What is the purpose of async_std::sync::Arc?

Question

I understand why structures that may block the current thread, like Mutex and RwLock, need to be reimplemented to rely on the futures/waker API, but why does it need to do so for Arc?

As far as I understand, no operations on the standard Arc are blocking, and in the latest version, async_std::sync::Arc is just a reexport of std::sync::Arc.

Is this future-proofing, in case a future implementation of Arc needs to rely on blocking?

Which one should I use in async code, and why?

Answer 1

The point is to ease the transition from std to async_std as much as possible.

The description of the crates starts with:

This crate provides an async version of std.

And one of its descriptive feature is:

Intuitive: Complete parity with the stdlib means you only need to learn APIs once.

As for the blog post that introduced it, it clearly mentions that async_std is mean to be a drop-in replacement of stdas much as possible:

With async-std​, all that’s needed is replace std​ with async_std​, add the prelude​, and sprinkle in a few .await​s

(emphasis is mine)

Given all that, it makes sense that the developers of async_std would chose to re-export all types from std that don't need to be adapted to async: Users don't need to worry about whether types are async or not and don't need to worry about what to use.