Add a new std::mem::copied function.

[jyn514]

Proposal

Add a new std::mem::copied function.

Problem statement, Motivation, use-cases

Right now if you want to pass a function pointer to a combinator, you need to define a new closure. This is a very small paper cut, but it comes up quite frequently. Having a named function reduces the line noise and makes the code easier to read.

let result_from_ffi_function: Result<(), &i32> = Err(&1);
let result_copied: Result<(), i32> = result_from_ffi_function.map_err(|x| *x);

Solution sketches

// core::mem
pub fn copied<T: Copy>(x: &T) -> T { *x }

The new function will make the code simpler:

use std::mem::copied;
let result_copied: Result<(), i32> = result_from_ffi_function.map_err(copied);

Links and related work

This mirrors core::mem::drop and core::mem::take.

rust-lang/rust#95534, rust-lang/rust#98262, rust-lang/rust#100006

What happens now?

This issue is part of the libs-api team API change proposal process. Once this issue is filed the libs-api team will review open proposals in its weekly meeting. You should receive feedback within a week or two.

[thomcc]

Note: the actual proposal here is to rename the unstable core::mem::copy function to core::mem::copied, and make it a const fn.

Background: this is a function which simply takes a &T where T: Copy and dereferences it, copying it. According to rust-lang/rust#95534, the main use case for this function is for use in combinators, which seems reasonable¹, as a.map(core::mem::copied) does seem clearer than a.map(|b| *b).

The motivation for making it a const fn is that you can already dereference a Copy reference in const contexts. That is, this grants no extra capabilities and it would seem inconsistent if not allowed.

The motivation for the rename to core::mem::copied is to distinguish it from core::ptr::copy, which performs a very different operation. The name copied is consistent with Iterator::copied, Option::copied, as those combinators are morally equivalent to something like thing.map(core::mem::copied)². Given that the use case is combinators, I do think this name makes sense, although I was initially opposed until I realized this. So I'm mostly in favor, especially given that it's unstable.

I think the main thing I'd wonder about this is if there's some module besides core::mem it could live in. It's hard for me to act like mem::copied doesn't bring memcpy to my mind (even if the name of the function on its own makes a lot of sense, and is an improvement).

Footnotes

1. I don't feel that strongly about its usefulness, and don't think I'd go to bat for the functions existence, as I don't know that I've needed to write a closure like |b| *b very many times (at least not since we added .copied() to things). ↩

2. This is of course less true for Iterator than Option (due to the return type), but you get the point. ↩

[jyn514]

at least not since we added .copied() to things

Note that copied() only exists for Iterator and Option; for it to be as useful as the standalone function there would need to be a Result::copied and Result::copied_err. And that still wouldn't support custom data structures or enums.

I agree mem::copied seems a little confusing, but not sure a better name for it ...

[scottmcm]

Copying from rust-lang/rust#95534 (comment):

I think this might really want to be something like

pub trait Clone {
    fn clone(&self) -> Self; // as today

    #[sealed]
    fn copy(&self) -> Self where Self: Copy { *self }
}

And I think someone might be working on an RFC for such an attribute...

[ChayimFriedman2]

@scottmcm This has the disadvantage that you can't use it fully.

[mbartlett21]

But it then has the advantage that it is always in scope, since Clone is in the prelude, and can always* be called as a method.

^{* Actually, it isn't quite always, since types can have a .copy() method, but those methods should be few and very far between...}

[joshtriplett]

👍 for the rename and constification.

(There's still a fair bit of debate about whether we want to stabilize this function, but we all agreed that the new name is better.)