Algebraic datatypes

[peq]

Just an idea.

Datatypes would be a generalization of tuples. Nonrecursive datatypes would be translated similar to tuples. Recursive datatypes would be translated similar to classes, but with some kind of reference counting to have automated deallocation. This is possible, because immutable datatypes cannot have cycles, and because datatypes will not be castable to int.

Some examples:

// could replace tuples:
data vec2 = vec2(real x, real y)
// can be generic and have cases:
data option<A> = Some(A a) | None
// can be recursive
data list<A> = Cons(A head, list<A> tail) | Nil

Discuss.

[peq]

Of course, datatypes could also replace enums. So the language could get smaller 👍

[Cokemonkey11]

I would use it. But what about standard library using options? we are already missing a lot of obvious stl functions and frequently PRs are a pain for user and maintainer.

Option is super nice but it means that pretty much any STL extension function that uses a handle (unit, item, destructable especially) should have another override for Option, I think.

I don't think this is more harm than good - I think it's nice. Just pointing it out.