Simple REST framework for Django

Devil aims to be simple to use REST framework for Django. Devil is influenced by piston.

Update 2.5.2016: Unmaintained. While the framework is being used in production, I have not tested it with recent versions of Django (Version 1.6 is what we are running).

Key Characteristics

• Resource is the key concept, everything builds around it.
• Builtin content negotiation (parsers / formatters).
• Gets out of your way
  • You can use additional features but you don't need to.
  • Everything is optional and works with default values.
  • Simple to get started.
• Flexible access control based on Django's users, groups and permissions.
  • Ability to assign CRUD operations per resource for each user (or group)
  • devil will auto-generate necessary permissions into the DB
  • You can use Django's admin interface to assign permissions to users and groups
  • After this devil automatically picks up request.user and performs authorization
• Intentionally doesn't give you CRUD for free as piston does
  • We can add this option later if it's concidered useful, but:
  • This rarely works for legacy systems anyway
  • For anything bigger, it's usually a good idea to decouple model and representation
• Ability to define representation using Django's forms
  • Automatic validation of incoming/outgoing data
  • Automatic documentation generation (Not implemented yet)

Table of Contents

• Installation
• Quick Example
• Step by Step Instructions
• URL Dispatching
• Method Dispatching
• Content Type Negotiation
• Dealing with Data
• HTTP Responses
• Defining Representations
• Using Object Factories
• Auto-generated Documentation
• Configuration
• License

Installation

pip install devil

Source code can be found at GitGub

Quick Example

resources.py:

from devil.resource import Resource

class MyTestResource(Resource):
    def get(self, request):
        return {'jedi': 'luke'}

urls.py:

from django.conf.urls.defaults import patterns, url
from devil.resource import Resource
import resources

mytestresource = resources.MyTestResource()
urlpatterns = patterns('',
    url(r'^test', mytestresource),
)

curl:

curl http://localhost:8000/test?format=json

> GET /contact/?format=json HTTP/1.1
> User-Agent: curl/7.21.4 (universal-apple-darwin11.0) libcurl/7.21.4
> Host: localhost:8000
> Accept: */*
>
< HTTP/1.0 200 OK
< Date: Tue, 14 Feb 2012 09:35:02 GMT
< Server: WSGIServer/0.1 Python/2.7.1
< Content-Type: application/json; charset=utf-8
<
    {
        "jedi": "luke"
    }

Anoher simple example is available here

Step by Step Instructions

$ pip install devil
$ django-admin.py startproject phone_book
$ cd phone_book
$ python manage.py startapp contacts

contacts/resources.py:

from devil.resource import Resource

class Contact(Resource):
    def get(self, request, *args, **kw):
        return {'name': 'Luke Skywalker'}

urls.py:

from django.conf.urls.defaults import patterns, url
from contacts import resources

contacts_resource = resources.Contact()

urlpatterns = patterns('',
    url(r'contact', contacts_resource),
)

start the server and in the console say (or you can use a browser):

$ python manage.py runserver
$ curl http://localhost:8000/contact?format=json

or if you want to be more HTTP friendly:

$ curl http://localhost:8000/contact -H 'Accept: application/json'

A more complete example can be found under examples/userdb. See the README for instructions on running the example and how to examine the code.

URL Dispatching

The relationship between URLs and RESTful resources is one to many. That is, one resource may have several URLs mapped to it. Conversely, one URL is always mapped into a single resource. Devil uses Django's built in URL dispatching to define these mappings. If you are familiar with Django's terms and concepts the resources in Devil become the views of Django.

