Expand float_cmp to structs with PartialEq

[RazrFalcon]

Example:

#[derive(PartialEq)]
struct Data {
    v: f64,
}

fn main() {
    let d1 = Data { v: 1.0 };
    let d2 = Data { v: 2.0 };

    // here should be a warning
    println!("{}", d1 == d2);
}

[clarfonthey]

But this should only be a warning if the structure also implements PartialOrd.

[RazrFalcon]

Doesn't PartialEq implemented without comparing the floats?

[clarfonthey]

@RazrFalcon generally, the suggestion for floats is that instead of doing something like x == y, you'd do (x - y).abs() < z.

Although in this case, we have no idea if that's necessarily possible with these sorts of structs. You could warn on PartialEq for structs that are just floats, though, suggesting other alternatives.

[oli-obk]

I think this should not lint on the comparison site, but on the struct definition. Any struct deriving PartialEq and having float fields should be linted against.

[cschwan]

I think this should not lint on the comparison site, but on the struct definition. Any struct deriving PartialEq and having float fields should be linted against.

I think it's a bit more complicated than this, for instance consider comparing two Vec<f64>; in this case I would like the linter to complain at the comparison site, but I agree that when a struct has non-generic float fields it's best linted on the struct definition.