Infer StableInterpolate on tuples

[mweatherley]

Objective

Make StableInterpolate "just work" on tuples whose parts are each StableInterpolate types. These types arise notably through Curve::zip (or just through explicit mapping of a similar form). It would otherwise be kind of frustrating to stumble upon such a thing and then realize that, e.g., automatic resampling just doesn't work, even though there is a very "obvious" way to do it.

Solution

Infer StableInterpolate on tuples of up to size 11. I can make that number bigger, if desired. Unfortunately, I don't think that our standard "fake variadics" tools actually work for this; the anonymous field accessors of tuples are :tt for purposes of macro expansion, which means that you can't simplify away the identifiers by doing something clever like using recursion (which would work if they were :expr). Maybe someone who knows some incredibly dark magic could chime in with a better solution.

The expanded impls look like this:

impl<
        T0: StableInterpolate,
        T1: StableInterpolate,
        T2: StableInterpolate,
        T3: StableInterpolate,
        T4: StableInterpolate,
    > StableInterpolate for (T0, T1, T2, T3, T4)
{
    fn interpolate_stable(&self, other: &Self, t: f32) -> Self {
        (
            <T0 as StableInterpolate>::interpolate_stable(&self.0, &other.0, t),
            <T1 as StableInterpolate>::interpolate_stable(&self.1, &other.1, t),
            <T2 as StableInterpolate>::interpolate_stable(&self.2, &other.2, t),
            <T3 as StableInterpolate>::interpolate_stable(&self.3, &other.3, t),
            <T4 as StableInterpolate>::interpolate_stable(&self.4, &other.4, t),
        )
    }
}

Testing

Expanded macros; it compiles.

Future

Make a version of the fake variadics workflow that supports this kind of thing.

[alice-i-cecile]

Can you add the fake_variadics doc attribute? We've added this elsewhere and it really improves the rustdoc output.

I would suggest using the all_tuples macro here, but it's good to not pull in that dependency.

[NiseVoid]

Kind of cursed, but it does look like this is the only way without introducing new deps.

[mockersf]

could you try to use https://docs.rs/bevy/latest/bevy/utils/macro.all_tuples.html?

[mweatherley]

Can you add the fake_variadics doc attribute? We've added this elsewhere and it really improves the rustdoc output.

Do you know how to do that outside of the context of the all_tuples macro? Everything I try basically winds up with the compiler complaining to me about how I'm not allowed to use #[doc(fake_variadics)], even when I add a module-level allow.

I would suggest using the all_tuples macro here, but it's good to not pull in that dependency.

It doesn't work anyway :( I spent quite a while trying to do it that way and reached the conclusion that it's basically impossible.

[alice-i-cecile]

I'm fine with this for now :)

[alice-i-cecile]

@BenjaminBrienen @bash y'all have any advice? You were very helpful for this in #15387

[mweatherley]

I'm fine with this for now :)

For what it's worth, I got it working with a new version of all_tuples in my branch now. At least long-term, I'd also prefer to do it that way, since I have other traits I want to derive in a similar way. I'd be happy to shove the more complicated thing into a follow-up, too, though.

[BenjaminBrienen]

I had the same errors he's describing. It only worked for me within all_tuples!.

[alice-i-cecile]

Follow-up it is then ;)

[bash]

@BenjaminBrienen @bash y'all have any advice? You were very helpful for this in #15387

Yeah #[doc(fake_variadic)] can be a bit of a hassle :))
I created a follow-up PR that adds it: #15933