Welcome to Messagio

It's sort of like pubsub, but evolved.

Messagio works out of the box with Django and Celery.

You can define events, aka messagios, with specific payloads, the messagios themselves are preferably python dataclasses, and their payloads are pickled, but you can in theory use any class as a messagio.

Now you can subscribe any function to listen for those events and do stuff when an event is received.

Any number of functions can subscribe to a single event. Every function can subscribe to one or more events. The events are transmitted using celery, so make sure you configure

Celery Config

As long as celery is configured in your django setup, you don't need to do anything. However, it might be wise to have a custom queue for your messagio events, to make sure they get the priority they deserve.

In django you can configure it as such in django settings.py or your celery settings:

CELERY_TASK_ROUTES = (
    [
        ('messagio.tasks.*', {'queue': 'messagio'}),
    ],
)

Don't forget to also set up the queues, for example:

CELERY_TASK_QUEUES = (
    Queue('celery', Exchange('celery'), routing_key='celery'),
    Queue("messagio", Exchange("messagio"), routing_key="messagio"),
)

And finally, you need to ensure that pickle is supported by your Celery config.

CELERY_ACCEPT_CONTENT = ["pickle"]

Keep in mind that if you are using json for pickling, you need to also add json to the array above.

Defining Messagios

A messagio is a class that can be pickled, and extends from the Messagio class. The best way to create messagios is with dataclasses, so just decorate your class with @dataclass. Your messagio class needs to extend Messagio, the @dataclass decorator is optional however.

Here are two ways to define messagios:

@dataclass
class PerformAction(Messagio):
   action_name:str
   obj:any

class ActionPerformed(Messagio):
   def __init__(action_name:str,obj:any):
       self.action_name=action_name
       self.obj = obj

Autodiscover the messagios and project structure

In order for your decorated messagio listeners to be respected, you must make sure they are imported at runtime.

The simplest way is to use the autodiscover_listeners function.

For autodiscovery to work your messagio definitions need to follow a certain structure.

Place your listener functions in any of the below files or folder structures.

- messagio.py

- tasks.py

- messagio
  - __init__.py

Then use the following code somewhere in your project that is always imported, for Django projects this can be in the main urls.py file.

from messagio import autodiscover_listeners
autodiscover_listeners()

Publishing messagios

Publishing messagios is simple, just fire it.

obj = Model.objects.get(pk=123)
PerformAction("foo_action", obj).fire()

Fire accepts certain additional parameters:

• sync:bool will call the messagio synchronously without going through celery (uses celery.task.apply)

Listening to messagios

Any function can be configured as a listener. A listener can listen to one or more messagio types and will receive the messagio object itself as the first and only parameter.

The subsciption can come with some extra arguments, these are all passed directly to celery.

• priority: TASK_PRIORITY set a priority that will be passed to celery, if your celery does not use priorities, this is ignored. The priorities are integers and depend on your celery configuration, some default options are available in const.py
• autoretry_for: tuple[Exception] a list of exceptions that trigger an auto-retry, again passed to celery
• max_retries: int number of times to auto retry if auto retry is enabled, default is no auto retry
• default_retry_delay: float how long to wait before auto retrying

# high priority listener
@listen_to_message(PerformAction, priority=10)
def high_prio_listener(messagio:PerformAction):
    # do something with your messagio
    pass

@listen_to_message(PerformAction, ActionPerformed)
def listener_of_many(messagio:typing.Union[PerformAction, ActionPerformed]):
    # do something with your messagio
    pass

# low priority listener
@listen_to_message(PerformAction, priority=1)
def log_me(messagio:PerformAction):
    logging.getLogger("foo").info("Action was performed %s" % messagio.action_name)

Messagio functions will be executed by your celery worker and they are executed sequentially in the order that they were subscribed. A single published messagio will be executed by a single worker and a messagio can be published as many times as you want.

You can also subscribe directly using the message center

def any_func(msg:PerformAction):
    #do stuff
    pass

MessageCenter.singleton().subscribe(PerformAction, any_function)

Abstract message center?

If you want, you can roll your own message center that does not use celery to transport messages, for example if you want to use a daemon or some other transport protocol like huey.