Simple asynchronous messages for django. Plays nicely with Celery.
What problem does this solve?
Suppose a user instigates an expensive task that you are processing offline (eg using Celery). This library provides a simple mechanism for notifying the user when the task is finished, utilising Django's messaging framework.
What's an example?
You might use Celery to generate a large report and then employ this library to notify the user that their report is ready to be downloaded. The user will see the message you've sent them when they make their next request after the message is submitted.
How does it work?
A cache is used to persist the messages, and middleware is used to pick these up and submit them to Django's messaging framework. All very simple.
Aren't there other libraries that do this?
Yes, there are - but they solve the problem in different ways:
- django-offline-messages - this provides an alternative storage backend that Django's messages framework can use.
What's good about this implementation?
- It's simple, fast and easy to use.
- It works cohesively with existing Django cache and messages framework. It will work no matter what cache backend your are using, and whatever storage backend is used for messages.
- A user may miss the message if they navigating quickly between pages. But this is a general problem of the Django messages framework.
From PyPI (stable):
pip install django-async-messages
From Github (unstable):
pip install git+git://github.com/codeinthehole/django-async-messages#egg=djang-async-messages
Add 'async_messages.middleware.AsyncMiddleware' to your MIDDLEWARE_CLASSES. Ensure it comes after 'django.contrib.messages.middleware.MessageMiddleware'.
You need to have CACHES configured in you settings for this to work. As usual, memcache is the best choice. Note that local memory caching is not suitable as each process has its own private cache and a Celery task can't communicate with the webserver process cache.
Send a message to a single user:
>>> from async_messages import message_user >>> from django.contrib.auth.models import User >>> barry = User.objects.get(username='barry') >>> message_user(barry, "Barry, your report is ready")
Send a message to lots of users:
>>> from async_messages import message_users >>> staff = User.objects.filter(is_staff=True) >>> message_users(staff, "All budgets must be spent by the end of the day")
Specify message level:
>>> from django.contrib.messages import constants >>> message_users(staff, "Boom!", constants.WARNING)
Send multiple messages to a single user:
>>> from async_messages import message_user >>> from django.contrib.auth.models import User >>> barry = User.objects.get(username='barry') >>> message_user(barry, "Barry, your report is queued up for processing") >>> # do more awesome stuff >>> message_user(barry, "Barry, your report is ready")
Alternative way to send a message to a single user, imitating the django.contrib.messages API:
>>> from async_messages import messages >>> barry = User.objects.get(username='barry') >>> messages.debug(barry, "Barry was here") >>> messages.info(barry, "Hi, Barry") >>> messages.success(barry, "Barry, your report is ready") >>> messages.warning(barry, "Barry, you didn't lock your session") >>> messages.error(barry, "You are not Barry")
Fork, clone and create a virtualenv. Then run:
Run tests with:
Please submit pull requests using 'develop' as the target branch.
- Fix bug around request instances that don't have a user attribute
- Mimic django.contrib.messages API for sending a message to a user
- Added possibility to queue multiple messages
- Altered dependency on Django to be only 1.2+
- Altered middleware to use process_response.
- Better docstrings
- Minimum viable product