Fork me on GitHub

jsonpublish – configurable JSON encoder

This package provides configurable JSON encoder based on simplejson or json module from Python’s standard library.

When and why you should use jsonpublish:

  • You want all JSON serialization code to be in one place.
  • You want your serialization code to be flexible and structured.
  • Sometimes you want to alter serialization for some objects.

Custom types serialization

Suppose you have some data of your application modeled as Python’s classes (it may be, for example, Django models or just plain old Python’s classes):

class User(object):

  def __init__(self, username, birthday):
    self.username = username
    self.birthday = birthday

Now if you want to serialize User objects as JSON documents you can’t simply use json module, because it just doesn’t know how to represent your objects as JSON documents. So you need to write a function which converts User objects to something which can be serialized, for example dict. With time your app grows and complexity grows along so you need somehow to structure you serialization machinery, let’s see how jsonpublish can help us there:

from jsonpublish import register_adapter

def adapt_user(user):
  return {
    "username": user.username,
    "birthday": user.birthday

Now you can serialize your User objects:

>>> from jsonpublish import dumps
>>> print dumps(User("andrey", 1987))
{"username": "andrey", "birthday": 1987}

Parametrized adapters

Sometimes you want to alter serialization of some object, For example, let’s write another adapter for User objects which can change it behaviour based on arguments given:

def adapt_user(user, include_birthday=True):
  if include_birthday:
    return {
      "username": user.username,
      "birthday": user.birthday
    return {"username": user.username}

The question now is how to pass include_birthday keyword argument right to adapter, the answer is to use jsonpublish.jsonsettings():

>>> from jsonpublish import jsonsettings
>>> user = User("andrey", 1987)
>>> user_m = jsonsettings(user, include_birthday=False)

>>> print dumps(user)
{"username": "andrey", "birthday": 1987}

>>> print dumps(user_m)
{"username": "andrey"}

As you can see, by wrapping our User object in jsonpublish.jsonsettings() we can pass arbitrary keyword arguments to corresponding adapter so we can alter serialization by per-object basis.

Function jsonsettings actually doesn’t alter object in any way, it just “annotates” it with some metadata needed for corresponding adapter. You can work with wrapped object as before – all methods and attributes are still there and even isinstance check works the right way:

>>> user_m == user
>>> user_m.username
>>> isinstance(user_m, User)

Reporting bugs and working on jsonpublish

Development takes place at GitHub, you can clone source code repository with the following command:

% git clone git://

In case submitting patch or GitHub pull request please ensure you have corresponding tests for your bugfix or new functionality.

API reference

jsonpublish.dumps(obj, skipkeys=False, ensure_ascii=True, check_circular=True, allow_nan=True, indent=None, separators=None, encoding='utf-8', default=None, use_decimal=True, namedtuple_as_object=True, tuple_as_array=True, **kw)

Serialize obj using globally configured JSON encoder

Accepted arguments are the same as json.dumps() accepts

jsonpublish.register_adapter(typ, adapter=None)

Register adapter for type typ

If no adapter supplied then this method returns decorator.

jsonpublish.jsonsettings(o, **settings)

Create a proxy which carries JSON encoder settings

class jsonpublish.JSONEncoder(*args, **kwargs)

Configurable JSON encoder

It serializes object by consulting adapter registry. Registry can be modified by accessing adapters attribute of encoder which is of type AdapterRegistry.

Attr adapters:instance of AdapterRegistry which is used for serialization by encoder

Return a JSON string representation of a Python data structure.

>>> JSONEncoder().encode({"foo": ["bar", "baz"]})
'{"foo": ["bar", "baz"]}'
iterencode(o, _one_shot=False)

Encode the given object and yield each string representation as available.

For example:

for chunk in JSONEncoder().iterencode(bigobject):
class jsonpublish.AdapterRegistry

Registry of adapters


Lookup adapter for typ

register_adapter(typ, adapter=None)

Register adapter for type typ

If no adapter supplied then this method returns decorator.