Yes, and that’s the worst part of async. That’s why you need to be very strategic about where you introduce it into your code in order to minimize the number of functions it infects, not give up and write a framework that’s all async for no good reason.
Yes. But you should be equally strategic about introducing sync code into your platform. Because making your platform sync basically makes it only be able to call sync functions of your code.
It's not that async infects. It's sync that infects and restricts. We are just used to it by default.
The fact that we started from sync was the cause of all the trouble of rpc because everything outside of CPU is innately async.
So make your utility functions sync whenever you can but make your platforms and frameworks async.
I just completely disagree. Async is syntactic sugar that can be reduced to sync code with callbacks. It doesn’t exist on equal footing. If you want to call sync code from async code, you just… call it. If it performs blocking IO, it’ll block, but that’s exactly what it would do if called from other sync code, too.
By contrast, calling async code from sync code requires a special blocking wrapper (Python) or unavoidably breaks control flow (JavaScript).
> By contrast, calling async code from sync code requires a special blocking wrapper (Python) ...
That's exaclty my point. If you don't have async by default in your platform you need to do stupid things to fake it. If function calls and main in Python were innately async you could be calling async code just as easily as sync code.
> [...] or unavoidably breaks control flow (JavaScript).
async/await syntax avoids it completely.
Tbh await should be default function call semantics and there should be special keyword for calling without awaiting. But since we come from sync primitives that would require small revolution that might happen at some point.
> Async is syntactic sugar
You could make sync code be syntactic sugar for await.
I hope that someone does an oral history of why gevent wasn't seen as the solution here. The existence of models like Twisted, and a general idea that yields to an event thread should be explicit in some way, I think caused the exact kind of fracturing of the ecosystem that everyone was trying to avoid. "Everyone will write async code" simply didn't happen in practice.
You had never tried gevent in production then.
As soon as workload and concurrency increases python programs with gevent or gevent based drivers, especially monkey patches causes unexpected crashes out of the blue, no way to debug, no error message,Memory leaks and whole slew of nightmares
It's not that async infects. It's sync that infects and restricts. We are just used to it by default.
The fact that we started from sync was the cause of all the trouble of rpc because everything outside of CPU is innately async.
So make your utility functions sync whenever you can but make your platforms and frameworks async.