/*===============================<o>=====================================Copyright 1996, 1997, 2004 Ian Kaplan, Bear Products International,www.bearcave.com.All Rights ReservedYou may use this software in software components for which you donot collect money (e.g., non-commercial software).  All commercialuse is reserved.===============================<o>=====================================*/#define NDEBUG 1#include "sysinc.h"#include "hash_serv.h"#define SIZE_INT 4/**   This function is passed a string.  It returns a hash key   formed from the string.  An attempt is made to make this   key as unique as possible.*/unsigned int hash_services::str_to_uint( const char *s ){    int i, byte_ix, word_cnt;    char c;    union {	unsigned char c[ SIZE_INT ];	unsigned int  u;    } val;    val.u = 0;    word_cnt = 0;    for (i = 0, byte_ix = 0; i < strlen( s ); i++, byte_ix = (byte_ix + 1) % SIZE_INT) {	if (i > 0 && (i % SIZE_INT) == 0) {	    word_cnt++; 	}	c = s[ i ];	// Every odd word, shift the character up by one bit        // It turns out that this makes a big difference in the        // uniqueness of the key generated from a string	if (word_cnt & 0x1) {	    c = c << 1;	}	val.c[ byte_ix ] ^= c;    }  return val.u;} // str_to_uint#undef SIZE_INT/**   Use a shift algorithm to distribute the key (e.g., make it more   pseudo-random).  I actually tried several different algorithms to   do this and this one worked the best.  It was suggested by   Brent Gregory at Synopsys, although any mistakes in translating his    suggestion to code are mine.  The code below is a loop, which has   been in-lined for speed.*/unsigned int hash_services::hash_value( const char *name ){  unsigned int val, new_val;    val = str_to_uint( name );  new_val = val ^ (val << 1);  new_val = new_val ^ (val << 2);  new_val = new_val ^ (val << 3);  new_val = new_val ^ (val << 4);  new_val = new_val ^ (val << 5);  new_val = new_val ^ (val << 6);  new_val = new_val ^ (val << 7);  new_val = new_val ^ (val << 8);  new_val = new_val ^ (val << 9);  new_val = new_val ^ (val << 10);  new_val = new_val ^ (val << 11);  new_val = new_val ^ (val << 12);  new_val = new_val ^ (val << 13);  new_val = new_val ^ (val << 14);  new_val = new_val ^ (val << 15);  new_val = new_val ^ (val << 16);  return new_val;  // adjust for hash table size} // hash_value