The question below was asked in 2008 about some code from 2003. As the OP's update shows, this entire post has been obsoleted by vintage 2008 algorithms and persists here only as a historical curiosity.
I need to do a fast case-insensitive substring search in C/C++. My requirements are as follows:
Here is the current implementation I am using (taken from the GNU C Library):
/* Return the offset of one string within another.
Copyright (C) 1994,1996,1997,1998,1999,2000 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
/*
* My personal strstr() implementation that beats most other algorithms.
* Until someone tells me otherwise, I assume that this is the
* fastest implementation of strstr() in C.
* I deliberately chose not to comment it. You should have at least
* as much fun trying to understand it, as I had to write it :-).
*
* Stephen R. van den Berg, berg@pool.informatik.rwth-aachen.de */
/*
* Modified to use table lookup instead of tolower(), since tolower() isn't
* worth s*** on Windows.
*
* -- Anders Sandvig (anders@wincue.org)
*/
#if HAVE_CONFIG_H
# include <config.h>
#endif
#include <ctype.h>
#include <string.h>
typedef unsigned chartype;
char char_table[256];
void init_stristr(void)
{
int i;
char string[2];
string[1] = '\0';
for (i = 0; i < 256; i++)
{
string[0] = i;
_strlwr(string);
char_table[i] = string[0];
}
}
#define my_tolower(a) ((chartype) char_table[a])
char *
my_stristr (phaystack, pneedle)
const char *phaystack;
const char *pneedle;
{
register const unsigned char *haystack, *needle;
register chartype b, c;
haystack = (const unsigned char *) phaystack;
needle = (const unsigned char *) pneedle;
b = my_tolower (*needle);
if (b != '\0')
{
haystack--; /* possible ANSI violation */
do
{
c = *++haystack;
if (c == '\0')
goto ret0;
}
while (my_tolower (c) != (int) b);
c = my_tolower (*++needle);
if (c == '\0')
goto foundneedle;
++needle;
goto jin;
for (;;)
{
register chartype a;
register const unsigned char *rhaystack, *rneedle;
do
{
a = *++haystack;
if (a == '\0')
goto ret0;
if (my_tolower (a) == (int) b)
break;
a = *++haystack;
if (a == '\0')
goto ret0;
shloop:
;
}
while (my_tolower (a) != (int) b);
jin:
a = *++haystack;
if (a == '\0')
goto ret0;
if (my_tolower (a) != (int) c)
goto shloop;
rhaystack = haystack-- + 1;
rneedle = needle;
a = my_tolower (*rneedle);
if (my_tolower (*rhaystack) == (int) a)
do
{
if (a == '\0')
goto foundneedle;
++rhaystack;
a = my_tolower (*++needle);
if (my_tolower (*rhaystack) != (int) a)
break;
if (a == '\0')
goto foundneedle;
++rhaystack;
a = my_tolower (*++needle);
}
while (my_tolower (*rhaystack) == (int) a);
needle = rneedle; /* took the register-poor approach */
if (a == '\0')
break;
}
}
foundneedle:
return (char*) haystack;
ret0:
return 0;
}Can you make this code faster, or do you know of a better implementation?
Note: I noticed that the GNU C Library now has a new implementation of strstr(), but I am not sure how easily it can be modified to be case-insensitive, or if it is in fact faster than the old one (in my case). I also noticed that the old implementation is still used for wide character strings, so if anyone knows why, please share.
Update
Just to make things clear—in case it wasn't already—I didn't write this function, it's a part of the GNU C Library. I only modified it to be case-insensitive.
Also, thanks for the tip about strcasestr() and checking out other implementations from other sources (like OpenBSD, FreeBSD, etc.). It seems to be the way to go. The code above is from 2003, which is why I posted it here in hope for a better version being available, which apparently it is. :)
The code you posted is about half as fast as strcasestr.
$ gcc -Wall -o my_stristr my_stristr.c
steve@solaris:~/code/tmp
$ gcc -Wall -o strcasestr strcasestr.c
steve@solaris:~/code/tmp
$ ./bench ./my_stristr > my_stristr.result ; ./bench ./strcasestr > strcasestr.result;
steve@solaris:~/code/tmp
$ cat my_stristr.result
run 1... time = 6.32
run 2... time = 6.31
run 3... time = 6.31
run 4... time = 6.31
run 5... time = 6.32
run 6... time = 6.31
run 7... time = 6.31
run 8... time = 6.31
run 9... time = 6.31
run 10... time = 6.31
average user time over 10 runs = 6.3120
steve@solaris:~/code/tmp
$ cat strcasestr.result
run 1... time = 3.82
run 2... time = 3.82
run 3... time = 3.82
run 4... time = 3.82
run 5... time = 3.82
run 6... time = 3.82
run 7... time = 3.82
run 8... time = 3.82
run 9... time = 3.82
run 10... time = 3.82
average user time over 10 runs = 3.8200
steve@solaris:~/code/tmp
The main function was:
int main(void)
{
char * needle="hello";
char haystack[1024];
int i;
for(i=0;i<sizeof(haystack)-strlen(needle)-1;++i)
{
haystack[i]='A'+i%57;
}
memcpy(haystack+i,needle, strlen(needle)+1);
/*printf("%s\n%d\n", haystack, haystack[strlen(haystack)]);*/
init_stristr();
for (i=0;i<1000000;++i)
{
/*my_stristr(haystack, needle);*/
strcasestr(haystack,needle);
}
return 0;
}
It was suitably modified to test both implementations. I notice as I am typing this up I left in the init_stristr call, but it shouldn't change things too much. bench is just a simple shell script:
#!/bin/bash
function bc_calc()
{
echo $(echo "scale=4;$1" | bc)
}
time="/usr/bin/time -p"
prog="$1"
accum=0
runs=10
for a in $(jot $runs 1 $runs)
do
echo -n "run $a... "
t=$($time $prog 2>&1| grep user | awk '{print $2}')
echo "time = $t"
accum=$(bc_calc "$accum+$t")
done
echo -n "average user time over $runs runs = "
echo $(bc_calc "$accum/$runs")