Inferring from context declarations

Lennart Augustsson lennart@mail.augustsson.net
Thu, 22 Feb 2001 09:10:18 -0500


Thomas Johnsson wrote:

> Lennart Augustsson writes:
>  > > Simon Peyton Jones' comments about dictionary passing are a red herring,
>  > > since they assume a particular form of compiler.  Various (MLj, MLton)
>  > > ML compilers already inline out all polymorphism.
>  > ML is a language where you can do this.  In Haskell it is not always
>  > possible to eliminate all polymorphism (due to polymorphic recursion).
>
> I'd like to see an example of this!

Sure, here's one where you can't bound the number of dictionaries:

main = interact $ show . f 'a' . read

f :: (Eq a) => a -> Integer -> Bool
f x 0 = x == x
f x n = f [x] (n-1)


If the input is 0 the comparison is of Char,
if the input is 1 the comparison is of [Char],
if the input is 2 the comparison is of [[Char]],
etc.

Incidentally, this has nothing to do with allowing polymorphic recursion
on functions in Haskell.  It could be done earlier too, but then it had
to be encoded using a class and instance declaration.

    -- Lennart