                raise ValueError('attempt to assign sequence of size %s '
                    'to slice of size %s' % (len(name), len(indexes)))
            # check they are the same keys
            # FIXME: Use set
            old_keys = self._main._sequence[index]
            new_keys = list(name)
            old_keys.sort()
            new_keys.sort()
            if old_keys != new_keys:
                raise KeyError('Keylist is not the same as current keylist.')
            orig_vals = [self._main[k] for k in name]
            del self._main[index]
            vals = zip(indexes, name, orig_vals)
            vals.sort()
            for i, k, v in vals:
                if self._main.strict and k in self._main:
                    raise ValueError('slice assignment must be from '
                        'unique keys')
                self._main.insert(i, k, v)
        else:
            raise ValueError('Cannot assign to keys')

    ### following methods pinched from UserList and adapted ###
    def __repr__(self): return repr(self._main._sequence)

    # FIXME: do we need to check if we are comparing with another ``Keys``
    #   object? (like the __cast method of UserList)
    def __lt__(self, other): return self._main._sequence <  other
    def __le__(self, other): return self._main._sequence <= other
    def __eq__(self, other): return self._main._sequence == other
    def __ne__(self, other): return self._main._sequence != other
    def __gt__(self, other): return self._main._sequence >  other
    def __ge__(self, other): return self._main._sequence >= other
    # FIXME: do we need __cmp__ as well as rich comparisons?
    def __cmp__(self, other): return cmp(self._main._sequence, other)

    def __contains__(self, item): return item in self._main._sequence
    def __len__(self): return len(self._main._sequence)
    def __iter__(self): return self._main.iterkeys()
    def count(self, item): return self._main._sequence.count(item)
    def index(self, item, *args): return self._main._sequence.index(item, *args)
    def reverse(self): self._main._sequence.reverse()
    def sort(self, *args, **kwds): self._main._sequence.sort(*args, **kwds)
    def __mul__(self, n): return self._main._sequence*n
    __rmul__ = __mul__
    def __add__(self, other): return self._main._sequence + other
    def __radd__(self, other): return other + self._main._sequence

    ## following methods not implemented for keys ##
    def __delitem__(self, i): raise TypeError('Can\'t delete items from keys')
    def __iadd__(self, other): raise TypeError('Can\'t add in place to keys')
    def __imul__(self, n): raise TypeError('Can\'t multiply keys in place')
    def append(self, item): raise TypeError('Can\'t append items to keys')
    def insert(self, i, item): raise TypeError('Can\'t insert items into keys')
    def pop(self, i=-1): raise TypeError('Can\'t pop items from keys')
    def remove(self, item): raise TypeError('Can\'t remove items from keys')
    def extend(self, other): raise TypeError('Can\'t extend keys')

class Items(object):
    """
    Custom object for accessing the items of an OrderedDict.
    
    Can be called like the normal ``OrderedDict.items`` method, but also
    supports indexing and sequence methods.
    """

    def __init__(self, main):
        self._main = main

    def __call__(self):
        """Pretend to be the items method."""
        return self._main._items()

    def __getitem__(self, index):
        """Fetch the item at position i."""
        if isinstance(index, types.SliceType):
            # fetching a slice returns an OrderedDict
            return self._main[index].items()
        key = self._main._sequence[index]
        return (key, self._main[key])

    def __setitem__(self, index, item):
        """Set item at position i to item."""
        if isinstance(index, types.SliceType):
            # NOTE: item must be an iterable (list of tuples)
            self._main[index] = OrderedDict(item)
        else:
            # FIXME: Does this raise a sensible error?
            orig = self._main.keys[index]
            key, value = item
            if self._main.strict and key in self and (key != orig):
                raise ValueError('slice assignment must be from '
                        'unique keys')
            # delete the current one
            del self._main[self._main._sequence[index]]
            self._main.insert(index, key, value)

    def __delitem__(self, i):
        """Delete the item at position i."""
        key = self._main._sequence[i]
        if isinstance(i, types.SliceType):
            for k in key:
                # FIXME: efficiency?
                del self._main[k]
        else:
            del self._main[key]

