Improve pipeline profiling

[juanmc2005]

Problem

The way to calculate the real-time latency of a pipeline is currently implemented in FileAudioSource. This is counter-intuitive and has a couple of problems, like including unwanted operations in the profiling (e.g. plotting or logging).

Idea

Implement this as a wrapper over an observable in a more flexible and RxPY way.

Example

class Chronometer:
    def __init__(self):
        self.current_start_time = None
        self.history = []
    
    def start(self):
        self.current_start_time = time.monotonic()
    
    def stop(self):
        end_time = time.monotonic() - self.current_start_time
        self.current_start_time = None
        self.history.append(end_time)


def profile(observable: rx.Observable, operations: List[Operator]) -> rx.Observable:
    chronometer = Chronometer()
    return observable.pipe(
        ops.do_action(lambda _: chronometer.start()),
        *operations,
        ops.do_action(
            on_next=lambda _: chronometer.stop(),
            on_completed=lambda: print(np.mean(chronometer.history))
        )
    )

[RahmaYasser]

Hello @juanmc2005 Is there a way to integrate this code with mobile app? I mean to take input from mic but in mobile app. Any hint? and thanks in advance

[juanmc2005]

Hi @RahmaYasser, thanks for your question. Would you mind creating a separate issue for this? Thanks!

[juanmc2005]

Implemented in #64