C++ Coroutines and Structured Concurrency in Practice - Dmitry Prokoptsev - CppCon 2024

[@ladnir]

37:11 is untilCanceledAnd a macro? Otherwise you have a typo as f will close immediately.

[@hosein9157]

Shouldn't the recursive call at 13:11 be co_await allOf(...)? It both makes more sense logically and is referenced as such in the following slide.

[@BigLongRandomNumberNameM-kf9vy]

You're probably right, it looks like a typo to me too.

[@codures]

Awkwardly, I find the nursery concept quite inspiring as I pulled this kind of a stunt in Delphi 15 years ago when I wrote a middleware. However, and contrary to what was said, there are quite a few methods to deal with thread creation overhead, at nursery level. The concept of "unsafe nursery" might work faster but, it should be isolated - also experimented: spawning child nurseries until resource exhaustion, checked.

[@andreaspokorny3089]

I do not see yet the difference to the role the "io_context" / "executor" fulfills when building applications with asio. You need it with coroutines and continuation style code..

[@jaybny]

this is nice

[@TheNovakon]

Why you guys didn't inspire in Javascript. Adopt promise-future concept and problem solved

[@Kitsmena]

We do have promise-future

[@lombeelo8780]

Because the concept is too simple for the inteded use-cases

[@dororofied]

There is no need for futures and promises with the existence of coroutines. Promise and futures were made to overcome the limitation of a thread-based task system where a task could not suspend mid execution to allow a child task to complete. Hence why there was the "promise" where some task promises to feed its outputs to another task. With coroutines, all we have to do is suspend the current coroutine, co_await on a subroutine and once the subroutine is done, the suspended coroutine immediately resumes execution while receiving the outputs from the subroutine.

[@lefteriseleftheriades7381]

the pacing of the presenter is too fast

[@alx9r]

Yup. I’m watching at 0.75x. That works ok.