Django EventStream

EventStream provides API endpoints for your Django application that can push data to connected clients. Data is sent using the Server-Sent Events protocol (SSE), in which data is streamed over a never-ending HTTP response.

For example, you could create an endpoint, /events/, that a client could connect to with a GET request:

GET /events/ HTTP/1.1
Host: api.example.com
Accept: text/event-stream

The client would receive a streaming HTTP response with content looking like this:

HTTP/1.1 200 OK
Transfer-Encoding: chunked
Connection: Transfer-Encoding
Content-Type: text/event-stream

event: message
data: {"foo": "bar"}

event: message
data: {"bar": "baz"}

...

Features:

Easy to consume from browsers or native applications.
Highly reliable. Events can be persisted to your database, so clients can recover if they get disconnected.
Set per-user channel permissions.
Clean API contract that could be exposed to third parties if desired.

Requirements

This library requires either:

Django Channels 3, for native asynchronous connection handling (needs Python 3.6+).

A GRIP-compatible proxy such as Pushpin or Fanout Cloud, for delegating the connection handling and keeping the Django app stateless.

Note that it is possible to combine the two. If the app is set up with Channels and a connection arrives through a GRIP proxy, then the handling will be delegated.

Setup

We recommend setting up your project with Channels as this will give you the most flexibility, including being able to run standalone or with runserver.

Otherwise, see Setup without Channels.

Setup with Channels

First, install this module and the channels module:

pip install django-eventstream channels

Add the channels and django_eventstream apps to your settings.py:

INSTALLED_APPS = [
    ...
    'channels',
    'django_eventstream',
]

Add the GripMiddleware:

MIDDLEWARE = [
    'django_grip.GripMiddleware',
    ...
]

Channels introduces an entirely separate routing system for handling async connections. You'll need to declare an ASGI application instead of (or in addition to) a WSGI application.

For example, create an asgi.py file in your Django project dir (next to settings.py) with an endpoint declared:

"""
ASGI entrypoint. Configures Django and then runs the application
defined in the ASGI_APPLICATION setting.
"""

import os
import django
from django.core.asgi import get_asgi_application
from django.conf.urls import url
from channels.routing import ProtocolTypeRouter, URLRouter
from channels.auth import AuthMiddlewareStack
import django_eventstream

os.environ.setdefault("DJANGO_SETTINGS_MODULE", "server.settings")

application = ProtocolTypeRouter({
    'http': URLRouter([
        url(r'^events/', AuthMiddlewareStack(
            URLRouter(django_eventstream.routing.urlpatterns)
        ), { 'channels': ['test'] }),
        url(r'', get_asgi_application()),
    ]),
})

Then set ASGI_APPLICATION in your settings.py file to your project's asgi module:

ASGI_APPLICATION = 'your_project.asgi.application'

For more information about setting up Channels in general, see the Channels Documentation.

That's it! If you run python manage.py runserver, clients will be able to connect to the /events/ endpoint and get a stream.

To send data to clients, call send_event:

from django_eventstream import send_event

send_event('test', 'message', {'text': 'hello world'})

The first argument is the channel to send on, the second is the event type, and the third is the event data. The data will be JSON-encoded using DjangoJSONEncoder.

Deploying with Channels

After following the instructions in the previous section, you'll be able to develop and run locally using runserver. However, you should not use runserver when deploying, and instead launch an ASGI server such as Daphne, e.g.:

daphne your_project.asgi:application

See the Channels Documentation for information about deployment.

Multiple instances and scaling

If you need to run multiple instances of your Django project for high availability, or need to push data from management commands, or need to be able to scale to a large number of connections, you can introduce a GRIP proxy layer (such as Pushpin or Fanout Cloud) into your architecture.

In your settings.py, set GRIP_URL with your proxy settings:

GRIP_URL = 'http://api.fanout.io/realm/your-realm?iss=your-realm&key=base64:your-realm-key'

Then configure the proxy to forward traffic to your project. E.g. with Fanout Cloud, set the host:port of your deployed project as your realm's Origin Server, and have clients connect to your realm's domain.

Setup without Channels

It is possible to use this library with a GRIP proxy only, without setting up Channels.

First, install this module:

pip install django-eventstream

A few changes need to be made to settings.py.

Add the django_eventstream app:

INSTALLED_APPS = [
    ...
    'django_eventstream',
]

Add the GripMiddleware:

MIDDLEWARE = [
    'django_grip.GripMiddleware',
    ...
]

Set GRIP_URL with your Pushpin or Fanout Cloud settings:

# pushpin
GRIP_URL = 'http://localhost:5561'

# fanout cloud
GRIP_URL = 'http://api.fanout.io/realm/your-realm?iss=your-realm&key=base64:your-realm-key'

Add an endpoint in urls.py:

from django.conf.urls import url, include
import django_eventstream

urlpatterns = [
    ...
    url(r'^events/', include(django_eventstream.urls), {'channels': ['test']}),
]

That's it! Clients can now connect to the /events/ endpoint through the proxy and get a stream.

