Improve Box<T> -> Pin<Box<T>> conversion

[Nemo157]

I found the From trait conversion for this very hard to find, having a named function for it is much more discoverable. Also fixes #56256 as I need that in the place I'm using this.

Has a placeholder tracking issue, will file an issue once I get feedback.

[rust-highfive]

r? @KodrAus

(rust_highfive has picked a reviewer for you, use r? to override)

[Nemo157]

r? @cramertj

[cramertj]

I feel sort of odd about having Box::pin create a Pin<Box<_>> and using into_pin to convert a Box<_> into a Pin<Box<_>>, but it seems basically fine to me. Assigning to a T-libs member to FCP.

r? @alexcrichton

[alexcrichton]

Since this is unstable it doesn't need an FCP, but should this also be added to Rc and Arc?

[cramertj]

@alexcrichton No, this method wouldn't be safe on Rc and Arc because if you have one Rc<_> and one Pin<Rc<_>> pointing to the same value (which could be created by cloning an Rc and using the method you describe) you can use Pin<Rc<_>> to witness that the inner value would never be moved, drop the Pin<Rc<_>>, then use Rc::try_unwrap to move the value out of the remaining Rc<_>, which would be unsound.

[alexcrichton]

Sure thing

@bors: r=cramertj

[bors]

📌 Commit d1a42ea has been approved by cramertj

[hellow554]

Funny enough, today I found this function and asked myself why the heck does it exist? It clearly states, that From already does the same, so why having another function doing the same? @Nemo157 what were your reaons apart from "it can be found easier"? Have you ever used it yet? ^^ I'm just curious, because to me it feels verbose and not necessary.

[Nemo157]

@hellow554 it's in use here: https://github.com/rust-lang-nursery/futures-rs/blob/8efde615a80aa23ece05d11edbe40db73f326f41/futures-util/src/future/join_all.rs#L119

[hellow554]

I see, but elems.into() would have done the same. I see the explicitness though... :/ I'm indecisive.

[Nemo157]

Yep, but when implementing that originally I didn’t even think of elems.into() and just used an unsafe block around Pin::new_unchecked, I think I only noticed that From worked when adding this method.

[czipperz]

If you want to be explicit, you could just use Pin::from(x) rather than x.into_pin().

[Nemo157]

@czipperz it's interesting that that works, but it's relying on the currently unique implementation of From<Box<_>> so could break if any more implementations are added (unless inference constrains the return type to be Pin<Box<_>>), e.g. playground.

[czipperz]

I misread the code as a method rather than a function. IE Pin::from(box) is equivalent to Box::into_pin(box) (as long as other implementations aren't in scope)

[czipperz]

I don't think this is really a useful feature to specialize in this way. Can't you just use the hyper fish operator to disambiguate?

[czipperz]

Like do people really implement this sort of functionality? It seems like there isn't really a reason for this feature to exist imo.