Question

I have the following function which does a crude job of parsing an XML file into a dictionary.

Unfortunately, since Python dictionaries are not ordered, I am unable to cycle through the nodes as I would like.

How do I change this so it outputs an ordered dictionary which reflects the original order of the nodes when looped with for.

def simplexml_load_file(file):
    import collections
    from lxml import etree

    tree = etree.parse(file)
    root = tree.getroot()

    def xml_to_item(el):
        item = None
        if el.text:
            item = el.text
        child_dicts = collections.defaultdict(list)
        for child in el.getchildren():
            child_dicts[child.tag].append(xml_to_item(child))
        return dict(child_dicts) or item

    def xml_to_dict(el):
        return {el.tag: xml_to_item(el)}

    return xml_to_dict(root)

x = simplexml_load_file('routines/test.xml')

print x

for y in x['root']:
    print y

Outputs:

{'root': {
    'a': ['1'],
    'aa': [{'b': [{'c': ['2']}, '2']}],
    'aaaa': [{'bb': ['4']}],
    'aaa': ['3'],
    'aaaaa': ['5']
}}

a
aa
aaaa
aaa
aaaaa

How can I implement collections.OrderedDict so that I can be sure of getting the correct order of the nodes?

XML file for reference:

<root>
    <a>1</a>
    <aa>
        <b>
            <c>2</c>
        </b>
        <b>2</b>
    </aa>
    <aaa>3</aaa>
    <aaaa>
        <bb>4</bb>
    </aaaa>
    <aaaaa>5</aaaaa>
</root>

Answer 1

You could use the new OrderedDictdict subclass which was added to the standard library’s collections module in version 2.7^✶. Actually what you need is an Ordered+defaultdict combination which doesn’t exist — but it’s possible to create one by subclassing OrderedDict as illustrated below:

^{✶ If your version of Python doesn’t have OrderedDict, you should be able use Raymond Hettinger’s Ordered Dictionary for Py2.4 ActiveState recipe as the base class instead.}

import collections

class OrderedDefaultdict(collections.OrderedDict):
    """ A defaultdict with OrderedDict as its base class. """

    def __init__(self, default_factory=None, *args, **kwargs):
        if not (default_factory is None or callable(default_factory)):
            raise TypeError('first argument must be callable or None')
        super(OrderedDefaultdict, self).__init__(*args, **kwargs)
        self.default_factory = default_factory  # called by __missing__()

    def __missing__(self, key):
        if self.default_factory is None:
            raise KeyError(key,)
        self[key] = value = self.default_factory()
        return value

    def __reduce__(self):  # Optional, for pickle support.
        args = (self.default_factory,) if self.default_factory else tuple()
        return self.__class__, args, None, None, iter(self.items())

    def __repr__(self):  # Optional.
        return '%s(%r, %r)' % (self.__class__.__name__, self.default_factory, self.items())

def simplexml_load_file(file):
    from lxml import etree

    tree = etree.parse(file)
    root = tree.getroot()

    def xml_to_item(el):
        item = el.text or None
        child_dicts = OrderedDefaultdict(list)
        for child in el.getchildren():
            child_dicts[child.tag].append(xml_to_item(child))
        return collections.OrderedDict(child_dicts) or item

    def xml_to_dict(el):
        return {el.tag: xml_to_item(el)}

    return xml_to_dict(root)

x = simplexml_load_file('routines/test.xml')
print(x)

for y in x['root']:
    print(y)

The output produced from your test XML file looks like this:

{'root':
    OrderedDict(
        [('a', ['1']),
         ('aa', [OrderedDict([('b', [OrderedDict([('c', ['2'])]), '2'])])]),
         ('aaa', ['3']),
         ('aaaa', [OrderedDict([('bb', ['4'])])]),
         ('aaaaa', ['5'])
        ]
    )
}

a
aa
aaa
aaaa
aaaaa

Which I think is close to what you want.

Minor update:

Added a __reduce__() method which will allow the instances of the class to be pickled and unpickled properly. This wasn’t necessary for this question, but came up in a similar one.