Subscribe/Publish style event system for Python 3.x
This library uses a simple decorator system to subscribe bound methods to a given event, which can be published to with a given set of args or kwargs, which are then passed off to the methods subscribed to the given event. The only requirement for a class to be viable for method subscription, is for it to inherit and initialize attributes from TrackRefs
. If you override __init__
, you must call super().__init__
, or call TrackRefs.__init__
.
Note: for dataclasses, one can always call TrackRefs.__init__
in __post_init__
Example: A bound method subscribes to a given event, determined by its name.
from subpub import Events, TrackRefs
class Adder(TrackRefs):
@Events.subscribe()
def on_nums_event(self, x: int, y: int, *, z: int) -> None:
print("Adder:", x + y + z)
class Subtractor(TrackRefs):
@Events.subscribe()
def on_nums_event(self, x: int, y: int, *, z: int) -> None:
print("Subtractor:", x - y - z)
The two given above subscribe to an event type nums_event
. When Events.publish
is called with the passed in event
argument being a string equal to "nums_event"
, it will call these methods from every instance of a class that has a subscribed method to that particular event.
Thus:
>>> add = Adder()
>>> Events.publish("nums_event", 10, 20, z=30)
Adder: 60
>>> sub = Subtractor()
>>> Events.publish("nums_event", 40, 50, z=60)
Adder: 150
Subtractor: -70
The main purpose of this library was for systematic calls to bound methods written with side effects in mind. This can be easily achieved with this library.
Absolutely. Be sure to leave Events.subscribe
as the last decorator, and wrapping a subscribed method, eg. with functools.cache
, works just fine.
class C(TrackRefs):
@functools.cache
@Events.subscribe()
def on_foo(self, spam):
return self.bar + spam.egg
This example does conflict with the core concept of the library to provide support for side effect methods, but these methods can still be called, regardless of whether or not its event has been published to. Also, it is worth noting that decorators just as easily provide new behaviours to methods, besides manipulating its return value (eg. loggers).
You can always contact me on Discord, or create an issue on the repository for something you think should be fixed or added. This library is still very early in development. Any criticism or ideas are greatly welcomed.