Variants/case classes/algebraic data types/sums/oh my! (original) (raw)

org.openjdk at io7m.com org.openjdk at io7m.com
Fri Jun 17 17:12:52 UTC 2016

On 2016-06-17T15:43:23 +0200 Remi Forax <forax at univ-mlv.fr> wrote:

Thinking a little bit more about that ... I we want to represent something like this, Expr = Value(int value) | Add(Expr left, Expr right) instead of using an interface wich is wrong because we want Value and Add to be Exprs and not subtypes of Expr, _we can use the Where clause to disambiguate between the Add and the Value.

I think there's some merit to this.

As JB Giraudeau pointed out in the original github thread:

btw, another problem of sealed hierarchies, often encountered in scala, is that they expose subtyping relationships, which often ruin type inference.

I think this is something that anyone coming from Haskell or OCaml has experienced. You write what would be a perfectly reasonable monadic expression and the compiler decides that all of the types are Any and gets upset.

M

More information about the valhalla-dev mailing list