🚋Streaming

One strong advantage of using ROAR_PY is that it supports streaming standard RoarPySensor / RoarPyActor / RoarPyWorld objects over websocket or any other protocols that you wish to implement.

To support streaming for non-official implementations of the ROAR_PY interface, you simply need to add a few annotations to your implementation code and you can spin up a ROAR_PY remote server quickly. Let's get started!

All annotation methods can be imported from roar_py_interface package

Streaming requires data to be passed between remote server and clients, and therefore usually requires the control code to support multi-threading.

ROAR_PY makes this simple by not requiring your code to support multi-threading natively, but just asking you to annotate methods that can only be called one thread at a time with @roar_py_thread_sync.

Do not worry about dead-locking because ROAR_PY's streaming service uses a reentrant lock.

To share a world with different clients, the server needs to know which actors belongs to which client, therefore you need to annotate methods that creates new actors / sensors with @roar_py_append_item and methods that removes actors / sensors with @roar_py_remove_item in any implementation of RoarPyWorld.

If you want to stream custom sensor datas, your custom data should be annotated with @serde, @dataclass, and @remote_support_sensor_data_register. Annotating with @serde and @dataclass tells the Python interpreter that the sensor data class is a data class that can be serialized, and @remote_support_sensor_data_register tells ROAR_PY's streaming service that this type of data is supported for streaming. The sensor data class should also extend RoarPyRemoteSupportedSensorData.

Refer to official sensor data types for examples on how to write streaming supported data types. You may also override to_data(...) method and create a new static method with signature from_data_custom(data: bytes, scheme: RoarPyRemoteSupportedSensorSerializationScheme) to support custom serialization flows without using the default pyserde