Hacker News Reader

Comment by galangalalgol | Hacker News Reader

The one lagging thing that isn't easy with rust's generics is expressions, and that looks to be getting kicked down the road indefinitely. You can't have Foo<N> * Foo<M> -> Foo<N+M> or similar. That is a useful thing for statically sized math libraries and for performance oriented metaprogramming. The latter can be clumsily handled with macros, but not the former. It is also blocking portable simd from stabilizing, apparently now indefinitely. I still wouldn't pick any other language a new production product, but I find the stalling out on const generic expressions frustrating, as simd and static math libraries are things I use daily.

View on HN

> You can't have Foo<N> * Foo<M> -> Foo<N+M> or similar.

You can though, unless I totally misunderstand your syntax

    use std::{marker::PhantomData, ops::{Add, Mul}};

    pub struct Foo<T> {
        _p: PhantomData<T>,
    }

    impl<N, M> Mul<Foo<M>> for Foo<N> 
    where
        N: Add<M>
    {
        type Output = Foo<<N as Add<M>>::Output>;
        fn mul(self, rhs: Foo<M>) -> Self::Output {
            todo!()
        }
    }

You misunderstood. N, M are supposed to be integers (const generics); in your example code you've made them types. Also, your `type Output = Foo<<N as Add<M>>::Output>;` just means "multiplication has the same return type as addition". But desired is that multiplying a Foo<4> with a Foo<3> results in a Foo<7>.

Rust decided that it's important to not have instantiation-time compiler errors, but this makes computation with const generics complicated: you're not allowed to write Foo<{N+M}> because N+M might overflow, even if it never actually overflows for the generic arguments used in the program.

Instantiation time errors unlock so much metaprogramming potential that Rust is going to be forced to allow them sooner or later.

I believe this kind of stuff is being worked on.

aiui this isn't inherently limited by instantiation-time error messages and is available on nightly today with the generic_const_exprs feature. It's in the pipeline.

I could go on and on about the limitations of min_const_generics, missing TAIT (type alias impl trait), unstable coroutines/generators, etc. but none of that stuff erases how much of a miracle Rust is as a language. The industry really needed this. It's practically the poster child of memory safety and security-critical industries are picking it up and Ferrocene is pushing it in safety-critical environments as well and it's just. Good. Please. Finally a serious contender against the terrible reign of C and C++.

Reply on news.ycombinator.com