error msg "Add/AddAssign (etc) is not implemented for the type _" poor for beginners

[pnkfelix]

Reduced version of code encountered during a tutorial:

fn main() {
    let x = 0;
    x += 1.0;
}

Compiler error message:

error: the trait bound `_: std::ops::AddAssign<_>` is not satisfied [--explain E0277]
 --> <anon>:3:5
3 |>     x += 1.0;
  |>     ^^^^^^^^
help: the following implementations were found:
help:   <f64 as std::ops::AddAssign>
help:   <f32 as std::ops::AddAssign>
help:   <u64 as std::ops::AddAssign>
help:   <i64 as std::ops::AddAssign>
help: and 21 others

error: aborting due to previous error

Namely, the error message says nothing about i32 nor f64 nor the use of let x = 0;, so a beginner really has no place to go for help here.

Can we leverage the on_unimplemented functionality to provide better feedback in the case where one is mixing i32 and f64? (Or does it not have a way to put in a more specific message depending on the types involved in this case?)

There may be an issue for this elsewhere (we all know that our error messaging when type-inference goes to numeric-fallback), but the error message here was particularly bad, and I couldn't find an existing issue quickly, so I am filing this one to make sure this doesn't fall through the cracks.

[eddyb]

I think we can print more than _, in cases where we don't have the expected/found long description.
Applying integer/floating-point defaults before printing errors might be useful, although they could be misleading in some cases. I'm not sure at which point the error is generated here.

integer type: std::ops::AddAssign<floating-point type> is already better, but we can combine this with #[on_unimplemented] to get floating-point type cannot be added to integer type.

We can also sneak in the literal which created the inference variable, but in the current ppaux setup that's a bit problematic (would have to put a mapping in TLS just to print such things, or redo ppaux).

[sanxiyn]

After #35080, we now print the trait bound {integer}: std::ops::AddAssign<{float}> is not satisfied.

[DiThi]

As a beginner I stumbled upon this problem, but with integers, and a more complicated expression. The problem was much more clearly explained (what errors in the expression) when I replaced x += by x = x +

edit: Putting x += back didn't work. Here's the actual code: http://exercism.io/submissions/0186f3e21e974b8f817a9bad500cf2c3

[estebank]

Current output:

error[E0277]: the trait bound `{integer}: std::ops::AddAssign<{float}>` is not satisfied
 --> src/main.rs:3:7
  |
3 |     x += 1.0;
  |       ^^ no implementation for `{integer} += {float}`
  |
  = help: the trait `std::ops::AddAssign<{float}>` is not implemented for `{integer}`

We could maybe add some suggestions to cast from float to integer or vice versa.

[estebank]

The casting suggestion could be added to *Assign traits' rustc_on_unimplemented attr once #44755 is implemented.

[estebank]

Current output:

error[E0277]: cannot add-assign `{float}` to `{integer}`
 --> src/main.rs:3:7
  |
3 |     x += 1.0;
  |       ^^ no implementation for `{integer} += {float}`
  |
  = help: the trait `std::ops::AddAssign<{float}>` is not implemented for `{integer}`

[estebank]

CC #38564

[estebank]

Closing in favor of #38564.