Do all bigram processing here, instead of in updatedb.csh; saves
a few cycles (it's runs like 2 seconds instead of 16) and the algorithm is hopefully a bit more clear
This commit is contained in:
parent
41053184b6
commit
a46f4702f8
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: locate.bigram.c,v 1.6 1997/10/19 04:11:52 lukem Exp $ */
|
||||
/* $NetBSD: locate.bigram.c,v 1.7 2000/03/20 19:17:35 jdolecek Exp $ */
|
||||
|
||||
/*
|
||||
* Copyright (c) 1989, 1993
|
||||
|
@ -46,7 +46,7 @@ __COPYRIGHT("@(#) Copyright (c) 1989, 1993\n\
|
|||
#if 0
|
||||
static char sccsid[] = "@(#)locate.bigram.c 8.2 (Berkeley) 4/28/95";
|
||||
#endif
|
||||
__RCSID("$NetBSD: locate.bigram.c,v 1.6 1997/10/19 04:11:52 lukem Exp $");
|
||||
__RCSID("$NetBSD: locate.bigram.c,v 1.7 2000/03/20 19:17:35 jdolecek Exp $");
|
||||
#endif /* not lint */
|
||||
|
||||
/*
|
||||
|
@ -57,12 +57,53 @@ __RCSID("$NetBSD: locate.bigram.c,v 1.6 1997/10/19 04:11:52 lukem Exp $");
|
|||
*/
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <sys/param.h> /* for MAXPATHLEN */
|
||||
|
||||
char buf1[MAXPATHLEN] = " ";
|
||||
char buf2[MAXPATHLEN];
|
||||
|
||||
int main __P((int, char **));
|
||||
static int compare_bigrams __P((const void *, const void *));
|
||||
static void add_bigram __P((u_char, u_char));
|
||||
|
||||
static char buf1[MAXPATHLEN] = " ";
|
||||
static char buf2[MAXPATHLEN];
|
||||
|
||||
struct bigram {
|
||||
int count;
|
||||
u_char b1, b2; /* needed for final sorting */
|
||||
};
|
||||
|
||||
struct bigram bigrams[256 * 256];
|
||||
|
||||
static void
|
||||
add_bigram(i1, i2)
|
||||
u_char i1, i2;
|
||||
{
|
||||
if (i1 == '\n') {
|
||||
bigrams[0].count++;
|
||||
i1 = 0;
|
||||
}
|
||||
if (i2 == '\n') {
|
||||
bigrams[0].count++;
|
||||
i2 = i1;
|
||||
i1 = 0;
|
||||
}
|
||||
|
||||
bigrams[(i1<<8)+i2].count++;
|
||||
}
|
||||
|
||||
static int
|
||||
compare_bigrams(item1, item2)
|
||||
const void *item1, *item2;
|
||||
{
|
||||
const struct bigram *it1=item1, *it2=item2;
|
||||
|
||||
if (it1->count != it2->count)
|
||||
return it2->count - it1->count;
|
||||
else if (it1->b1 != it2->b1)
|
||||
return it2->b1 - it1->b1;
|
||||
else
|
||||
return it2->b2 - it2->b2;
|
||||
}
|
||||
|
||||
int
|
||||
main(argc, argv)
|
||||
|
@ -71,6 +112,15 @@ main(argc, argv)
|
|||
{
|
||||
char *cp;
|
||||
char *oldpath = buf1, *path = buf2;
|
||||
struct bigram *bg;
|
||||
int i;
|
||||
|
||||
/* initialize bigram array */
|
||||
memset(bigrams, 0, sizeof(bigrams));
|
||||
for(i=0; i < 65536; i++) {
|
||||
bigrams[i].b1 = i / 256;
|
||||
bigrams[i].b2 = i % 256;
|
||||
}
|
||||
|
||||
while ( fgets ( path, sizeof(buf2), stdin ) != NULL ) {
|
||||
|
||||
|
@ -78,18 +128,31 @@ main(argc, argv)
|
|||
for ( cp = path; *cp == *oldpath; cp++, oldpath++ )
|
||||
if ( *oldpath == '\0' )
|
||||
break;
|
||||
|
||||
/*
|
||||
* output post-residue bigrams only
|
||||
*/
|
||||
while ( *cp != '\0' && *(cp + 1) != '\0' ) {
|
||||
putchar ( *cp++ );
|
||||
putchar ( *cp++ );
|
||||
putchar ( '\n' );
|
||||
}
|
||||
if ( path == buf1 ) /* swap pointers */
|
||||
for(; cp[0] != '\0' && cp[1] != '\0'; cp += 2)
|
||||
add_bigram((u_char)cp[0], (u_char)cp[1]);
|
||||
|
||||
if (path == buf1) /* swap pointers */
|
||||
path = buf2, oldpath = buf1;
|
||||
else
|
||||
path = buf1, oldpath = buf2;
|
||||
}
|
||||
|
||||
/* sort the bigrams by how many times it appeared and their value */
|
||||
heapsort((void *)bigrams, 256 * 256, sizeof(struct bigram),
|
||||
compare_bigrams);
|
||||
|
||||
/* write 128 most frequent bigrams out */
|
||||
bg = bigrams;
|
||||
for (i = 0; i < 128 && bg->count > 0; i++, bg++) {
|
||||
if (bg->b1 != '\0')
|
||||
fputc(bg->b1, stdout);
|
||||
if (bg->b2 != '\0')
|
||||
fputc(bg->b2, stdout);
|
||||
}
|
||||
|
||||
return (0);
|
||||
}
|
||||
|
|
Loading…
Reference in New Issue