- Get better runtime type validation
- Evaluate moving to pydantic model instead of dataclass for Event
- What about avro, or protobuf?
In [9]: import datetime
In [10]: from dataclasses import dataclass, asdict
In [11]: @dataclass
...: class Event:
...: data: str
In [12]: bus.register_event("airbase.event_store.event", Event)
In [13]: event = Event(data=datetime.datetime.now())
In [14]: bus.send(event)
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
One side-effect of using dataclass in the eventbus is dataclass values aren't type validated during creation. As result the failure happens at the lowest layer while converting to JSON. This becomes painful to debug when there one of the field is nested dictionary and can't be converted to JSON by json.dumps.
If we use pydantic the type check will kick in during object creation and can prevent the error happening at lower level.