Say you define your resources in a module called resources. Then in your urls.py file you would instantiate and map your resources to URLs, like so:

user_resource = resources.UserResource()

urlpatterns = patterns('',
    url(r'/user', user_resource),
)

And to define aliases, you can just add new mappings to the same resource:

urlpatterns = patterns('',
    url(r'/user', user_resource),
    url(r'/jedi', user_resource),
    url(r'/sith', user_resource),
)

You can use Django's built-in regexp features like named parameters:

urlpatterns = patterns('',
    url(r'/user(?P<id>\d{1,7})?', user_resource),
)

In this case, the id property would be available in the resource method:

class UserResource(Resource):
    def get(self, request, id, *args, **kw):
        print id

or

class UserResource(Resource):
    def get(self, request, *args, **kw):
        print kw['id']

Method Dispatching

Devil maps the HTTP request methods into functions of the resource directly. So, if Devil receives an HTTP POST request, it will try and find an instance method called post in the resource and invoke it. If the resource doesn't define post method, Devil will automatically return 405 Mehod Not Allowed to the client. The signature for the method for PUT and POST requests is:

def post(self, data, request):

and for others methods:

def get(self, request):

so, PUTs and POSTs will have additional data attribute that contains the (possibly parsed) content body of the request. Also, bear in mind that function parameters may also include named parameters from url mappings.

Content Type Negotiation

Devil uses the terms parser and formatter for data decoding and encoding respectively. They are collectively referred to as data mappers. Devil uses data mappers to parse all data that comes in with PUT and POST requests (e.g. JSON, XML or plaintext). Similarly, devil uses data mappers to automatically format all outgoing data when it is present. The mapper to be used for a given request can be defined in one of the following places (note that this list is not sorted by precedence):

• In the URL:
  • either with ?format=json
  • or with .json suffix
• HTTP Accept header. The Accept header supports the full format, as in: Accept: audio/*; q=0.2, audio/basic
• HTTP Content-Type header (meaningful only for PUT and POST)
• A resource may define its own mapper which will take precedence over anything else
  • define mapper in your derived Resource class (see examples)
• A resource may define a default mapper that will be used if the client specifies no content type
  • define default_mapper in your derived Resource class (see examples)
• Application may define one system wide default mapper by registering a mapper with content type */*

If the client specifies a content type that is not supported, devil responds with 406 Not Acceptable. Out of the box, devil supports plain/text, application/json and text/xml. You can register more mappers for your application of course. It should be noted that the built-in XML mapper has some restrictions (see the docstring).

Following picture formally defines how a correct formatter is chosen for encoding the outgoing data:

Likewise, the next picture defines how a correct parser is chosen for the incoming (via PUT or POST) data:

See the docstrings in the DataMapper class and the example resources in tests for instructions on how to implement your own mappers.

Dealing with Data

Once the appropriate data mapper has been chosen, devil can perform decoding for the incoming data and encoding for the outgoing data. For example, if a json mapper is chosen your post and get functions would look something like this (in terms of data passing):

$ curl http://localhost:8000/contact -X POST -d '{"name": "Darth Maul", "age": 24}' -H 'Content-Type: application/json'

def post(self, data, request):
    # data is available as a python dict in the data parameter
    print data['name']       # Darth Maul
    print type(data['age'])  # <type 'int'>

$ curl http://localhost:8000/contact/1 -X GET -H 'Accept: application/json'

    def get(self, request):
        # you can return a python dictionary
        return {
            'name': 'Yoda',
            'age': '876',
        }

Devil's built-in json and xml mappers will convert to and from python dictionaries and lists. However, the built-in text (text/plain) mapper will only convert between strings and unicode objects.

HTTP Responses

A resource (that is, any of its post/get/put/delete methods) may return following values:

• dictionary
• list
• string
• None
• Devil's http.Response
• Django's HttpResponse

If the resource returns Django's HttpResponse, devil doesn't touch the return value at all but just passes it on to the client. If the return type is any of the other five, devil tries to encode the data using the appropriate mapper. Furthermore, devil's Response object provides a way for the resource to include HTTP response code and headers along with the data. Devil will automatically use response code 200 OK in cases where the response code isn't explicitly defined by the resource. Also, devil will automatically add Content-Type HTTP header based on the used data mapper.

Error situations may be handled using exceptions defined in devil.errors package. So whenever there's an error situation and you want to return a certain response code, you can raise a HttpStatusCodeError and devil will catch it and turn it into appropriate HTTP response object.

from devil import errors
def post(self, data, request):
    if data['age'] > 50:
        # too old
        raise errors.BadRequest("you're too old")

In the example, the client would receive 400 BAD REQUEST with the string "you're too old" in the body whenever the age is above 50.

