Path: utzoo!attcan!uunet!cs.utexas.edu!wuarchive!udel!princeton!drops.Princeton.EDU!pfalstad
From: pfalstad@drops.Princeton.EDU (Paul John Falstad)
Newsgroups: comp.lang.c
Subject: Re: Comparing strings...
Message-ID: <3343@idunno.Princeton.EDU>
Date: 14 Oct 90 21:45:37 GMT
References: <1990Oct13.190106.15615@ux1.cso.uiuc.edu> <3330@idunno.Princeton.EDU> <11486@alice.att.com>
Sender: news@idunno.Princeton.EDU
Organization: Princeton University, Princeton, New Jersey
Lines: 34

In article <11486@alice.att.com> ark@alice.att.com (Andrew Koenig) writes:
>In article <3330@idunno.Princeton.EDU>, pfalstad@phoenix.Princeton.EDU (Paul John Falstad) writes:
>> int strcmp(char *s,char *t)
>> {
>> 	for (; *s && *s == *t; s++,t++);
>> 	return *s-*t;
>> }
>Whether or not this works depends on your definition of the problem.
>If you really want to follow the normal lexical convention that
>the null string compares <= anything else, and you're on a machine
>on which chars can be negative, then it doesn't work.

OK.  Declare the chars unsigned then.  (I copied this with minor changes
from K&R2, so I assumed it would be right.)

Other optimizations could be made:

int strcmp(char *s,char *t)  /* declared register if you like */
{
unsigned char c,d;

	while((c = (unsigned char) *s++) == (d == (unsigned char) *t++) && c);
	return c-d;
}

I've heard of faster ways that take advantage of parallelism by
comparing a word or longword at a time.  Can't remember if it was
worth it or not.

--
pfalstad@phoenix.princeton.edu   The Germans are disputing it! Hegel is
arguing that the reality is merely an a priori adjunct of non-absolutistic
ethics; Kant, by the categorical imperative, is holding it ontologically
exists only in the imagination; and Marx is claiming it was off sides.