Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!samsung!munnari.oz.au!goanna!ok
From: ok@goanna.cs.rmit.oz.au (Richard A. O'Keefe)
Newsgroups: comp.lang.scheme
Subject: Re: Destructuring / pattern-matching (was: Multiple return values)
Message-ID: <5261@goanna.cs.rmit.oz.au>
Date: 15 Apr 91 10:37:38 GMT
References: <12542.9104111347@subnode.aiai.ed.ac.uk>
Organization: Comp Sci, RMIT, Melbourne, Australia
Lines: 52

In article <12542.9104111347@subnode.aiai.ed.ac.uk>, jeff@aiai.edinburgh.ac.UK (Jeff Dalton) writes:
> Another problem with patterns in general is that they usually do not
> respect the abstract structure of objects.  If something is the third
> element of a list, a pattern will depend on it being the third
> element; if something is a slot of a struct, the pattern will depend
> on it being a slot (rather than being accessed some other way).  This
> is a serious hassle in Prolog, where changing the structure of a term
> often requires changing patterns all over the program; and since the
> contents of terms are accessed positionally, it's easy to make a
> mistake.

> When people talk of the advantages of pattern-matching in Prolog and
> functional languages, they seldom mention that it has this disadvantage
> too.

(Self-advertisement:  "The Craft of Prolog" _does_ talk about this and
what to do about it.)

A possible answer in Lisp-like languages would be to introduce a
new form

	(define-pattern (<name> <parameter>...)
		<simple backquote expression>)

where <simple backquote expressions> are sufficiently restricted for
a simple program to figure out how to invert them (say: each
<parameter> name must appear exactly once in the pattern, ,@ may appear
only at the end of a list, something like that).  One also introduces a
new , variant:  ,=(<pattern name> <backquote expression>...).

If I used
	(define-pattern (assoc-link Key Val Rest)
	    `((,Key ,@Val) ,@Rest))

then
	(let ((foo `,=(assoc-link foo 9 ,bar)))
	    ...)
would have the effect of
	(let ((foo `((foo . 9) ,@bar)))
	    ...)
and
	(destructure `,=(assoc-link ,k ,v ,bar) foo
	    ...)
would have the effect of
	(destructure `((,k ,@v) ,bar) foo
	    ...)

It's certainly doable.
-- 
It is indeed manifest that dead men are formed from living ones;
but it does not follow from that, that living men are formed from dead ones.
			-- Tertullian, on reincarnation.