    ### following methods pinched from UserList and adapted ###
    def __repr__(self): return repr(self._main.items())

    # FIXME: do we need to check if we are comparing with another ``Items``
    #   object? (like the __cast method of UserList)
    def __lt__(self, other): return self._main.items() <  other
    def __le__(self, other): return self._main.items() <= other
    def __eq__(self, other): return self._main.items() == other
    def __ne__(self, other): return self._main.items() != other
    def __gt__(self, other): return self._main.items() >  other
    def __ge__(self, other): return self._main.items() >= other
    def __cmp__(self, other): return cmp(self._main.items(), other)

    def __contains__(self, item): return item in self._main.items()
    def __len__(self): return len(self._main._sequence) # easier :-)
    def __iter__(self): return self._main.iteritems()
    def count(self, item): return self._main.items().count(item)
    def index(self, item, *args): return self._main.items().index(item, *args)
    def reverse(self): self._main.reverse()
    def sort(self, *args, **kwds): self._main.sort(*args, **kwds)
    def __mul__(self, n): return self._main.items()*n
    __rmul__ = __mul__
    def __add__(self, other): return self._main.items() + other
    def __radd__(self, other): return other + self._main.items()

    def append(self, item):
        """Add an item to the end."""
        # FIXME: this is only append if the key isn't already present
        key, value = item
        self._main[key] = value

    def insert(self, i, item):
        key, value = item
        self._main.insert(i, key, value)

    def pop(self, i=-1):
        key = self._main._sequence[i]
        return (key, self._main.pop(key))

    def remove(self, item):
        key, value = item
        try:
            assert value == self._main[key]
        except (KeyError, AssertionError):
            raise ValueError('ValueError: list.remove(x): x not in list')
        else:
            del self._main[key]

    def extend(self, other):
        # FIXME: is only a true extend if none of the keys already present
        for item in other:
            key, value = item
            self._main[key] = value

    def __iadd__(self, other):
        self.extend(other)

    ## following methods not implemented for items ##

    def __imul__(self, n): raise TypeError('Can\'t multiply items in place')

class Values(object):
    """
    Custom object for accessing the values of an OrderedDict.
    
    Can be called like the normal ``OrderedDict.values`` method, but also
    supports indexing and sequence methods.
    """

    def __init__(self, main):
        self._main = main

    def __call__(self):
        """Pretend to be the values method."""
        return self._main._values()

    def __getitem__(self, index):
        """Fetch the value at position i."""
        if isinstance(index, types.SliceType):
            return [self._main[key] for key in self._main._sequence[index]]
        else:
            return self._main[self._main._sequence[index]]

    def __setitem__(self, index, value):
        """
        Set the value at position i to value.
        
        You can only do slice assignment to values if you supply a sequence of
        equal length to the slice you are replacing.
        """
        if isinstance(index, types.SliceType):
            keys = self._main._sequence[index]
            if len(keys) != len(value):
                raise ValueError('attempt to assign sequence of size %s '
                    'to slice of size %s' % (len(name), len(keys)))
            # FIXME: efficiency?  Would be better to calculate the indexes
            #   directly from the slice object
            # NOTE: the new keys can collide with existing keys (or even
            #   contain duplicates) - these will overwrite
            for key, val in zip(keys, value):
                self._main[key] = val
        else:
            self._main[self._main._sequence[index]] = value

    ### following methods pinched from UserList and adapted ###
    def __repr__(self): return repr(self._main.values())

    # FIXME: do we need to check if we are comparing with another ``Values``
    #   object? (like the __cast method of UserList)
    def __lt__(self, other): return self._main.values() <  other
    def __le__(self, other): return self._main.values() <= other
    def __eq__(self, other): return self._main.values() == other
    def __ne__(self, other): return self._main.values() != other
    def __gt__(self, other): return self._main.values() >  other
    def __ge__(self, other): return self._main.values() >= other
    def __cmp__(self, other): return cmp(self._main.values(), other)

    def __contains__(self, item): return item in self._main.values()
    def __len__(self): return len(self._main._sequence) # easier :-)
    def __iter__(self): return self._main.itervalues()
    def count(self, item): return self._main.values().count(item)
    def index(self, item, *args): return self._main.values().index(item, *args)

    def reverse(self):
        """Reverse the values"""
        vals = self._main.values()
        vals.reverse()
        # FIXME: efficiency
        self[:] = vals

    def sort(self, *args, **kwds):
        """Sort the values."""
        vals = self._main.values()
        vals.sort(*args, **kwds)
        self[:] = vals

    def __mul__(self, n): return self._main.values()*n
    __rmul__ = __mul__
    def __add__(self, other): return self._main.values() + other
    def __radd__(self, other): return other + self._main.values()

