Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!elroy.jpl.nasa.gov!decwrl!world!iecc!compilers-sender
From: cs450a03@uc780.umd.edu (Raul Rockwell)
Newsgroups: comp.compilers
Subject: RE: Help needed on SML compiler
Keywords: ML, functional
Message-ID: <27MAR91.08353858@uc780.umd.edu>
Date: 27 Mar 91 08:35:38 GMT
References: <9103260248.AA22854@enuxha.eas.asu.edu>
Sender: compilers-sender@iecc.cambridge.ma.us
Reply-To: cs450a03@uc780.umd.edu (Raul Rockwell)
Organization: The University of Maryland University College
Lines: 32
Approved: compilers@iecc.cambridge.ma.us

Surendar Chandra, working on a compiler for a functional language,
gives an example that basically runs:
  define function A
  use function A
  redefine function A
  use function A

Which, I think, points out the reason functional language advocates
tend to look at reassignment as a totally different operation than
assignment.  Many would say that reassignment should be unimplemented
as much as possible.

However, the best way I can see to implement something like the above
is to simply say that the last definition of a name is the one which
is advertised outside the module.  (e.g. given A@1, A@2 and A@3, where
@n indicates serial number, use A@3).

If the function associated with a name can change on-the-fly, for the
same body of code, you'll have to provide an interface for those
functions.  This might be as simple as a "case statement" or, for ill
conditioned code, might be as complex as a call to some generalized
function resolver.

The case of "redefined on the fly" occurs with the use of global
references or where function definition occurs inside a loop (on
non-constant data) or where a function's definition depends on I/O
(e.g. construct fndef from file).

Raul Rockwell
-- 
Send compilers articles to compilers@iecc.cambridge.ma.us or
{ima | spdcc | world}!iecc!compilers.  Meta-mail to compilers-request.