Path: utzoo!attcan!uunet!wuarchive!brutus.cs.uiuc.edu!apple!bionet!csd4.csd.uwm.edu!mailrus!ncar!ico!ism780c!news
From: news@ism780c.isc.com (News system)
Newsgroups: comp.lang.c
Subject: Re: swap(x,y)
Message-ID: <32074@ism780c.isc.com>
Date: 24 Aug 89 20:01:00 GMT
References: <MYwDEFy00WB842JEUx@andrew.cmu.edu> <8350@boring.cwi.nl> <8514@galbp.LBP.HARRIS.COM> <3040@solo5.cs.vu.nl>
Reply-To: marv@ism780.UUCP (Marvin Rubenstein)
Organization: Interactive Systems Corp., Santa Monica CA
Lines: 31

In article <3040@solo5.cs.vu.nl> maart@cs.vu.nl (Maarten Litmaath) writes:
>jsb@advdev.LBP.HARRIS.COM (FLEA) writes:
>\In article <8350@boring.cwi.nl> tromp@piring.cwi.nl (John Tromp) writes:
>\:		 (Timothy R. Gottschalk) writes:
>\...
>\:>       x += y;
>\:>       y = x - y;
>\:>       x -= y;
>\...
>\Sorry, folks, both of these are illegal if x and y are pointers.
>
>Or if they're structs! :-)

It is obvious to the casual observer that the above is illegal if the
operations + and - are not defined for the data.  The question really is: for
the case where the program segment (as written) compiles, will it swap the
values of x and y.  The answer is that it will work on two's complement
binary machines that ignore overflow and carry out provided x and y are the
same integral type.  It does not work when the types differ (for example
char x; long y;).  It also does not work for all values of floating types.

But most important, it is a bad piece of code for two reasons.  1) it is
obtuse and  2) it requires more space and time than the simpiler:

     temp=x; x=y; y=temp;

As may have been pointed the EXOR method may be faster then the above for
a machine like the 370 that provides a memory to memory EXOR operation on
data blocks.  But there is no way to express such an operation in C.

     Marv Rubinstein