    ## following methods not implemented for values ##
    def __delitem__(self, i): raise TypeError('Can\'t delete items from values')
    def __iadd__(self, other): raise TypeError('Can\'t add in place to values')
    def __imul__(self, n): raise TypeError('Can\'t multiply values in place')
    def append(self, item): raise TypeError('Can\'t append items to values')
    def insert(self, i, item): raise TypeError('Can\'t insert items into values')
    def pop(self, i=-1): raise TypeError('Can\'t pop items from values')
    def remove(self, item): raise TypeError('Can\'t remove items from values')
    def extend(self, other): raise TypeError('Can\'t extend values')

class SequenceOrderedDict(OrderedDict):
    """
    Experimental version of OrderedDict that has a custom object for ``keys``,
    ``values``, and ``items``.
    
    These are callable sequence objects that work as methods, or can be
    manipulated directly as sequences.
    
    Test for ``keys``, ``items`` and ``values``.
    
    >>> d = SequenceOrderedDict(((1, 2), (2, 3), (3, 4)))
    >>> d
    SequenceOrderedDict([(1, 2), (2, 3), (3, 4)])
    >>> d.keys
    [1, 2, 3]
    >>> d.keys()
    [1, 2, 3]
    >>> d.setkeys((3, 2, 1))
    >>> d
    SequenceOrderedDict([(3, 4), (2, 3), (1, 2)])
    >>> d.setkeys((1, 2, 3))
    >>> d.keys[0]
    1
    >>> d.keys[:]
    [1, 2, 3]
    >>> d.keys[-1]
    3
    >>> d.keys[-2]
    2
    >>> d.keys[0:2] = [2, 1]
    >>> d
    SequenceOrderedDict([(2, 3), (1, 2), (3, 4)])
    >>> d.keys.reverse()
    >>> d.keys
    [3, 1, 2]
    >>> d.keys = [1, 2, 3]
    >>> d
    SequenceOrderedDict([(1, 2), (2, 3), (3, 4)])
    >>> d.keys = [3, 1, 2]
    >>> d
    SequenceOrderedDict([(3, 4), (1, 2), (2, 3)])
    >>> a = SequenceOrderedDict()
    >>> b = SequenceOrderedDict()
    >>> a.keys == b.keys
    1
    >>> a['a'] = 3
    >>> a.keys == b.keys
    0
    >>> b['a'] = 3
    >>> a.keys == b.keys
    1
    >>> b['b'] = 3
    >>> a.keys == b.keys
    0
    >>> a.keys > b.keys
    0
    >>> a.keys < b.keys
    1
    >>> 'a' in a.keys
    1
    >>> len(b.keys)
    2
    >>> 'c' in d.keys
    0
    >>> 1 in d.keys
    1
    >>> [v for v in d.keys]
    [3, 1, 2]
    >>> d.keys.sort()
    >>> d.keys
    [1, 2, 3]
    >>> d = SequenceOrderedDict(((1, 2), (2, 3), (3, 4)), strict=True)
    >>> d.keys[::-1] = [1, 2, 3]
    >>> d
    SequenceOrderedDict([(3, 4), (2, 3), (1, 2)])
    >>> d.keys[:2]
    [3, 2]
    >>> d.keys[:2] = [1, 3]
    Traceback (most recent call last):
    KeyError: 'Keylist is not the same as current keylist.'

    >>> d = SequenceOrderedDict(((1, 2), (2, 3), (3, 4)))
    >>> d
    SequenceOrderedDict([(1, 2), (2, 3), (3, 4)])
    >>> d.values
    [2, 3, 4]
    >>> d.values()
    [2, 3, 4]
    >>> d.setvalues((4, 3, 2))
    >>> d
    SequenceOrderedDict([(1, 4), (2, 3), (3, 2)])
    >>> d.values[::-1]
    [2, 3, 4]
    >>> d.values[0]
    4
    >>> d.values[-2]
    3
    >>> del d.values[0]
    Traceback (most recent call last):
    TypeError: Can't delete items from values
    >>> d.values[::2] = [2, 4]
    >>> d
    SequenceOrderedDict([(1, 2), (2, 3), (3, 4)])
    >>> 7 in d.values
    0
    >>> len(d.values)
    3
    >>> [val for val in d.values]
    [2, 3, 4]
    >>> d.values[-1] = 2
    >>> d.values.count(2)
    2
    >>> d.values.index(2)
    0
    >>> d.values[-1] = 7
    >>> d.values
    [2, 3, 7]
    >>> d.values.reverse()
    >>> d.values
    [7, 3, 2]
    >>> d.values.sort()
    >>> d.values
    [2, 3, 7]
    >>> d.values.append('anything')
    Traceback (most recent call last):
    TypeError: Can't append items to values
    >>> d.values = (1, 2, 3)
    >>> d
    SequenceOrderedDict([(1, 1), (2, 2), (3, 3)])
    
