xmlsjon converts XML into Python dictionary structures (trees, like in JSON) and vice-versa.
Clone or download



xmljson converts XML into Python dictionary structures (trees, like in JSON) and vice-versa.


XML can be converted to a data structure (such as JSON) and back. For example:

        <name value="Alice"/>
        <name value="Bob"/>

can be converted into this data structure (which also a valid JSON object):

    "employees": [{
        "person": {
            "name": {
                "@value": "Alice"
    }, {
        "person": {
            "name": {
                "@value": "Bob"

This uses the BadgerFish convention that prefixes attributes with @. The conventions supported by this library are:

  • Abdera: Use "attributes" for attributes, "children" for nodes
  • BadgerFish: Use "$" for text content, @ to prefix attributes
  • Cobra: Use "attributes" for sorted attributes (even when empty), "children" for nodes, values are strings
  • GData: Use "$t" for text content, attributes added as-is
  • Parker: Use tail nodes for text content, ignore attributes
  • Yahoo Use "content" for text content, attributes added as-is

Convert data to XML

To convert from a data structure to XML using the BadgerFish convention:

>>> from xmljson import badgerfish as bf
>>> bf.etree({'p': {'@id': 'main', '$': 'Hello', 'b': 'bold'}})

This returns an array of etree.Element structures. In this case, the result is identical to:

>>> from xml.etree.ElementTree import fromstring
>>> [fromstring('<p id="main">Hello<b>bold</b></p>')]

The result can be inserted into any existing root etree.Element:

>>> from xml.etree.ElementTree import Element, tostring
>>> result = bf.etree({'p': {'@id': 'main'}}, root=Element('root'))
>>> tostring(result)
'<root><p id="main"/></root>'

This includes lxml.html as well:

>>> from lxml.html import Element, tostring
>>> result = bf.etree({'p': {'@id': 'main'}}, root=Element('html'))
>>> tostring(result, doctype='<!DOCTYPE html>')
'<!DOCTYPE html>\n<html><p id="main"></p></html>'

For ease of use, strings are treated as node text. For example, both the following are the same:

>>> bf.etree({'p': {'$': 'paragraph text'}})
>>> bf.etree({'p': 'paragraph text'})

By default, non-string values are converted to strings using Python's str, except for booleans -- which are converted into true and false (lower case). Override this behaviour using xml_fromstring:

>>> tostring(bf.etree({'x': 1.23, 'y': True}, root=Element('root')))
>>> from xmljson import BadgerFish              # import the class
>>> bf_str = BadgerFish(xml_tostring=str)       # convert using str()
>>> tostring(bf_str.etree({'x': 1.23, 'y': True}, root=Element('root')))

Convert XML to data

To convert from XML to a data structure using the BadgerFish convention:

>>> bf.data(fromstring('<p id="main">Hello<b>bold</b></p>'))
{"p": {"$": "Hello", "@id": "main", "b": {"$": "bold"}}}

To convert this to JSON, use:

>>> from json import dumps
>>> dumps(bf.data(fromstring('<p id="main">Hello<b>bold</b></p>')))
'{"p": {"b": {"$": "bold"}, "@id": "main", "$": "Hello"}}'

To preserve the order of attributes and children, specify the dict_type as OrderedDict (or any other dictionary-like type) in the constructor:

>>> from collections import OrderedDict
>>> from xmljson import BadgerFish              # import the class
>>> bf = BadgerFish(dict_type=OrderedDict)      # pick dict class

By default, values are parsed into boolean, int or float where possible (except in the Yahoo method). Override this behaviour using xml_fromstring:

>>> dumps(bf.data(fromstring('<x>1</x>')))
'{"x": {"$": 1}}'
>>> bf_str = BadgerFish(xml_fromstring=False)   # Keep XML values as strings
>>> dumps(bf_str.data(fromstring('<x>1</x>')))
'{"x": {"$": "1"}}'
>>> bf_str = BadgerFish(xml_fromstring=repr)    # Custom string parser
'{"x": {"$": "\'1\'"}}'

xml_fromstring can be any custom function that takes a string and returns a value. In the example below, only the integer 1 is converted to an integer. Everything else is retained as a float:

>>> def convert_only_int(val):
...     return int(val) if val.isdigit() else val
>>> bf_int = BadgerFish(xml_fromstring=convert_only_int)
>>> dumps(bf_int.data(fromstring('<p><x>1</x><y>2.5</y><z>NaN</z></p>')))
'{"p": {"x": {"$": 1}, "y": {"$": "2.5"}, "z": {"$": "NaN"}}}'


To use a different conversion method, replace BadgerFish with one of the other classes. Currently, these are supported:

>>> from xmljson import abdera          # == xmljson.Abdera()
>>> from xmljson import badgerfish      # == xmljson.BadgerFish()
>>> from xmljson import cobra           # == xmljson.Cobra()
>>> from xmljson import gdata           # == xmljson.GData()
>>> from xmljson import parker          # == xmljson.Parker()
>>> from xmljson import yahoo           # == xmljson.Yahoo()


Conventions may support additional options.

The Parker convention absorbs the root element by default. parker.data(preserve_root=True) preserves the root instance:

>>> from xmljson import parker, Parker
>>> from xml.etree.ElementTree import fromstring
>>> from json import dumps
>>> dumps(parker.data(fromstring('<x><a>1</a><b>2</b></x>')))
'{"a": 1, "b": 2}'
>>> dumps(parker.data(fromstring('<x><a>1</a><b>2</b></x>'), preserve_root=True))
'{"x": {"a": 1, "b": 2}}'


This is a pure-Python package built for Python 2.6+ and Python 3.0+. To set up:

pip install xmljson


  • Test cases for Unicode
  • Support for namespaces and namespace prefixes
  • Support XML comments