Is it possible to implement `PartialEq`-like traits with `#[recursive]`?

[TheVeryDarkness]

Hello! Thanks for your great work! I've used this in several scenarios, but I have a trait Equiv, which compares enums by each variant. It's like below:

trait Equiv {
    fn equiv(&self, other: &Self) -> bool;
}

Given an enum with the most common usage of derive_recursive below:

#[derive(Recursive)]
#[recursive(
    impl<T: Equiv, U: Equiv> Equiv for Self<T, U> {
        fn equiv(&self, other: &Self) -> bool { aggregate = && }
    }
)]
enum Either<T, U> {
    Left(T),
    Right(U),
}

It will be compiled to:

impl<T: Equiv, U: Equiv> Equiv for Either<T, U> {
    fn equiv(&self, arg0: &Self) -> bool {
        match self {
            Self::Left(f0) => <T as Equiv>::equiv(f0, arg0),
            Self::Right(f0) => <U as Equiv>::equiv(f0, arg0),
        }
    }
}

However, <T as Equiv>::equiv accepts 2 arguments of &T, but not 1 argument of &T and 1 argument of &Either<T, U>.

Is there a way to implement this trait correctly in current version?

[SmakoszJan]

No, currently it's not possible. I will look into this, as it should be relatively straightforward to implement. Thank you for bringing this up!

[SmakoszJan]

Unfortunately it most certainly isn't straightforward to implement. There's a way to do it elegantly but it's pretty complicated would require me to rethink some of the logic (after almost a year, so there's also that). I need my mind fully in shape next week, so I'm taking a break from programming and this would be a bit too much of a strain. You can have a go at it, if you want. If you don't, I'll probably look into it in a week or so.

Generally the idea is to dump the distinction between associated and non-associated and instead look at function parameters and figure out which of them are Self typed and which are not. Then, based on whether the implemented type is a struct or an enum, the parameters would be appropriately destructured. In case of structs (and enum variants) it would be a simple destructuring, no biggie. In case of enums, instead of the current match self there would have to be a match over all the Self-typed params. Any validation related to associated functions would have to instead trigger when these params would be present.

As I said, I'll look into it in a week, but I encourage you to have a go at it yourself.

[TheVeryDarkness]

Yes, I'm glad to have a try.