df_websockets

df_websockets

df_websockets extends django-channels to simplify communications between clients and servers and to process heavy tasks in background processes.

df_websockets is based on two main concepts:

• signals, that are functions triggered both on the server or the browser window by either the server or the client,
• topics to allow the server to send signals to any group of browser windows.

Signals are exchanged between the browser window and the server using a single websocket. Signals that are triggered by the browser on the server are processed as background tasks (so the websocket endpoint does almost nothing). Signals that are triggered by the server can be processed as background tasks on the serveur and as Javascript functions on the browser.

Background processes can use celery, channels workers, or simply different processes or threads.

Requirements

df_websockets works with:

• Python >= 3.6,
• django >= 2.0,
• channels >= 2.0,
• daphne >= 2.0.

For production use or any multiprocess setup (even in development mode), you also need:

• redis >= 5.0,
• channels_redis >= 3.3,
• daphne >= 2.0.

If you want to process signals in Celery tasks rather in Channel workers, you need to setup a Celery infrastructure: Celery setup.

Installation

python -m pip install df_websockets

In your settings, you must add the following values:

# the ASGI application to use with gunicorn or daphne
ASGI_APPLICATION = "df_websockets.routing.application"
# add the required Middleware
MIDDLEWARES = [..., "df_websockets.middleware.WebsocketMiddleware", ...]
INSTALLED_APPS = [..., "channels", "daphne", "df_websockets", ...]
WEBSOCKET_WORKERS = "thread"
# a channel layer, required by channels_redis
CHANNEL_LAYERS = {
    'default': {
        'BACKEND': 'channels_redis.core.RedisChannelLayer',
        'CONFIG': {
            "hosts": [('localhost', 6379)],
        },
    },
}

If you use df_config and you use a local Redis, you have nothing to do: settings are automatically set and everything is working as soon as a Redis is running on your machine.

Now, include js/df_websockets.min.js in your HTML and call df_websockets.tasks.set_websocket_topics(request) somewhere in the Django view. A bidirectionnal websocket connection will be established in your page.

You can start the development server:

python manage.py runserver

daphne must be separately installed and added to INSTALLED_APPS with channels>=4.0

If you use Channels workers (WEBSOCKET_WORKERS = "channels"), you also need to start a Channel worker:

python manage.py run_worker celery

If you use Celery (WEBSOCKET_WORKERS = "celery"), you also need to start a Celery worker:

python manage.py worker -Q celery

basic usage

A signal is a string attached to Python or Javascript functions. When this signal is triggered, all these functions are called. Of course, you can target the platforms on which the functions will be executed: the server (for Python code) or chosen browser windows.

First, we connect our code to the signal "myproject.first_signal".

from df_websockets.decorators import everyone, signal
import time

@signal(path="myproject.first_signal", is_allowed_to=everyone, queue="celery")
def my_first_signal(window_info, content=None):
    print(content)

@signal(path="myproject.first_signal", is_allowed_to=everyone, queue="slow")
def my_first_signal_slow(window_info, content=None):
    time.sleep(100)
    print(content)

/* static file "js/df_websockets.min.js" must be included first */
document.addEventListener("DOMContentLoaded", () => {
    window.DFSignals.connect('myproject.first_signal', (opts) => {
        console.warn(opts.content);
    });
});

Now, we can trigger this signal to call this functions. In both cases, both functions will be called on the server and in the browser window.

window.DFSignals.call('myproject.first_signal', {content: "Hello from browser"});

from df_websockets.tasks import WINDOW, trigger, SERVER
from df_websockets.decorators import everyone, signal
from django.http.response import HttpResponse

def any_view(request):  # this is a standard Django view
    trigger(request, 'myproject.first_signal', to=[SERVER, WINDOW], content="hello from a view")
    return HttpResponse()

@signal(path="myproject.second_signal", is_allowed_to=everyone, queue="slow")
def second_signal(window_info):
    trigger(window_info, 'myproject.first_signal', to=[SERVER, WINDOW], content="hello from Celery")

In this case, the to parameter targets both the server and the window. You can even open a shell and call df_websockets.tasks.trigger(None, 'myproject.first_signal', to=[BROADCAST], content="hello from a shell"). All open windows will react.

Topics

You can select the set of connected browser windows that receive a signal triggered by the server, in addition of processing this signal on the server.

