diagnostics: suggest using smaller type if constant is small enough and smaller type impls conversion

[matthiaskrgr]

this code

fn main() {
    let x = 123.1234_f32 - f32::from(100);
}

fails to compile as follows:

error[E0277]: the trait bound `f32: std::convert::From<i32>` is not satisfied
 --> src/main.rs:2:28
  |
2 |     let x = 123.1234_f32 - f32::from(100);
  |                            ^^^^^^^^^ the trait `std::convert::From<i32>` is not implemented for `f32`
  |
  = help: the following implementations were found:
            <f32 as std::convert::From<u8>>
            <f32 as std::convert::From<i16>>
            <f32 as std::convert::From<u16>>
            <f32 as std::convert::From<i8>>
  = note: required by `std::convert::From::from
error: aborting due to previous error
For more information about this error, try `rustc --explain E0277`.

We know that 100 < <u8>::max_value()and that there is <f32 as std::convert::From<u8>> thus maybe we could suggest using f32::from(100_u8) instead?

[leonardo-m]

thus maybe we could suggest using f32::from(100_u8) instead?

Even better is to see 100 as a untyped value that can fit into the <f32 as std::convert::From<u16>>. So the _u8 is unnecessary. Is this possible?

[matthiaskrgr]

Hmm, on second thought, it might be confusing if

let y = f32::from(100);

behaves differently from

let x = 100; // i32
let y = f32::from(x);

so forcing to explicitly make it u8 might be more clear.