在python中使用有序的dict作为对象字典

Question

I don't know why this doesn't work:

I'm using the odict class from PEP 372 , but I want to use it as a __dict__ member, ie:

class Bag(object):
    def __init__(self):
        self.__dict__ = odict()

But for some reason I'm getting weird results. This works:

>>> b = Bag()
>>> b.apple = 1
>>> b.apple
1
>>> b.banana = 2
>>> b.banana
2

But trying to access the actual dictionary doesn't work:

>>> b.__dict__.items()
[]
>>> b.__dict__
odict.odict([])

And it gets weirder:

>>> b.__dict__['tomato'] = 3
>>> b.tomato
3
>>> b.__dict__
odict.odict([('tomato', 3)])

I'm feeling very stupid. Can you help me out?

Answer 1

The closest answer to your question that I can find is at http://mail.python.org/pipermail/python-bugs-list/2006-April/033155.html .

Basically, if __dict__ is not an actual dict() , then it is ignored, and attribute lookup fails.

The alternative for this is to use the odict as a member, and override the getitem and setitem methods accordingly.

>>> class A(object) :
...     def __init__(self) :
...             self.__dict__['_odict'] = odict()
...     def __getattr__(self, value) :
...             return self.__dict__['_odict'][value]
...     def __setattr__(self, key, value) :
...             self.__dict__['_odict'][key] = value
... 
>>> a = A()
>>> a
<__main__.A object at 0xb7bce34c>
>>> a.x = 1
>>> a.x
1
>>> a.y = 2
>>> a.y
2
>>> a.odict
odict.odict([('x', 1), ('y', 2)])

Answer 2

Everything in sykora's answer is correct. Here's an updated solution with the following improvements:

1. works even in the special case of accessing a.__dict__ directly
2. supports copy.copy()
3. supports the == and != operators
4. uses collections.OrderedDict from Python 2.7.

...

from collections import OrderedDict

class OrderedNamespace(object):
    def __init__(self):
        super(OrderedNamespace, self).__setattr__( '_odict', OrderedDict() )

    def __getattr__(self, key):
        odict = super(OrderedNamespace, self).__getattribute__('_odict')
        if key in odict:
            return odict[key]
        return super(OrderedNamespace, self).__getattribute__(key)

    def __setattr__(self, key, val):
        self._odict[key] = val

    @property
    def __dict__(self):
        return self._odict

    def __setstate__(self, state): # Support copy.copy
        super(OrderedNamespace, self).__setattr__( '_odict', OrderedDict() )
        self._odict.update( state )

    def __eq__(self, other):
        return self.__dict__ == other.__dict__

    def __ne__(self, other):
        return not self.__eq__(other)

Answer 3

If you're looking for a library with attribute access to OrderedDict, the orderedattrdict package provides this.

>>> from orderedattrdict import AttrDict
>>> conf = AttrDict()
>>> conf['z'] = 1
>>> assert conf.z == 1
>>> conf.y = 2
>>> assert conf['y'] == 2
>>> conf.x = 3
>>> assert conf.keys() == ['z', 'y', 'x']

Disclosure: I authored this library. Thought it might help future searchers.