Defining Representations

Devil uses Django's form fields to define representations. By defining representation for your resource, you gain three advantages: automatic data validation, possibility to use object factories to automatically create python objects from your data and the possibility to have auto-generated documentation for your resources. The easiest way to define a representation for your resource is to subclass from Devil's Representation class and use fields defined in Django and Devil to define all the properties.

Following example shows a simple representation using Django's form fields:

from django import forms
from devil import Representation

class UserRepresentation(Representation):
    username = forms.CharField(max_length=30)
    joined = forms.DateField(input_formats='%Y-%m-%d')

Devil provides couple of handy fields for defining more complex representations, mainly NestedField and ListField. These can be used for speficying composite and collection fields respectively:

from devil.fields import ListField, NestedField, EnumField

class EmailAddress(NestedField):
    email = forms.EmailField()
    type = forms.CharField(required=False)

class UserRepresentation(Representation):
    username = forms.CharField(max_length=30)
    joindate = forms.DateField(input_formats=('%Y-%m-%d',))
    emailAddresses = EmailAddress()
    accountType = EnumField('root', 'novice', 'intermediate')
    tags = ListField(forms.CharField(max_length=20), required=False)

The EmailField defined here will be reusable in any representation that embeds email addresses. The UserRepresentation on the other hand, is a complete representation that can be used in a resource to perform automatic data validation. To see how to put UserRepresentation into use, see configuration.

Using object factories

After a data mapper has converted the data into python dictionary, an object factory may further convert the dictionary into python object. Against common convention for implementing factories, Devil's factory also provides serialize() function which is the counter operation of create().

Factory uses representation object as a specification when performing object creation and serialization. Normally a field speficied in the representation knows how to marshal/unmarshal itself but if not, a subclassed factory may provide create_foo() and serialize_foo() functions to provide custom creation/serialization of a field (foo in our example). Following example demonstrates the definition of UserFactory.

from devil import Factory

class User(object):
    pass

class UserFactory(Factory):

    klass = User
    spec = UserRepresentation()

An example of manually using UserFactory:

testdata = {
    'username': 'jedi',
    'joindate': '2012-01-18',
    'emailAddress': {
        'email': 'jedi@rebillion.net',
        'type': 'work',
    },
    'accountType': 'root',
    'tags': ['jedi', 'active', 'rebel'],
}

factory = UserFactory()
user = factory.create(testdata)
print user.username  # 'jedi'
print user.joindate  # datetime.date(2012, 1, 18)
print user.emailAddress.email  # 'jedi@rebillion.net'

To see how to instruct Devil to automatically convert incoming data into python objects using factories, see configuration.

Factories also support name mangling of properties. For example:

emailAddress -> email_address
email_address -> emailAddress

This is achieved by providing alias property for a field. For Example:

class UserRepresentation(Representation):
    emailAddress = EmailField(alias='email_address')

This definition wouldn't have any affect on data when in serialized format. That is, in JSON or in dictionary the property name is still emailAddress. However, when the field is present in python object (i.e. after object has been created with factory.create() or when it is given to factory.serialize()) the name of the property is email_address.

Auto-generated Documentation

NOTE This is still a work in progress but there is a very bare (but working) implementation in the doc branch.

Basically, by subclassing your resources from DocumentedResource instead of Resource you add support for auto- generated documenatation. When client provides _doc in the query string of the request, a documentation is generated instead of the actual method being executed. The documentation includes:

• docstring of the implementing resource-class
• all supported methods and their docstrings
• representations (if defined) and their constraints

Configuration

Resources in Devil can be configured by overriding any of the following class properties defined in the devil.Resource class:

class Resource(object):
    access_controller = None
    allow_anonymous = True
    authentication = None
    representation = None
    post_representation = None
    factory = None
    post_factory = None
    default_mapper = None
    mapper = None

It is usually a good idea to derive devil.Resource and use the derived class as a base class for your project's resources. That way, you only need to define these configurations once. The values shown above are the default values that are used if not overridden.

