[MLton] expansive expressions yielding polymorphic values

[Stephen Weeks]

>  Note,  even  though  f  is  the  result  of an expansive expression (function
>  composition), the signature Z can make its type polymorphic.
>
>  Can people confirm that this really is supposed to be legal?

I think it is supposed to be legal.  What is going on is not that "f"
is polymorphic, it is that f's type (unit -> int) is at least as
general as the type specified in the signature ('a t -> int), where
the signature matching gets to assume that 'a t = unit.  A simpler
example might be

structure Z : sig
  type 'a t
  val s : 'a t
end = struct
  type 'a t = string
  val s = "foo"
end

Also note that once you have hidden the fact that "t" is monomorphic,
clients can no longer take advantage of that fact in type checking
their own expansive expressions.