A Django view using this signal system must call set_websocket_topics to add some ”topics” to this view. When you trigger a signal on the server, you can target any set topic. All windows featuring this topic will receive this signal.

For example, assume that multiple clients open a specific article on a blog. At any time, you can open Python shell in a terminal and trigger a signal on all these windows.

from df_websockets.tasks import set_websocket_topics
from django.contrib.auth.models import Group
from django.template.response import TemplateResponse

def any_view(request):  # this is a standard Django view
    # useful code
    obj1 = Group.objects.get(id=42)
    set_websocket_topics(request, [obj1])
    return TemplateResponse("my/template.html", {})  # do not forget to add `js/df_websockets.min.js` to this HTML

obj1 must be a Python object that is handled by the WEBSOCKET_TOPIC_SERIALIZER function. By default, any string and Django models are valid. Each window also has a unique identifier that is automatically added to this list, as well as the connected user id and the BROADCAST.

The following code will call the JS function on every browser window having the obj topic and to the displayed window.

from df_websockets.tasks import WINDOW, trigger
from df_websockets.tasks import set_websocket_topics
from django.contrib.auth.models import Group
from django.http.response import HttpResponse
def another_view(request, obj_id):
    obj = Group.objects.get(id=42)
    trigger(request, 'myproject.first_signal', to=[WINDOW, obj], content="hello from a view")
    set_websocket_topics(request, ["other topics"])
    return HttpResponse()

There are three special values:

• df_websockets.tasks.WINDOW: the original browser window,
• df_websockets.tasks.USER: all windows currently displayed by the connected user,
• df_websockets.tasks.BROADCAST: all active windows.

Some information about the original window (like its unique identifier or the connected user) must be provided to the triggered Python code, allowing it to trigger JS events on any selected window.
These data are stored in the WindowInfo object, automatically built from the HTTP request by the trigger function and provided as first argument to the triggered code. The trigger function accepts WindowInfo or HTTPRequest objects as first argument.

settings

There are a few settings:

• WEBSOCKET_WORKERS: one of "celery" (use Celery tasks), "channels" (use Channels workers), "multithread" (process signals in threads), "multiprocess" (process signals in new processes). The first two choices require at least one valid worker.
• WEBSOCKET_DEFAULT_QUEUE the default queue for signals ("celery" by default)

Other settings are:

• WEBSOCKET_CACHE_EXPIRE: the validity of the association between a websocket connection and the associated topics
• WEBSOCKET_CACHE_PREFIX: prefix of keys used to cache data
• WEBSOCKET_CACHE_BACKEND: the cache backend (default by default) — you cannot use LocMemCache with Celery or Channels workers, nor DummyCache with any kind of workers
• WEBSOCKET_SIGNAL_ENCODER: the JSON encoder to encode signal arguments
• WEBSOCKET_SIGNAL_DECODER: the JSON decoder to decode signal arguments
• WEBSOCKET_TOPIC_SERIALIZER: the function used to transform Python topics into valid topic names
• WEBSOCKET_POOL_SIZES: a dict associating a queue name to a number of threads (or processes)
• WINDOW_INFO_MIDDLEWARES: a list of middlewares for transforming a HttpRequest to a WindowInfo
• WEBSOCKET_URL: URL prefix (/ws/ by default)
• ASGI_APPLICATION: the ASGI application

Cache backends

Task data are passed by the server process to the workers using the Django cache infrastructure. For obvious reasons, you cannot use DummyCache nor LocMemCache with Celery or Channels workers, since these caching methods are not shared accross processes. So, you need either to use a shared cache backend as default backend, or dedicate a cache backend to websockets.

