# Maybe-RFC: trait for \`as\` casting **URL:** \ **Category:** language design\ **Created:** [June 17, 2019, 6:43pm UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419 "2019-06-17T18:43:17Z")\ **Posts on this page:** 8\ **Page:** 1
**Author:** ![CodeSandwich](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/codesandwich/32/3479_2.png) [@CodeSandwich](https://internals.rust-lang.org/u/CodeSandwich)\ **Post date:** [June 17, 2019, 6:43pm UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/1 "2019-06-17T18:43:17Z")
This is a maybe-RFC, because I’m not even sure if the idea is actually as interesting as it seems to me. # Idea To create an `As` trait, which represents possibility of casting with `as` keyword: ```rust trait As { fn as(self) -> T; } ``` By default the trait is defined in stdlib for all types that can be casted with an `as` keyword. # Motivation Currently there’s no way to define a type that is castable to another one. One can’t write a generic structure that holds anything castable to u32. Or a function that accepts fn item that can be casted to a specific fn pointer. Another thing is elevation of `as` keyword to a fully functional operator, that can be defined by user for types, same as `+`, `[]` and `?`. Automatic coercion between custom types becomes a thing. This potentially makes the experimental `CoerceUnsized` trait fully covered and obsolete. ---
**Author:** ![H2CO3](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/h2co3/32/2849_2.png) [@H2CO3](https://internals.rust-lang.org/u/H2CO3)\ **Post date:** [June 17, 2019, 7:18pm UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/2 "2019-06-17T19:18:03Z")
> [@CodeSandwich](#): > > Currently there’s no way to define a type that is castable to another one Why aren't `From` and `Into` good enough for this purpose? These are the standard traits with which user-defined conversions are supposed to be defined. ---
**Author:** ![CodeSandwich](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/codesandwich/32/3479_2.png) [@CodeSandwich](https://internals.rust-lang.org/u/CodeSandwich)\ **Post date:** [June 17, 2019, 9:00pm UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/3 "2019-06-17T21:00:01Z")
`From` and `Into` are definitely enough. Yet some conversions are defined with `as` and that gives them automatic coercion. ---
**Author:** ![scottmcm](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/scottmcm/32/2355_2.png) [@scottmcm](https://internals.rust-lang.org/u/scottmcm)\ **Post date:** [June 17, 2019, 9:00pm UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/4 "2019-06-17T21:00:02Z")
This has come up many times; the core problem is that `as` is too random in the kinds of things that it does to be usable generically. Personally, I’d like to see subsets, like the `WrappingFrom` in > Since that at least has a justifiable semantics instead of “do something to turn a thing into a thing and who knows what will happen”. ---
**Author:** ![197g](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/197g/32/7276_2.png) [@197g](https://internals.rust-lang.org/u/197g)\ **Post date:** [June 18, 2019, 6:44am UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/5 "2019-06-18T06:44:20Z")
Why a new trait in particular? For integers a wrapper type alone could suffice. ```rust struct WrappingU16(u16); impl From for u8 { … } impl u16 { pub fn wrapping(self) -> WrappingU16 { WrappingU16(self) } } let x: u16 = 0; let y: u8 = x.wrapping().into(); ``` ---
**Author:** ![CodeSandwich](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/codesandwich/32/3479_2.png) [@CodeSandwich](https://internals.rust-lang.org/u/CodeSandwich)\ **Post date:** [June 18, 2019, 8:01am UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/6 "2019-06-18T08:01:26Z")
The wrapper is going to work in non-generic case, but it’s a workaround adding custom cruft to code. The new trait is clean, global and generic solution. At least it’s supposed to be. ---
**Author:** ![scottmcm](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/scottmcm/32/2355_2.png) [@scottmcm](https://internals.rust-lang.org/u/scottmcm)\ **Post date:** [June 20, 2019, 3:58am UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/7 "2019-06-20T03:58:13Z")
> [@197g](#): > > Why a new trait in particular? For integers a wrapper type alone could suffice. Because coherence. (`std::num::Wrapping` isn't fundamental.) It's legal to ```rust impl From for Foo { ... } ``` But using wrapping you get ```rust error[E0117]: only traits defined in the current crate can be implemented for arbitrary types --> src/lib.rs:6:1 | 6 | impl From> for Wrapping { | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ impl doesn't use types inside crate ``` ---
**Author:** ![system](https://sea2.discourse-cdn.com/flex002/user_avatar/internals.rust-lang.org/system/32/14092_2.png) [@system](https://internals.rust-lang.org/u/system)\ **Post date:** [September 18, 2019, 4:05am UTC](https://internals.rust-lang.org/t/maybe-rfc-trait-for-as-casting/10419/8 "2019-09-18T04:05:07Z")
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.