Using existential types

[Derek Elkins]

On Mon, 13 Oct 2003 11:37:43 -0400
Jan-Willem Maessen <jmaessen at MIT.EDU> wrote:

> > So, the existential quantification permits more nuanced
> > transformations -- and yet the nuances aren't observable. So, they
> > don't matter?
> 
> Whether they matter or not depends upon our aims.  I'm curious about
> what's possible---in the hopes that I can get a better handle on
> what's really desirable in a particular situation.

Indeed.  One of the issues is useability or at least naturality.  It
seems that existentials in some situations better achieve the
intent/goal.  Anyways, if another challenge is desired, I would be
interested in seeing how one would go about converting the
implementation of Dynamics in "A Lightweight Implementation of Generics
and Dynamics" into a form that doesn't use existentials* but still
provides the same (or very similar or I guess simpler) interface.

* No coding existentials as universals :P
data Showable = Showable (forall r.(forall a.Show a => a -> r) -> r)
mapM_ (\(Showable v) -> v print) [Showable ($ 5),Showable ($ True)]
5
True