Easy to use class mixins for Django Rest Framework and Pusher to keep your API realtime.

Download from PyPI:

pip install drf_model_pusher

In your settings set your Pusher App Id and credentials, the cluster is optional

PUSHER_APP_ID=""  
PUSHER_KEY=""  
PUSHER_SECRET=""
PUSHER_CLUSTER=""  

Add drf_model_pusher to your INSTALLED_APPS:

INSTALLED_APPS = [
    "...",
    "drf_model_pusher",
]

Define some PusherBackends for your models and serializers in a pusher_backends.py file. The PusherBackend class just needs to define a serializer_class attribute which inherits from ModelSerializer.

# example/pusher_backends.py

from django.db.models import Model
from rest_framework.serializers import ModelSerializer
from drf_model_pusher.backends import PusherBackend, PrivatePusherBackend

class MyModel(Model):
    pass

class MyModelSerializer(ModelSerializer):
    class Meta:
        model = MyModel

class MyModelPrivateSerializer(ModelSerializer):
    class Meta:
        model = MyModel

class MyModelPusherBackend(PusherBackend):
    serializer_class = MyModelSerializer
    
class MyModelPrivatePusherBackend(PrivatePusherBackend):
    serializer_class = MyModelPrivateSerializer

Add the ModelPusherViewMixin mixin class to your views and define a get_pusher_channels method which should return a list of strings to use as channels.

# example/views.py

from rest_framework.viewsets import ModelViewSet
from drf_model_pusher.views import ModelPusherViewMixin
from example.serializers import MyModelSerializer

class MyModelViewSet(ModelPusherViewMixin, ModelViewSet):
    serializer_class = MyModelSerializer
    
    def get_pusher_channels(self):
        return ["<channel_id>"]

If you want to ignore the current connection when sending messages you should set a x-pusher-socket-id header on your requests. This may be useful if you're modifying resources and receiving the results in a response, you may not want the current connection to listen on these events to prevent duplicating content.

The PusherBackend.push_change method accepts an ignore boolean keyword argument which can toggle whether the pusher socket id is used, it defaults to True, so including the pusher socket id in the request will ignore the current connection for all pusher events sent by default.

• DRF_MODEL_PUSHER_BACKENDS_FILE (default: pusher_backends.py) - The file in your applications to import PusherBackends.
• DRF_MODEL_PUSHER_DISABLED (default: False) - Determines whether or not to trigger Pusher events.
• DRF_MODEL_PUSHER_WEBHOOK_OPTIMISATION_ENABLED (default: False) - Determines whether or not to check if the channel is occupied before sending an event. See Occupied Channels Optimisation.

If you have followed the above steps correctly and your backends are not registering, your app config may not be running it's ready method. To force this, in your apps __init__.py add the line default_app_config = 'myapp.apps.MyAppConfig'

Be aware of any pusher limits and consult their documentation at https://pusher.com/docs for some common questions.

Pusher has a 10kb default size limit on messages, this can be increased to 256kb by contacting support.

If you want to extend PusherBackend or PrivatePusherBackend rather than declaring a new concrete backend, you need to make sure the class is abstract. For example your new base class would be similar to this:

class MyPusherBackend(PusherBackend):
    class Meta:
        abstract = True

    # Override whatever methods here


class MyModelBackend(MyPusherBackend):
    class Meta:
        model = MyModel

This optimisation store channel occupation state locally through Django's cache and only send events to channels that are occupied. A view class (ChannelExistenceWebhook) is provided to act as a webhook to receive updates from Pusher as to whether a channel is still occupied or has been vacated. If the package hasn't seen a channel before it will re-sync the cache with Pusher to make sure we have the most up to date view of channel occupations and acts as a warm up for the cache on the first event. Note that Pusher occasionally delays channel_vacated events to combat network interruptions.

You can enabled this feature by setting DRF_MODEL_PUSHER_WEBHOOK_OPTIMISATION_ENABLED = True in your Django settings. You must also have a Django cache set up and create a route for Pusher to send webhook events to:

# urls.py
from drf_model_pusher.views import ChannelExistenceWebhook

urlpatterns = [
    url(r"^pusher/channel-existence/$", ChannelExistenceWebhook.as_view(), name="pusher-channel-existence-webhook"),
]

It's early days, but if you'd like to report any issues or work on an improvement then please check for any similar existing issues before you report them.