Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!rutgers!iuvax!bsu-cs!dhesi
From: dhesi@bsu-cs.UUCP (Rahul Dhesi)
Newsgroups: comp.lang.c
Subject: Re: typeof isn't enough to define swap correctly
Message-ID: <1106@bsu-cs.UUCP>
Date: Fri, 4-Sep-87 18:08:50 EDT
Article-I.D.: bsu-cs.1106
Posted: Fri Sep  4 18:08:50 1987
Date-Received: Sat, 5-Sep-87 20:48:40 EDT
References: <557@rocky.STANFORD.EDU>
Reply-To: dhesi@bsu-cs.UUCP (Rahul Dhesi)
Organization: CS Dept, Ball St U, Muncie, Indiana
Lines: 34
Keywords: a can of worms by any other name

In article <557@rocky.STANFORD.EDU> andy@rocky.UUCP (Andy Freeman) quotes
Walter Bright's swap macro:
>> 	#define swap(a,b)	{ typeof(a) tmp; tmp = a; a = b; b = tmp; }

and says,

     Even given typeof, this definition of swap is buggy...I've never
     figured out how to define a C swap macro that works - I'm willing
     to believe it is impossible.

Way back, when Algol's call-by-name parameter passing method was
invented, they never realized what a can of worms it would be.  One of
these worms is the proven impossibility of writing a general swap
function in Algol.  (I think I'm thinking of Algol68, but it might be
true of earlier versions too.)

As far as I can tell, the swap macro is the equivalent of an Algol
subprogram that uses call-by-name, so it's not surprising that a
generized swap macro seems to be impossible.

The trouble occurs in things like

     swap(i, a[i]);

because the macro changes i before it changes a[i].  An unexpected effect
is:

     a[a[i]] = something;

(P.S.  Niklaus Wirth has a joke about how his name is pronounced:  you
can call him by name, pronounced Veert, or call him by value,
pronounced Worth.)
-- 
Rahul Dhesi         UUCP:  {ihnp4,seismo}!{iuvax,pur-ee}!bsu-cs!dhesi