Higher rank types

Graham Klyne GK@ninebynine.org
Tue, 08 Jul 2003 11:18:34 +0100


I took a quick skim through this -- not closely enough to follow all the 
details but I still found some useful background understanding.

For me, it clarified slightly the role of types on the left and right hand 
sides of a function arrow, which I've seen mentioned in passing but never 
explained, but there's still a gap in my grasp of this notion (something to 
do with types on LHS of arrow being subject to pattern matching?).

A lingering question in my mind, which may be pretty irrelevant to your 
purpose, is the relationship between these higher rank types and GHC's 
existentially bound types (which also use the 'forall' keyword).

(Working as I do with Semantic Web technologies and ideas, I was interested 
to note your "simple rules ... rather than trying to exploit every ounce of 
inference".  I have similar views regarding the role of inference in 
semantic web applications.  I think it will be interesting to study your 
inference patterns in detail (when I find time!), and in particular to 
compare your approach to subsumption with the Description Logic approach 
favoured by many in the SWeb community.)

#g
--

At 11:23 07/07/03 +0100, Simon Peyton-Jones wrote:
>Dear Haskellers
>
>For a long time now, Mark Shields and I have been writing a paper
>describing GHC's approach to type inference for higher-rank types.
>We've finally finished a complete draft:
>
>         Practical type inference for arbitrary-rank types
>         Simon Peyton Jones and Mark Shields
>         http://research.microsoft.com/~simonpj/papers/putting/index.htm
>
>The paper has a strongly tutorial flavour, and comes complete with a
>prototype implementation that you can play with.  My hope is that, in
>addition to describing higher-rank stuff, the paper may serve as a kind
>of tutorial on type inference generally, and the use of monads to
>support the plumbing in particular.  There's nothing really new about
>this aspect of the paper, but I don't know of a comparable tutorial that
>takes a monadic approach.
>
>It's still very much a draft, so I'd be very interested in your feedback
>about it.
>
>Simon
>
>_______________________________________________
>Haskell mailing list
>Haskell@haskell.org
>http://www.haskell.org/mailman/listinfo/haskell

-------------------
Graham Klyne
<GK@NineByNine.org>
PGP: 0FAA 69FF C083 000B A2E9  A131 01B9 1C7A DBCA CB5E