monomorphic pattern binding

[Paul Hudak <hudak-paul@cs.yale.edu>]

Original-Via: uk.ac.nsfnet-relay; Mon, 10 Sep 90 17:47:58 BST
Date: Mon, 10 Sep 90 10:11:19 EDT
From: Paul Hudak <hudak-paul@cs.yale.edu>
To: haskell@cs.yale.edu
Cc: rjmh@cs.glasgow.ac.uk, peterson-john@cs.yale.edu, hudak@cs.yale.edu
In-Reply-To: "Prof. John Hughes"'s message of Fri, 7 Sep 90 14:27:40 BST <9009071505.AA25056@BULLDOG.CS.YALE.EDU>
Subject: monomorphic pattern binding
Sender: haskell-request@cs.glasgow.ac.uk

As John Hughes points out, the issue of pattern-bindings was a point
of real concern on the committee.  The solution chosen was not the
clear-cut best choice, but one that we eventually agreed was at least
the best of a bad lot.  John Peterson points out some of the problems
with the current solution, some of which are now obvious only after
trying to implement the rule (as Kevin Hammond will I believe attest
to as well).  In my opinion, serious problems in an implementation are
often a sign that something is wrong with the language, and this may
be a case in point; indeed that's why we wanted to have a year after
issuing the report to reconsider some of the design decisions (please
re-read the preface, which in fact lists this very problem as one to
be reconsidered).

What we need are more alternative solutions.  Being more liberal (i.e.
removing the single-overloading rule) has the problems that John H.
points out.  To ameliorate this, a decent implementation could always
issue a warning when such a variable was used with more than one
overloading; but then there are those who do not believe in compiler
warnings.  However, if it could result in a drastic simplification to
the language, I could believe in it.

  -Paul