Latency in Asynchronous Python(nullprogram.com)
nullprogram.com
Latency in Asynchronous Python
https://nullprogram.com/blog/2020/05/24/
6 comments
I wrote and use https://github.com/CaliDog/asyncpool (CaliDog graciously created a module of it for me) to solve this issue. Here it is forked on my repo: https://github.com/thehesiod/asyncpool
async def process():
time.sleep(JOB_DURATION) # simulate CPU time
A better solution is to make the CPU intensive code cooperative by running it in a different thread or process.
> Never, ever use unbounded queues.
Hmm... I rather like them. But the consumers must be able to keep up with the producers or their must be some form of back pressure.
I think the curio project addresses some of these asyncio design concerns.
asyncio is very poorly designed overall. This is just one offender.
> asyncio.sleep(0) is nearly always wrong.
This, however, is VERY bad advice. a sleep(0) (or, a checkpoint as it is better called) is very useful in things like branches where asynchronous operations only run in one branch. This ensures you don't end up deadlocking half your app if your branches go wrong.