First case needs to update your settings.py file:

CACHES = {
    "default": {
        "BACKEND": "django.core.cache.backends.redis.RedisCache",
        "LOCATION": "redis://127.0.0.1:6379",
    }
}

Second case, still in your settings.py file:

CACHES = {
    "default": {
        "BACKEND": "django.core.cache.backends.locmem.LocMemCache",
        "LOCATION": "unique-snowflake",
    },
    "websockets": {
        "BACKEND": "django.core.cache.backends.redis.RedisCache",
        "LOCATION": "redis://127.0.0.1:6379",
    },
}
WEBSOCKET_CACHE_BACKEND = "websockets"

HTML forms

df_websockets comes with some helper functions when you signals to be trigger on the server when a form is submitted or changed. Assuming that you have a signals.py file that contains:

from df_websockets.decorators import signal
from df_websockets.tasks import WINDOW, trigger
from df_websockets.utils import SerializedForm
from django import forms


class MyForm(forms.Form): 
    title = forms.CharField()

@signal(path='signal.name')
def my_signal_function(window_info, form_data: SerializedForm(MyForm)=None, title=None, id=None):
    print(form_data and form_data.is_valid())
    trigger(window_info, 'myproject.first_signal', to=WINDOW, title=title)

@signal(path='signal.name')
def my_signal_function_raw(window_info, form_data=None, title=None, id=None):
    print(form_data and form_data.is_valid())
    trigger(window_info, 'myproject.first_signal', to=WINDOW, title=title)

Using on a HTML form:

<form data-df-signal='[{"name": "signal.name", "on": "change", "form": "form_data", "opts": {"id": 42} }]'>
    <input type="text" name="title" value="df_websockets">
</form>

or, using the Django templating system:

{% load websockets %}
<form {% js_call "signal.name" on="change" form="form_data" id=42 %}>
    <input type="text" name="title" value="df_websockets">
</form>

When the field "title" is modified, my_signal_function(window_info, form_data = [{"name": "title", "value": "df_websockets"}], id=43) is called.

Using on a HTML form input field:

<form>
    <input type="text" name="title" data-df-signal='[{"name": "signal.name", "on": "change", "value": "title", "opts": {"id": 42} }]'>
</form>

or, using the Django templating system:

{% load websockets %}
<form>
    <input type="text" name="title" {% js_call "signal.name" on="change" value="title" id=42 %}>
</form>

When the field "title" is modified, my_signal_function(window_info, title="new title value", id=43) is called.

Testing signals

In production, the signal framework requires a working Redis and worker processes. However, if you only want to check if a signal has been called in unitary tests, you can use :class:df_websockets.utils.SignalQueue. Both server-side and client-side signals are kept into memory:

• df_websockets.testing.SignalQueue.ws_signals,

  • keys are the serialized topics
  • values are lists of tuples (signal name, arguments as dict)

• df_websockets.testing.SignalQueue.python_signals

  • keys are the name of the queue

  • values are lists of (signal_name, window_info_dict, kwargs=None, from_client=False, serialized_client_topics=None, to_server=False, queue=None)

    • signal_name is … the name of the signal
    • window_info_dict is a WindowInfo serialized as a dict,
    • kwargs is a dict representing the signal arguments,
    • from_client is True if this signal has been emitted by a web browser,
    • serialized_client_topics is not None if this signal must be re-emitted to some client topics,
    • to_server is True if this signal must be processed server-side,
    • queue is the name of the selected Celery queue.

from df_websockets.tasks import trigger, SERVER
from df_websockets.window_info import WindowInfo
from df_websockets.testing import SignalQueue

from df_websockets.decorators import signal
# noinspection PyUnusedLocal
@signal(path='test.signal', queue='demo-queue')
def test_signal(window_info, value=None):
  print(value)

wi = WindowInfo()
with SignalQueue() as fd:
  trigger(wi, 'test.signal1', to=[SERVER, 1], value="value1")
  trigger(wi, 'test.signal2', to=[SERVER, 1], value="value2")

# fd.python_signals looks like {'demo-queue': [ ['test.signal1', {…}, {'value': 'value1'}, False, None, True, None], 
# # ['test.signal2', {…}, {'value': 'value2'}, False, None, True, None]]}
# fd.ws_signals looks like {'-int.1': [('test.signal1', {'value': 'value1'}), ('test.signal2', {'value': 'value2'})]}

JavaScript signals

Many JS signals are available out-of-the-box. These signals can be triggered either by the JS code or by the Python code.

Signals must be defined in a Python file that is imported during Django's startup, or in any signals.py file inside a Django app (like models.py). For example, you can update the content of a HTML node with the following lines:

from df_websockets.tasks import trigger, WINDOW
from df_websockets.decorators import signal

@signal(path='test.signal', queue='demo-queue')
def test_signal(window_info, word="hellow"):
    trigger(window_info, 'html.content', to=WINDOW, selector="#obj", content= "<span>%s</span>" % word)

window.DFSignals.call('html.content', {selector: "#obj", content: "<span>hello</span>"});

Please read the content of npm/df_websockets/base.js for the whole list of available signals. You can also create some shortcuts for the most common signals. Another way is to run the demo:

python demo_manage.py runserver

Checklist

Everything must be correctly setup to have working signals.

