Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!world!iecc!compilers-sender
From: brm@Neon.Stanford.EDU (Brian R. Murphy)
Newsgroups: comp.compilers
Subject: Re: Static type-checking with dynamic scoping
Keywords: types, design
Message-ID: <9101160042.AA28163@Neon.Stanford.EDU>
Date: 16 Jan 91 00:42:07 GMT
Sender: compilers-sender@iecc.cambridge.ma.us
Reply-To: Brian R. Murphy <brm@Neon.Stanford.EDU>
Organization: Computer Science Department, Stanford University
Lines: 31
Approved: compilers@iecc.cambridge.ma.us

It seems to me that one could consider (dynamically bound) free
variables in a procedure to simply be additional parameters to the
procedure.  Then standard ML-like type inference could be done.

Not being fluent with ML syntax, I give an example in (dynamically
scoped) Lisp:
	(defun f (x)
	    (y x)) ;; y free, bound dynamically
Here f might taken to have type:
	([y: a->b], a) -> b
that is, it takes an argument of type a, a binding of y with type a->b,
and returns a value of type b.

A place where f is used might look like:
	(let ((y sub1))
	    (f 3))
where y has type
	int -> int
and f is applied to ([y:int->int], int) which may be unified with the
above, yielding a result of int as well.

Keeping track of all of the current bindings and determining which
onese were relevant in a particular application would certainly
clutter up the type inference, but it shouldn't make it terribly
more difficult.

					-Brian Murphy
					 brm@cs.stanford.edu
-- 
Send compilers articles to compilers@iecc.cambridge.ma.us or
{ima | spdcc | world}!iecc!compilers.  Meta-mail to compilers-request.