|
|
researchv8 Dan Cross
/* @(#)hash.c 1.1 */
#include "hash.h"
#define LOCHWIDTH 3
#define HICHWIDTH 3
#define CHARWIDTH (LOCHWIDTH+HICHWIDTH)
#define LOCHMASK ((1<<LOCHWIDTH)-1)
/* The hash function is first a substitution cipher into 6-bit
* characters. Then the string of 6-bit bytes is interpreted
* as a huge integer, and taken mod a big prime (hashsize).
* Unhashable trash produces an otherwise impossible value; it
* would be prudent not to have that value in the dictionary.
*/
/* if HASHWIDTH + CHARWIDTH < bitsizeof(long)
* one could make LOCHWIDTH=6 and HICHWIDTH=0
* and simplify accordingly; the hanky-panky
* is to avoid overflow in long multiplication
*/
#define NC 30
long hashsize = HASHSIZE;
long pow2[NC*2];
static char hashtab[] = {
-1, -1, -1, -1, -1, -1, 0, 31, /* &' */
-1, -1, -1, -1, -1, -1, -1, -1,
2, 25, 20, 35, 54, 61, 40, 39, /* 0-7 */
42, 33, -1, -1, -1, -1, -1, -1,
-1, 60, 43, 30, 5, 16, 47, 18, /* A-G */
41, 36, 51, 6, 13, 56, 55, 58,
49, 12, 59, 46, 21, 32, 63, 34,
57, 52, 3, -1, -1, -1, -1, -1,
-1, 22, 29, 8, 7, 10, 1, 28, /* a-g */
11, 62, 37, 48, 15, 50, 9, 4,
19, 38, 45, 24, 23, 26, 17, 44,
27, 14, 53, -1, -1, -1, -1, -1
};
long hash(s)
char *s;
{
register c;
register long *lp;
long h = 0;
for(lp=pow2; c = *s++&0177; ) {
if((c-=' ') >= 0 && (c=hashtab[c]) >= 0
&& lp < pow2+(sizeof pow2/sizeof *pow2)) {
h += (c&LOCHMASK) * *lp++;
h += (c>>LOCHWIDTH) * *lp++;
h %= hashsize;
} else
return (1L<<HASHWIDTH) - 1; /*trash value*/
}
return(h);
}
hashinit()
{
register i;
if(1L<<(HASHWIDTH+LOCHWIDTH)<=0
|| 1L<<(HASHWIDTH+HICHWIDTH)<=0)
abort(); /* overflow is imminent */
pow2[0] = 1L<<(HASHWIDTH-CHARWIDTH-2);
for(i=0; i<2*NC-3; i+=2) {
pow2[i+1] = (pow2[i]<<LOCHWIDTH) % hashsize;
pow2[i+2] = (pow2[i+1]<<HICHWIDTH) % hashsize;
}
}
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.