Back to Questions
1.1k
views2
comments
A little typing challenge
f#

F# does not accept the definition of f below. Can anybody think of a way to change the definition of f so that the full code including the last three lines compile without further changes? It doesn't matter whether inline of not, direct or indirect, dynamic or static, I just need f to upcast the second argument if necessary and don't want to explicitly cast in the last lines.

type Base = class
val a: int
new (a) = {a = a}
end

type Child = class
inherit Base
val b: int
new (b) = {inherit Base(-1); b = b}
end

type C<'a> when 'a :> Base = class
val c: 'a
new (c) = {c = c}
end

let f (c: C<'a>) (a: #'a): int = // F# will further further constrain the signature...
c.c.a + a.a

let v1 = (f (C(Base(1))) (Base(1)))
let v2 = (f (C(Child(1))) (Child(1)))
let v3 = (f (C(Base(1))) (Child(1))) //... so that this line doesn't compile

Stephan

.

Does F# support some kind of dependent type? If not, I can't see a clear answer.

By on 11/8/2007 12:53 AM ()

Good question!

You ought to be able to write

1

let f (c : C<'a>) (a : 'b) : int when 'a :> Base and 'b :> 'a = c.c.a + a.a

but for whatever reason F# decides to unify 'a and 'b which is not what we want. So we could try making the type variables rigid like so

1

let f<'a, 'b when 'a :> Base and 'b :> 'a> (c : C<'a>) (a : 'b) : int = c.c.a + a.a

But this produces many interesting compiler errors. It would seem that the compiler is treating 'a in the constraint 'b :> 'a as a sealed type, which is not really correct, and concluding that the constaint is invalid. My guess is that this is a compiler bug.

By on 11/7/2007 8:22 PM ()

    Topic tags

    IntelliFactory Offices Copyright (c) 2011-2012 IntelliFactory. All rights reserved.
    Home | Products | Consulting | Trainings | Blogs | Jobs | Contact Us | Terms of Use | Privacy Policy | Cookie Policy     		Built with WebSharper