Hacker News Reader

Comment by gregoryl | Hacker News Reader

Isn't generics just a part of the compiler?

No, .NET generics are real and are reified at runtime, unlike Java where generics are just some compile-time typechecking.

On platforms that don't allow runtime code generation, such as iOS, generics are actually compiled ahead of time.

It's reasonable to ask if the same is done with NanoFramework.

To clarify, I was under-educated on how generics work, rather than being this clever.

Does this imply that generics in Ocaml/Haskell are not real?

"Generics" in Haskell are real, but they're not what is in the .NET lingo called reified. Just like Java does not monomorphise.

On the other hand, C#, F# or Rust monomorphise (have reified generics).

Are “monomorphism” and “reification” interchangeable terms though? Considering how .NET now has covariant/contravariant generics I wonder if it could be described as having “polymorphic” generics as opposed to monomorphic generics - (unrelated to OOP polymorphism, ofc).

AFAIUI _monomorphisation_ and reification are interchangable.

That means, that if I have a polymorphic (generic) class `MyClassA<T>`, the compiler (regardless if it is JIT in the runtime as in C#, or AOT as in Rust) monomorphises (reifies) `MyClassA` for each instantiation of `T`. So `MyClassA<MyClassB>` and `MyClassA<MyClassC>` are distinct classes.

For this reason, co-/contra-variance in C# works only for interfaces. Type parameters in classes are invariant. `MyClassA<MyClassB>` and `MyClassA<MyClassC>` are distinct classes, neither sub-/super-class of the other, even if `MyClassB` were a subclass of `MyClassC`.

My understanding is it doesn't matter in Haskell since it lacks reflection, they are real because you can't reflect and see they have been type erased at runtime.

In C# and Java reflection is used extensively and the generics still exist in C# when reflecting they are gone in Java.

Reply on news.ycombinator.com