aiodjango 0.1.0

Utilities for running async aiohttp based routes in the context of a Django project.

This is a proof-of-concept experiment to combine a Django WSGI app mixed with async views/websocket handlers using aiohttp. The API is highly unstable and I wouldn’t recommend that you use this code for anything other than wild experimentation.

How It Works

aidjango.get_aio_application builds an application which combines both request handlers/views from Django and aiohttp.web. Views are defined using the normal Django url pattern syntax but any handler which is a coroutine is handled by the aiohttp application while the rest of the views are handled by the normal Django app.

Internal this makes use of aiohttp-wsgi which runs the Django WSGI app in a thread-pool to minimize blocking the async portions of the app.

Running the Demo

The example project requires Python 3.4+ to run. You should create a virtualenv to install the necessary requirements:

$ git clone https://github.com/mlavin/aiodjango.git
$ cd aiodjango/
$ mkvirtualenv aiodjango -p `which python3.4`
(aiodjango) $ add2virtualenv .
(aiodjango) $ cd example
(aiodjango) $ pip install -r requirements.txt
(aiodjango) $ python manage.py migrate
(aiodjango) $ python manage.py runserver

This starts the server on http://localhost:8000/ with a new version of Django’s built-in runserver. For a more multi-process server you can run using the aiohttp worker along with Gunicorn:

(aiodjango) $ gunicorn example.wsgi:app --worker-class aiohttp.worker.GunicornWebWorker --workers 2