Hacker News Reader

Comment by ddek | Hacker News Reader

Parser combinators in F# were where I really grasped the utility of a Monad. F# doesn’t have HKT’s, and computational expressions are a little different from Haskell ‘do’.

It was the ‘(>>=) = Parser<‘a> -> (‘a -> Parser<‘b>) -> Parser<‘b>’ that flicked the lightswitch in my head.

View on HN

In .NET I still find it weird that people sometimes implement the visitor pattern without the output type. You'd have to truly think imperatively to consider:

    void Structure(IVisitor visitor);

a good function signature, instead of:

    A Structure(IParser<A> parser);

I mean fair enough that you can't be bothered to implement the functor and monad methods, but it's downright silly to define a function that is incapable of returning output.

In .NET specifically a reason I can think about doing that is that you'd need to implement your visitor twice, as the type argument cannot be void (if I recall correctly).

It can't be void, but writing

    public sealed class Unit {
        private static Unit m_value = new Unit();

        public static Unit Value => m_value;

        public Unit() { }

        public override string ToString() => "()";

        public override bool Equals(object obj) => obj is Unit;

        public override int GetHashCode(object obj) => 0;
    }

and then using it ("return Unit.Value") in place of void is really not that big a problem.

True, but I just never quite managed to understand why you'd want your visitor to return nothing. And if you did why it was too much hassle to just return '0' or something.

Reply on news.ycombinator.com