On Mon, May 11, 2015 at 7:36 AM, Clinton Mead <clintonmead@gmail.com> wrote:

The "letSTRef" is what's new. The value it returns can change based on the parts that make up it's function changing.

Certainly that sounds very close to FRP, which propounds the concept of _time-varying_ values.

I understand this syntax above isn't going to work (I'd have to use applicative at least I'd imagine) but my main question is that does something like this exist?

Let's assume you want to adopt the classic effectful approach to FRP. Now when you write:

x <- newSTRef 2

y <- newSTRef 3

z <- letSTRef (x + y)

you're really working in a pure fragment. Moreover, letSTRef, while attractive, is sort of on the wrong track.

Suppose we have two time-varying values mouseX and mouseY. They are both at least applicative values, i.e. f Int for some applicative, possibly even monadic, functor f.

They are also primitive, like the way a Haskell Int is primitive and involves hardware-wired details, meaning you'll do the actual implement at a lower level than the following:

We have a Euclidean distance function dist :: Int -> Int -> Int.

To track the distance of the mouse pointer from the origin, we'd write liftA2 dist mouseX mouseY :: f Int, thus deriving another time-varying Int.

Does this help clarify?

-- Kim-Ee