Event loops

An event loop is a frequently used pattern in concurrent programming: event loops perhaps most famously used in Javascript – and Node, Android has a MessageQueue, along with most GUI implementations; and of course, Python has Asyncio.

To paraphrase W. C. Williams, there are no ideas but in things so I'm going to build an extremely simple one in Python as a demonstration.

The core of the idea is – surprisingly enough – a loop doing pieces of work – generally referred to as messages – passed into it. Messages are generally functions, or carefully delineated parts of functions.

def run_event_loop(queue):
    while queue:
        msg = queue.pop(0)

    lambda: print("Hello"),
    lambda: print("World!")

That's cute, but not remarkably useful or better than just calling the functions directly. As a first step, I'm going to introduce a little bit of concurrency into the mix.

Consider two functions that I'd like to run simultaneously for reasons:

def first():
    print("Function 1, part 1")
    print("Function 1, part 2")
    print("Function 1, part 3")

def second():
    print("Function 2, part 1")
    print("Function 2, part 2")
    print("Function 2, part 3")

Before we dive deeply, I should remind you that concurrency allows interleaving different units of work; parallelism means running different units of work simultaneously.

One of the biggest advantages of using an event loop that I've noticed is being able to avoid all the messy synchronization you might otherwise need: the loop and all the messages on it are in the same thread. For your mental model: I'd recommend defining messages as functions, or carefully delineated parts of functions.

view source