AsyncIO In Depth

If you find Python's async / await to be mostly easy to use, yet somewhat magical and opaque I think you'll find this series of articles valuable. I spent several weeks digging into asyncio's internals and these articles are the result of my exploration.

They aren't a way to quickly become productive with asyncio: this series is the one you should read after you can write a little bit of async code, but aren't quite sure how or why things work.

This is an in-progress depth-first-traversal into the Python's event loop implementation. After several aborted attempts at writing about asyncio in 2020, I'm publishing this as a series in pieces instead of trying to complete the full and then sharing.¹

Part of my exploration last year also involved building and open sourcing Panopticon – a Python tracer to visualize async execution, so I'll also touch upon that along the way, with some other projects I have in mind.

As far as possible, I'll also be describing how I went about spelunking into asyncio along the way as well.

tree ~/expLog/src/drafts/asyncio

import asyncio
import random

async def hello_world():
    for ch in "Hello, world!":
        x = max(0, random.gauss(.08, .03))
        await asyncio.sleep(x)
        print(ch, end="", flush=True)
    print()

asyncio.run(hello_world())

import random
import time

def sync_hello_world():
    for ch in "Hello, world!":
        x = max(0, random.gauss(.08, .03))
        time.sleep(x)
        print(ch, end="", flush=True)
    print()

sync_hello_world()