Suggest async_recursion crate when writing a recursive async fn

[estebank]

We should suggest the use of https://docs.rs/async-recursion/0.3.2/async_recursion/ when writing

async fn fib(n : u32) -> u64 {
   match n {
       0     => panic!("zero is not a valid argument to fib()!"),
       1 | 2 => 1,
       3     => 2,
       _ => fib(n-1).await + fib(n-2).await
   }
}

which currently only says:

error[E0733]: recursion in an `async fn` requires boxing
 --> src/lib.rs:1:26
  |
1 | async fn fib(n : u32) -> u64 {
  |                          ^^^ recursive `async fn`
  |
  = note: a recursive `async fn` must be rewritten to return a boxed `dyn Future`

[Aaron1011]

Is there any precedent for the compiler suggesting an external crate?

[thomaseizinger]

Is there any precedent for the compiler suggesting an external crate?

Yes, for async traits the compiler suggests to use async-trait: https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=b80e16a6aadbc2c76a3ce4c2296af1bd

[estebank]

As @thomaseizinger pointed out, we do for async-trait. We normally wouldn't recommend one crate over another ("don't play favorites"), but for small, single use, specialized crates that fill a common language "hole" that we might be working towards "plugging", I think it is a much better experience to immediately unblock newcomers than send them to a user forum to ask how to get around the limitation. The current error gives some idea of how to fix this, but not nearly enough for an even moderately knowledgeable Rust user to fix the problem without some searching.

That being said, thanks for double checking and mildly pushing back on a potentially controversial and "uncommon" request. That is important for the health of the project.