ack/modules/src/idf/idf_pkg.body

/*	SYMBOL TABLE HANDLING	*/

#include	<alloc.h>

/*	Each character of the identifier is xored with an 8-bit mask which
	depends on the position of the character; the sum of these results
	is the hash value.  The random masks are obtained from a
	congruence generator.
*/

#define	HASHSIZE	256	/* size of hashtable, must be a power of 2 */
#ifndef IDF_HSIZE
#define IDF_HSIZE	16	/* # of significant characters for hashing.
				   This is NOT the number of significant
				   characters!
				*/
#endif
#define	HASH_X		0253	/* Knuth's X */
#define	HASH_A		77	/* Knuth's a */
#define	HASH_C		153	/* Knuth's c */

#define	HASHMASK		(HASHSIZE-1)	/* since it is a power of 2 */
#define	STARTHASH(hs)		(hs = 0)
#define	ENHASH(hs,ch,hm)	(hs += (ch ^ hm))
#define	STOPHASH(hs)		(hs &= HASHMASK)

static char hmask[IDF_HSIZE];

static struct idf *id_hashtable[HASHSIZE];
	/*	All identifiers can in principle be reached through
		id_hashtable; id_hashtable[hc] is the start of a chain of
		idf's whose tags all hash to hc.
		Any identifier is entered into this
		list, regardless of the nature of its declaration
		(variable, selector, structure tag, etc.).
	*/

static struct idf *
new_idf(tg, size, cpy)
	register char *tg;
	register int size;
{
	static int nidf;
	static struct idf *pidf;
#define NIDS 50
#define IBUFSIZ	2048
	static unsigned int icnt;
	static char *ip;
	register char *p;


	if (! nidf--) {
		nidf += NIDS;
		pidf = (struct idf *) Malloc(NIDS * sizeof (struct idf));
		clear((char *) pidf, NIDS * sizeof(struct idf));
	}

	if (cpy) {
		if (size > icnt) {
			icnt =  size > IBUFSIZ ? size : IBUFSIZ;
			p = Malloc(icnt);
		}
		else p = ip;
		icnt -= size;
		pidf->id_text = p;
		while (size--) {
			*p++ = *tg++;
		}
		ip = p;
	}
	else	pidf->id_text = tg;
	return pidf++;
}

#ifdef	IDF_DEBUG
hash_stat()
{
	register int i;

	print("Hash table tally:\n");
	for (i = 0; i < HASHSIZE; i++)	{
		register struct idf *notch = id_hashtable[i];
		register int cnt = 0;

		while (notch)	{
			cnt++;
			notch = notch->id_next;
		}
		print("%d %d\n", i, cnt);
	}
	print("End hash table tally\n");
}
#endif	IDF_DEBUG

struct idf *
str2idf(tg, cpy)
	char tg[];
{
	/*	str2idf() returns an entry in the symbol table for the
		identifier tg.  If necessary, an entry is created.
	*/
	register char *cp = tg;
	register char *phm = &hmask[0];
	struct idf **hook;
	register struct idf *notch;
	register int hash;
	int size;

	STARTHASH(hash);
	while (*cp && phm < &hmask[IDF_HSIZE]) {
		ENHASH(hash, *cp++, *phm++);
	}
	STOPHASH(hash);
	while (*cp++) /* nothing. Find end of string */ ;
	size = cp - tg;

	/*	The tag tg with length size and known hash value hash is
		looked up in the identifier table; if not found, it is
		entered if cpy >= 0. A pointer to it is returned.
		Notice that the chains of idf's are sorted alphabetically.
	*/
	hook = &id_hashtable[hash];

	while ((notch = *hook))	{
		register char *s1 = tg;
		int cmp;

		cp = notch->id_text;

		while (!(cmp = (*s1 - *cp++))) {
			if (*s1++ == '\0') {
				break;
			}
		}

		if (cmp == 0) return notch;
		if (cmp < 0) break;
		hook = &notch->id_next;
	}
	/* a new struct idf must be inserted at the hook */
	if (cpy < 0) return 0;
	notch = new_idf(tg, size, cpy);
	notch->id_next = *hook;
	*hook = notch;		/* hooked in */
	return notch;
}

init_idf()	{
	/*	A simple congruence random number generator, as
		described in Knuth, vol 2.
	*/
	register int rnd = HASH_X;
	register char *phm;
	
	for (phm = &hmask[0]; phm < &hmask[IDF_HSIZE];)	{
		*phm++ = rnd;
		rnd = (HASH_A * rnd + HASH_C) & HASHMASK;
	}
}