representation

When defined, this representation will be used for validating both, incoming and outgoing data. For example: representation = UserRepresentation(). More information on representations is available here

post_representation

When defined, this representation will be used for validating client-data inside a POST request. If not defined representation will be used for POST requests too. post_representation will never be used for validating outgoing data (i.e. data returned to the client).

factory

When defined, Devil uses this factory to automatically convert incoming data into python objects and outgoing objects into serialized format. An example definition looks like this: factory = UserFactory(). More information on factories is available here .

post_factory

When defined, this factory will be used for creating a python object from data inside a POST request. If not defined, factory will be used for POST requests too. post_factory will never be used for serializing outgoing data.

authentication

Defines the authentication handler. When provided, it should be an object that has a function called authentication. This function takes only one paremeter: the request object. The function should raise devil.errors.Unauthorized if the request cannot be authorized otherwise it must populate the request.user property.

Instead of implementing your own authentication you are free to use devil's HTTP basic authentication or one of Django's authentication middlewares. To use Devil's authentication, you can simply add this to your resource class:

from devil.auth import HttpBasic

class MyResource(Resource):
    authentication = HttpBasic()

allow_anonymous

If the resource allows anonymous access, set this to True (default). Note, that this is different from authentication as you can turn authentication on but still allow anonymous access (for example with limited access).

access_controller

If access_controller is set, devil uses it to authorize requests. The value of access_controller must be an object that has a method called check_perm. This method takes two parameters: the request object and the resource that is being queried. If the request is permitted, the function should return nothing. However, if the request is not permitted, the function should raise devil.errors.Forbidden.

While you are free to implement your own access controller, devil comes with an access controller based on Django's users, groups and permissions. To make use of it, you would say this in your resource code:

from devil.resource import Resource
from devil.perm.acl import PermissionController

class MyResource(Resource):
    access_controller = PermissionController()

After this, all requests to MyResource will be authorized based on the user who sent the request and the method that was used. So, each resource has four possible permissions associated with it: post, get, put and delete. Devil will automatically check that the user (or the user's group) making the request has appropriate permission to access the resource.

Devil can automatically add all necessary permissions into the database but you (or the site administrator) will need to assign these permissions for users (or groups). To have devil populate all possible permissions into the database you need to tell devil what are the resources you want to protect and then use syncdb to insert them. For all this, you would first add two things into your Django settings.py file:

1. ACL_RESOURCES to define your protected resources
2. devil.perm among INSTALLED_APPS

For example:

ACL_RESOURCES = (
    'myproject.myapp.urls.acl_resources',
    )

INSTALLED_APPS = (
    'devil.perm',
    )

Now, you also need to have your protected resources as a list in the urls.py file. For example:

myresource = MyResource()

acl_resources = (
    myresource,
    )

urlpatterns = (
    url(r'^test', myresource),
    )

After this, you can run python manage.py syncdb and have devil to insert all necessary permissions into the database. To be exact, devil inserts them into Django's auth_permission table. The format for the permission name is prefix_resourcename_method. The prefix is "resource", the resourcename is whatever your resource class´ name() function returns and method is the request method (i.e. "post", "get", "put" or "delete"). If you haven't overridden the name() function in your resource class, it returns the name of the class with CamelCase converted to slashes. So, in our example running the syncdb would generate following four permissions into the database:

• "resource_my/resource_post"
• "resource_my/resource_get"
• "resource_my/resource_put"
• "resource_my/resource_delete"

Now, you are ready to assign these permissions to your users or groups by using the Django admin interface.

Few notes:

• Devil's access controller depends on the request.user property. This means that the request must have been authenticated before it lands to the access controller. The easiest way to handle authentication is to use devil's HTTP basic authentication or one of Django's authentication middlewares.
• request.user may not be Django's AnonymourUser
• If the user making the request is a super user, permission is always granted.
• The user needs to be active in order to get access to any resource that is protected by devin's access controller.
• Devil's access controller uses Django's get_all_permissions function to figure out whether the user has the permission.

License

(The MIT License)

Copyright (c) 2012 Janne Kuuskeri

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the 'Software'), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.