The first step is to test tasks from the command-line:

1. Redis is running and accepting connections
2. Celery is working
3. at least one worker is running with all required queues
4. the triggered signal is exists
5. open a console python manage.py shell and manually trigger a task

from df_websockets.tasks import trigger, SERVER
from df_websockets.window_info import WindowInfo
trigger(WindowInfo(), 'test.signal', to=[SERVER], value="value2")

The second step is to check the web part:

1. the web server must be running and accepting connections
2. Celery is working
3. at least one worker is running with all required queues
4. the triggered signal exists
5. df_websockets.middleware.WebsocketMiddleware is included
6. check the used domain name, since tokens are passed through cookies: "localhost" is different than "127.0.0.1"
7. df_websockets.tasks.set_websocket_topics is used somewhere in the view
8. static/js/df_websockets.min.js is included in the page
9. check if the WS tries to connect
10. check if the WS is connected
11. open a console python manage.py shell and manually trigger a task

from df_websockets.tasks import trigger, BROADCAST
from df_websockets.window_info import WindowInfo
trigger(WindowInfo(), 'html.text', to=[BROADCAST], selector="body", content= "<span>hello</span>")

Do not hesitate to use a verbose logging:

LOGGING = {
    "version": 1,
    "disable_existing_loggers": True,
    "formatters": {
        "verbose": {
            "format": (
                "%(asctime)s [%(process)d] [%(levelname)s] "
                + "pathname=%(pathname)s lineno=%(lineno)s "
                + "funcname=%(funcName)s %(message)s"
            ),
            "datefmt": "%Y-%m-%d %H:%M:%S",
        },
        "django.server": {
            "()": "django.utils.log.ServerFormatter",
        },
        "nocolor": {
            "()": "logging.Formatter",
            "fmt": "%(asctime)s [%(name)s] [%(levelname)s] %(message)s",
            "datefmt": "%Y-%m-%d %H:%M:%S",
        },
    },
    "filters": {
    },
    "handlers": {
        "stdout.info": {
            "class": "logging.StreamHandler",
            "level": "DEBUG",
            "stream": "ext://sys.stdout",
            "formatter": "verbose",
        },
        "stderr.debug.django.server": {
            "class": "logging.StreamHandler",
            "level": "DEBUG",
            "stream": "ext://sys.stderr",
            "formatter": "django.server",
        },
    },
    "loggers": {
        "django": {"handlers": [], "level": "INFO", "propagate": True},
        "django.db": {"handlers": [], "level": "INFO", "propagate": True},
        "django.db.backends": {"handlers": [], "level": "INFO", "propagate": True},
        "django.request": {"handlers": [], "level": "DEBUG", "propagate": True},
        "django.security": {"handlers": [], "level": "INFO", "propagate": True},
        "df_websockets.signals": {"handlers": [], "level": "DEBUG", "propagate": True},
        "gunicorn.error": {"handlers": [], "level": "DEBUG", "propagate": True},
        "pip.vcs": {"handlers": [], "level": "INFO", "propagate": True},
        "py.warnings": {
            "handlers": [],
            "level": "INFO",
            "propagate": True,
        },
        "daphne": {"handlers": [], "level": "INFO", "propagate": True},
        "mail.log": {"handlers": [], "level": "INFO", "propagate": True},
        "aiohttp.access": {
            "handlers": ["stderr.debug.django.server"],
            "level": "INFO",
            "propagate": False,
        },
        "django.server": {
            "handlers": ["stderr.debug.django.server"],
            "level": "INFO",
            "propagate": False,
        },
        "django.channels.server": {
            "handlers": ["stderr.debug.django.server"],
            "level": "INFO",
            "propagate": False,
        },
        "gunicorn.access": {
            "handlers": ["stderr.debug.django.server"],
            "level": "INFO",
            "propagate": False,
        },
    },
    "root": {"handlers": ["stdout.info"], "level": "DEBUG"},
}