fanout/django-eventstream

Server-Sent Events for Django

Django EventStream

EventStream provides Django views 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: 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.
  • Reliable delivery. Events can be persisted to your database, so clients can recover if they get disconnected.
  • Per-user channel permissions.
  • Works with Django REST Framework.

Setup

First, install this module and the daphne module:

pip install django-eventstream daphne

Or, for Django REST Framework support, include the optional drf dependency:

pip install django-eventstream[drf] daphne

Add the daphne and django_eventstream apps to settings.py:

INSTALLED_APPS = [
    ...
    "daphne",
    "django_eventstream",
]

Add an endpoint in urls.py:

from django.urls import path, include
import django_eventstream

urlpatterns = [
    ...
    path("events/", include(django_eventstream.urls), {"channels": ["test"]}),
]

For DRF views, see Views with Django REST Framework

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

Sending events

To send data to clients, call send_event:

from django_eventstream import send_event

# send_event(<channel>, <event_type>, <event_data>)
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.

Note: In a basic setup, send_event must be called from within the server process (e.g. called from a view). It won't work if called from a separate process, such as from the shell or a management command. To send events from separate processes, see Multiple instances and scaling.

Deploying

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 How to deploy with ASGI.

WSGI mode can work too, but only in combination with a GRIP proxy. See Multiple instances and scaling.

Multiple instances and scaling

If you need to run multiple instances of your Django app for high availability or scalability, or need to send events from management commands, then you can introduce a GRIP proxy such as Pushpin or Fastly Fanout into your architecture. Alternatively, you could use Redis or another message queue to communicate between processes, and call send_event from each one. Otherwise, events originating from an instance will only be delivered to clients connected to that instance.

For example, to use Pushpin with your app, you need to do three things:

  1. In your settings.py, add the GripMiddleware and set GRIP_URL to reference Pushpin's private control port:
MIDDLEWARE = [
    "django_grip.GripMiddleware",
    ...
]
GRIP_URL = 'http://localhost:5561'

The middleware is part of django-grip, which should have been pulled in automatically as a dependency of this module.

  1. Configure Pushpin to route requests to your app, by adding something like this to Pushpin's routes file (usually /etc/pushpin/routes):
* localhost:8000 # Replace `localhost:8000` with your app's URL and port
  1. Configure your consuming clients to connect to the Pushpin port (by default this is port 7999). Pushpin will forward requests to your app and handle streaming connections on its behalf.

If you would normally use a load balancer in front of your app, it should be configured to forward requests to Pushpin instead of your app. For example, if you are using Nginx you could have configuration similar to:

location /api/ {
    proxy_pass http://localhost:7999
}

The location block above will pass all requests coming on /api/ to Pushpin.

Views with Django REST Framework

To set up views with DRF, register them on the router:

from django.urls import path, include
from django_eventstream.viewsets import EventsViewSet, configure_events_view_set

router = DefaultRouter()

# register by function
router.register(
    "events1",
    configure_events_view_set(channels=["channel1", "channel2"],
    messages_types=["message", "info"]),
    basename="events1")

# register by class
router.register(
    "events2",
    EventsViewSet(channels=["channel1", "channel2"]),
    basename="events2")

urlpatterns = [
    ...
    path("", include(router.urls)),
]

Next, use DEFAULT_RENDERER_CLASSES in settings.py to manage the renderers you want to use. The django_eventstream.renderers.SSEEventRenderer is required to enable SSE functionality. If you also want the Browsable API view, add django_eventstream.renderers.BrowsableAPIEventStreamRenderer.

Example:

REST_FRAMEWORK = {
    'DEFAULT_RENDERER_CLASSES': [
        'rest_framework.renderers.JSONRenderer',
        'rest_framework.renderers.BrowsableAPIRenderer',
        'django_eventstream.renderers.SSEEventRenderer',
        'django_eventstream.renderers.BrowsableAPIEventStreamRenderer'
         # Add other renderers as needed
    ]
}

Be careful to not add the eventstream renderers before the JSONRenderer and BrowsableAPIRenderer (or other Renderer), otherwise the API will probably not work as expected.

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
path('foo/events/', include(django_eventstream.urls), {'channels': ['foo']})

# specify a list of dynamic channels using formatting based on view keywords
path('objects/<obj_id>/events/', include(django_eventstream.urls),
    {'format-channels': ['object-{obj_id}']})

# client selects a single channel using a path component
path('events/<channel>/', include(django_eventstream.urls))

# client selects one or more channels using query parameters
path('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_ORIGINS, EVENTSTREAM_ALLOW_CREDENTIALS, and EVENTSTREAM_ALLOW_HEADERS, respectively.

Examples:

EVENTSTREAM_ALLOW_ORIGINS = ['http://example.com', 'https://example.com']
EVENTSTREAM_ALLOW_CREDENTIALS = True
EVENTSTREAM_ALLOW_HEADERS = 'Authorization'

Note that EVENTSTREAM_ALLOW_HEADERS only takes a single string value and does not process a list.

If more advanced CORS capabilities are needed, see django-cors-headers.

Stars
624
100.00% more than last month
Forks
83
Open Issues
40