Friday, 4 May 2012

F# for Scheme programmers

Scheme is a dynamically typed functional programming language commonly taught on undergraduate courses in the US. This post looks at the key concepts that Scheme programmers will have to learn in order to use F# effectively.
Static typing is the major difference between Scheme and F#. This facilitates a style called typeful programming where the type system is used to encode constraints about functions and data such that the compiler proves these aspects of the program correct at compile time and any violations of the constraints are caught immediately.
For example, a sequence of one or more elements of the same type might be conveyed by a value of the following type:
type list1<'a> = List1 of 'a * 'a list

let xs = List1(1, [])
let ys = List1(2, [3; 4])
The compiler now guarantees that any attempt to use an empty one of these sequences will be caught at compile time as an error.
Now, the reduce function makes no sense on an empty sequence so the built-in implementation for lists barfs at run-time with an exception if it encounters an empty sequence:
> List.reduce (+) [];;
System.ArgumentException: The input list was empty.
Parameter name: list
   at Microsoft.FSharp.Collections.ListModule.Reduce[T](FSharpFunc`2 reduction, FSharpList`1 list)
   at <StartupCode$FSI_0271>.$FSI_0271.main@()
Stopped due to error
With our new sequence of one or more elements, we can now write a reduce function that never barfs at run-time with an exception because its input is guaranteed by the type system to be non-empty:
let rec reduce f = function
  | List1(x, []) -> x
  | List1(x0, x1::xs) -> f x0 (reduce f (List1(x1, xs)))
This is a great way to improve the reliability of software by eliminating sources of run-time errors and it is something that dynamically typed languages cannot do.

No comments: