Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!helios!bcm!dimacs.rutgers.edu!rutgers!modus!gear!am!alex
From: alex@am.sublink.org (Alex Martelli)
Newsgroups: comp.unix.shell
Subject: Re: string comparisons
Message-ID: <1991Mar19.002033.1305@am.sublink.org>
Date: 19 Mar 91 00:20:33 GMT
References: <2847@kluge.fiu.edu> <1991Mar14.225901.3973@mrspoc.Transact.COM> <1991Mar17.061632.4096@cbnewsh.att.com>
Organization: Premiata Famiglia Martelli & Figli
Lines: 94

wcs@cbnewsh.att.com (Bill Stewart 908-949-0705 erebus.att.com!wcs) writes:
:In article <1991Mar14.225901.3973@mrspoc.Transact.COM> steven@Transact.COM writes:
:]acmfiu@serss0.fiu.edu (ACMFIU) writes:
:]>under csh or sh, how do i compare two strings?
:]Either use expr(1) or awk(1).  Depends on how elaborate you want to get.
:The test program (aka "[") is accessible from either shell, and is a
:builtin for modern shells like ksh.  In sh, just do
:    if [ "$foo" = "$bar" ]
:    then true-stuff
:    else false-stuff

I believe that the following approach should be often faster:

case $foo in
    "$bar")    true-stuff ;;
    *)        false-stuff ;;
esac

since "case", and the "globbing-like" behavior of its pattern-matching,
IS guaranteed to be builtin to sh.  NOT earth-shattering differences,
mind you!, but given these two scripts on my 386/20, Interactive 2.2:
---cut: pep
#!/bin/sh
fai='abe*cou'
ogu='abe*cou'
for i in 0 1 2 3 4 5 6 7 8 9
do
	case $fai in
		"$ogu")	echo "$fai" si! ;;
		*)		echo "$fai" no! ;;
	esac >/dev/null 
done
fai='abetocou'
for i in 0 1 2 3 4 5 6 7 8 9
do
	case $fai in
		"$ogu")	echo "$fai" si! ;;
		*)		echo "$fai" no! ;;
	esac >/dev/null 
done
---cut: pap
#!/bin/sh
fai='abe*cou'
ogu='abe*cou'
for i in 0 1 2 3 4 5 6 7 8 9
do
	if [ $fai = $ogu ]
		then	echo "$fai" si! ;
		else	echo "$fai" no! ;
	fi >/dev/null 
done
fai='abetocou'
for i in 0 1 2 3 4 5 6 7 8 9
do
	if [ $fai = $ogu ]
		then	echo "$fai" si! ;
		else	echo "$fai" no! ;
	fi >/dev/null 
done
---end cut

/bin/time says, for pap:
real        1.6
user        0.3
sys         1.2

while for pep:
real        0.5
user        0.0
sys         0.4

with very repeatable measurements (+-0.1 second max on each number
over quite a few repetitions), so the case would appear to be 2 to 3
times faster than the if [ ] (despite the "dilution" of the for,
echos to /dev/null, and other trinkets in common between the two
scripts).

Besides, the globbing in the case statement turns out handy quite
often... (sometimes it's a bother instead - see the above quotes around
"$ogu").

I learned Unix on underpowered/overloaded machines, so my istincts
push me to "optimize" even such trifles.  I realize that "if [ ]"
will look more readable/understandable/maintainable than "case" to
most people, and no doubt it may even be faster in shells with test
built-in... so please don't flame on these accounts!  I'm just trying
to suggest an alternative technique that may avoid some extra forks
in more "traditional" sh's.

-- 
Alex Martelli - (home snailmail:) v. Barontini 27, 40138 Bologna, ITALIA
Email: (work:) martelli@cadlab.sublink.org, (home:) alex@am.sublink.org
Phone: (work:) ++39 (51) 371099, (home:) ++39 (51) 250434; 
Fax: ++39 (51) 366964 (work only), Fidonet: 332/401.3 (home only).