    >>> d = SequenceOrderedDict(((1, 2), (2, 3), (3, 4)))
    >>> d
    SequenceOrderedDict([(1, 2), (2, 3), (3, 4)])
    >>> d.items()
    [(1, 2), (2, 3), (3, 4)]
    >>> d.setitems([(3, 4), (2 ,3), (1, 2)])
    >>> d
    SequenceOrderedDict([(3, 4), (2, 3), (1, 2)])
    >>> d.items[0]
    (3, 4)
    >>> d.items[:-1]
    [(3, 4), (2, 3)]
    >>> d.items[1] = (6, 3)
    >>> d.items
    [(3, 4), (6, 3), (1, 2)]
    >>> d.items[1:2] = [(9, 9)]
    >>> d
    SequenceOrderedDict([(3, 4), (9, 9), (1, 2)])
    >>> del d.items[1:2]
    >>> d
    SequenceOrderedDict([(3, 4), (1, 2)])
    >>> (3, 4) in d.items
    1
    >>> (4, 3) in d.items
    0
    >>> len(d.items)
    2
    >>> [v for v in d.items]
    [(3, 4), (1, 2)]
    >>> d.items.count((3, 4))
    1
    >>> d.items.index((1, 2))
    1
    >>> d.items.index((2, 1))
    Traceback (most recent call last):
    ValueError: list.index(x): x not in list
    >>> d.items.reverse()
    >>> d.items
    [(1, 2), (3, 4)]
    >>> d.items.reverse()
    >>> d.items.sort()
    >>> d.items
    [(1, 2), (3, 4)]
    >>> d.items.append((5, 6))
    >>> d.items
    [(1, 2), (3, 4), (5, 6)]
    >>> d.items.insert(0, (0, 0))
    >>> d.items
    [(0, 0), (1, 2), (3, 4), (5, 6)]
    >>> d.items.insert(-1, (7, 8))
    >>> d.items
    [(0, 0), (1, 2), (3, 4), (7, 8), (5, 6)]
    >>> d.items.pop()
    (5, 6)
    >>> d.items
    [(0, 0), (1, 2), (3, 4), (7, 8)]
    >>> d.items.remove((1, 2))
    >>> d.items
    [(0, 0), (3, 4), (7, 8)]
    >>> d.items.extend([(1, 2), (5, 6)])
    >>> d.items
    [(0, 0), (3, 4), (7, 8), (1, 2), (5, 6)]
    """

    def __init__(self, init_val=(), strict=True):
        OrderedDict.__init__(self, init_val, strict=strict)
        self._keys = self.keys
        self._values = self.values
        self._items = self.items
        self.keys = Keys(self)
        self.values = Values(self)
        self.items = Items(self)
        self._att_dict = {
            'keys': self.setkeys,
            'items': self.setitems,
            'values': self.setvalues,
        }

    def __setattr__(self, name, value):
        """Protect keys, items, and values."""
        if not '_att_dict' in self.__dict__:
            object.__setattr__(self, name, value)
        else:
            try:
                fun = self._att_dict[name]
            except KeyError:
                OrderedDict.__setattr__(self, name, value)
            else:
                fun(value)

if __name__ == '__main__':
    if INTP_VER < (2, 3):
        raise RuntimeError("Tests require Python v.2.3 or later")
    # turn off warnings for tests
    warnings.filterwarnings('ignore')
    # run the code tests in doctest format
    import doctest
    m = sys.modules.get('__main__')
    globs = m.__dict__.copy()
    globs.update({
        'INTP_VER': INTP_VER,
    })
    doctest.testmod(m, globs=globs)