To send data to clients, call send_event:

from django_eventstream import send_event

send_event('test', 'message', {'text': 'hello world'})

The first argument is the channel to send on, the second is the event type, and the third is the event data. The data will be JSON-encoded using DjangoJSONEncoder.

Local development without Channels

If you're developing locally without Channels and want to test with Fanout Cloud, we recommend using ngrok to register a public host that routes to your local instance.

As a convenience, this module comes with a Django command runserver_ngrok that acts like runserver except it additionally configures your Fanout Cloud realm to use a detected tunnel as the origin server.

From a separate shell, run ngrok:

ngrok http 8000

Then run the runserver_ngrok command:

python manage.py runserver_ngrok

You should see output like this:

Setting ngrok tunnel 4f91f84e.ngrok.io as GRIP origin
...
Starting development server at http://127.0.0.1:8000/
Quit the server with CONTROL-C.

Note that it may take a minute or so for the changes to take effect.

Now if you make client requests to your realm's domain (e.g. {realm-id}.fanoutcdn.com) they will be routed to your local instance.

Event storage

By default, events aren't persisted anywhere, so if clients get disconnected or if your server fails to send data, then clients can miss messages. For reliable delivery, you'll want to enable event storage.

First, set up the database tables:

python manage.py migrate

Then, set a storage class in settings.py:

EVENTSTREAM_STORAGE_CLASS = 'django_eventstream.storage.DjangoModelStorage'

That's all you need to do. When storage is enabled, events are written to the database before they are published, and they persist for 24 hours. If clients get disconnected, intermediate proxies go down, or your own server goes down or crashes at any time, even mid-publish, the stream will automatically be repaired.

To enable storage selectively by channel, implement a channel manager and override is_channel_reliable.

Receiving in the browser

Include client libraries on the frontend:

<script src="{% static 'django_eventstream/eventsource.min.js' %}"></script>
<script src="{% static 'django_eventstream/reconnecting-eventsource.js' %}"></script>

Listen for data:

var es = new ReconnectingEventSource('/events/');

es.addEventListener('message', function (e) {
    console.log(e.data);
}, false);

es.addEventListener('stream-reset', function (e) {
    // ... client fell behind, reinitialize ...
}, false);

Authorization

Declare a channel manager class with your authorization logic:

from django_eventstream.channelmanager import DefaultChannelManager

class MyChannelManager(DefaultChannelManager):
    def can_read_channel(self, user, channel):
        # require auth for prefixed channels
        if channel.startswith('_') and user is None:
            return False
        return True

Configure settings.py to use it:

EVENTSTREAM_CHANNELMANAGER_CLASS = 'myapp.channelmanager.MyChannelManager'

Whenever permissions change, call channel_permission_changed. This will cause clients to be disconnected if they lost permission to the channel.

from django_eventstream import channel_permission_changed

channel_permission_changed(user, '_mychannel')

Note: OAuth may not work with the AuthMiddlewareStack from Django Channels. See this token middleware.

Routes and channel selection

The channels the client listens to are specified using Django view keyword arguments on the routes. Alternatively, if no keyword arguments are specified, then the client can select the channels on its own by providing one or more channel query parameters in the HTTP request.

Examples:

# specify fixed list of channels
url(r'^foo/events/', include(django_eventstream.urls), {'channels': ['foo']})

# specify a list of dynamic channels using formatting based on view keywords
url(r'^objects/(?P<obj_id>\w+)/events/', include(django_eventstream.urls),
    {'format-channels': ['object-{obj_id}']})

# client selects a single channel using a path component
url(r'^events/(?P<channel>\w+)/', include(django_eventstream.urls))

# client selects one or more channels using query parameters
url(r'^events/', include(django_eventstream.urls))

Note that if view keywords or a channel path component are used, the client cannot use query parameters to select channels.

If even more advanced channel mapping is needed, implement a channel manager and override get_channels_for_request.

Cross-Origin Resource Sharing (CORS) Headers

There are settings available to set response headers Access-Control-Allow-Origin, Access-Control-Allow-Credentials, and Access-Control-Allow-Headers, which are EVENTSTREAM_ALLOW_ORIGIN, EVENTSTREAM_ALLOW_CREDENTIALS, and EVENTSTREAM_ALLOW_HEADERS, respectively.