alf

.. image:: https://travis-ci.org/globocom/alf.svg?branch=master :target: https://travis-ci.org/globocom/alf

Python OAuth 2 Client

alf is an OAuth 2 Client based on requests.Session <http://docs.python-requests.org/en/latest/user/advanced/#session-objects>_ with seamless support for the Client Credentials Flow <http://tools.ietf.org/html/draft-ietf-oauth-v2-31#section-1.3.4>_.

.. image:: /assets/alf.jpeg?raw=true

Features

• Automatic token retrieving and renewing
• Token expiration control
• Automatic token storage
• Automatic retry on status 401 (UNAUTHORIZED)

Usage

Initialize the client and use it as a requests.Session <http://docs.python-requests.org/en/latest/user/advanced/#session-objects>_ object.

.. code-block:: python

from alf.client import Client

alf = Client(
    token_endpoint='http://example.com/token',
    client_id='client-id',
    client_secret='secret')

resource_uri = 'http://example.com/resource'

alf.put(
    resource_uri, data='{"name": "alf"}',
    headers={'Content-Type': 'application/json'})

alf.get(resource_uri)

alf.delete(resource_uri)

Using your custom token storage

Now passing an object with get and set attributes you can store or retrieve a token.

This object can be a Redis, Memcache or your custom object.

.. code-block:: python

from alf.client import Client
from redis import StrictRedis

redis = StrictRedis(host='localhost', port=6379, db=0)

alf = Client(
    token_endpoint='http://example.com/token',
    client_id='client-id',
    client_secret='secret',
    token_storage=redis)

resource_uri = 'http://example.com/resource'

alf.put(
    resource_uri, data='{"name": "alf"}',
    headers={'Content-Type': 'application/json'})

alf.get(resource_uri)

alf.delete(resource_uri)

How does it work?

Before the request, a token will be requested on the authentication endpoint and a JSON response with the access_token and expires_in keys will be expected.

Multiple attempts will be issued after an error response from the endpoint if the token_retries argument is used. Check token-retrying_ for more info.

alf keeps the token until it is expired according to the expires_in value.

The token will be used on a Bearer authorization header <http://tools.ietf.org/html/draft-ietf-oauth-v2-31#section-7.1>_ for the original request.

.. code-block::

GET /resource/1 HTTP/1.1
Host: example.com
Authorization: Bearer token-12312

If the request fails with a 401 (UNAUTHORIZED) status, a new token is retrieved from the endpoint and the request is retried. This happens only once, if it fails again the error response is returned.

The token will be reused for every following request until it is expired.

.. _token-retrying:

Token Retrying

The client supports the retry interface from urllib3 <https://urllib3.readthedocs.org/en/latest/helpers.html?highlight=retry#module-urllib3.util.retry>_ to repeat attempts to retrieve the token from the endpoint.

The following code will retry the token request 5 times when the response status is 500 and it will wait 0.3 seconds longer after each error (known as backoff <https://en.wikipedia.org/wiki/Exponential_backoff>_).

.. code-block:: python

from requests.packages.urllib3.util import Retry
from alf.client import Client

alf = Client(
    token_endpoint='http://example.com/token',
    client_id='client-id',
    client_secret='secret',
    token_retry=Retry(total=5, status_forcelist=[500], backoff_factor=0.3))

Workflow

.. image:: /assets/workflow.png?raw=true

Troubleshooting

In case of an error retrieving a token, the error response will be returned, the real request won't happen.

Related projects

djalf <https://github.com/viniciuschagas/djalf>_ ''''''''''''''''''''''''''''''''''''''''''''''''''

An extended client that uses Django's cache backend to share tokens between server instances.

tornado-alf <https://github.com/globocom/tornado-alf>_ ''''''''''''''''''''''''''''''''''''''''''''''''''''''''

A port of the alf client using tornado's AsyncHTTPClient.