Home | History | Annotate | Line # | Download | only in make
hash.c revision 1.47
      1 /*	$NetBSD: hash.c,v 1.47 2020/10/18 12:47:43 rillig Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1988, 1989, 1990 The Regents of the University of California.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to Berkeley by
      8  * Adam de Boor.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  * 3. Neither the name of the University nor the names of its contributors
     19  *    may be used to endorse or promote products derived from this software
     20  *    without specific prior written permission.
     21  *
     22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     32  * SUCH DAMAGE.
     33  */
     34 
     35 /*
     36  * Copyright (c) 1988, 1989 by Adam de Boor
     37  * Copyright (c) 1989 by Berkeley Softworks
     38  * All rights reserved.
     39  *
     40  * This code is derived from software contributed to Berkeley by
     41  * Adam de Boor.
     42  *
     43  * Redistribution and use in source and binary forms, with or without
     44  * modification, are permitted provided that the following conditions
     45  * are met:
     46  * 1. Redistributions of source code must retain the above copyright
     47  *    notice, this list of conditions and the following disclaimer.
     48  * 2. Redistributions in binary form must reproduce the above copyright
     49  *    notice, this list of conditions and the following disclaimer in the
     50  *    documentation and/or other materials provided with the distribution.
     51  * 3. All advertising materials mentioning features or use of this software
     52  *    must display the following acknowledgement:
     53  *	This product includes software developed by the University of
     54  *	California, Berkeley and its contributors.
     55  * 4. Neither the name of the University nor the names of its contributors
     56  *    may be used to endorse or promote products derived from this software
     57  *    without specific prior written permission.
     58  *
     59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     69  * SUCH DAMAGE.
     70  */
     71 
     72 /*
     73  * This module contains routines to manipulate a hash table.
     74  * See hash.h for a definition of the structure of the hash
     75  * table.  Hash tables grow automatically as the amount of
     76  * information increases.
     77  */
     78 
     79 #include "make.h"
     80 
     81 /*	"@(#)hash.c	8.1 (Berkeley) 6/6/93"	*/
     82 MAKE_RCSID("$NetBSD: hash.c,v 1.47 2020/10/18 12:47:43 rillig Exp $");
     83 
     84 /*
     85  * The ratio of # entries to # buckets at which we rebuild the table to
     86  * make it larger.
     87  */
     88 #define rebuildLimit 3
     89 
     90 /* This hash function matches Gosling's emacs. */
     91 static unsigned int
     92 hash(const char *key, size_t *out_keylen)
     93 {
     94 	unsigned h = 0;
     95 	const char *p = key;
     96 	while (*p != '\0')
     97 		h = (h << 5) - h + (unsigned char)*p++;
     98 	if (out_keylen != NULL)
     99 		*out_keylen = (size_t)(p - key);
    100 	return h;
    101 }
    102 
    103 static HashEntry *
    104 HashTable_Find(HashTable *t, unsigned int h, const char *key)
    105 {
    106 	HashEntry *e;
    107 	unsigned int chainlen = 0;
    108 
    109 #ifdef DEBUG_HASH_LOOKUP
    110 	DEBUG4(HASH, "%s: %p h=%x key=%s\n", __func__, t, h, key);
    111 #endif
    112 
    113 	for (e = t->buckets[h & t->bucketsMask]; e != NULL; e = e->next) {
    114 		chainlen++;
    115 		if (e->key_hash == h && strcmp(e->key, key) == 0)
    116 			break;
    117 	}
    118 
    119 	if (chainlen > t->maxchain)
    120 		t->maxchain = chainlen;
    121 
    122 	return e;
    123 }
    124 
    125 /* Sets up the hash table. */
    126 void
    127 Hash_InitTable(HashTable *t)
    128 {
    129 	unsigned int n = 16, i;
    130 	HashEntry **hp;
    131 
    132 	t->numEntries = 0;
    133 	t->maxchain = 0;
    134 	t->bucketsSize = n;
    135 	t->bucketsMask = n - 1;
    136 	t->buckets = hp = bmake_malloc(sizeof(*hp) * n);
    137 	for (i = 0; i < n; i++)
    138 		hp[i] = NULL;
    139 }
    140 
    141 /* Removes everything from the hash table and frees up the memory space it
    142  * occupied (except for the space in the HashTable structure). */
    143 void
    144 Hash_DeleteTable(HashTable *t)
    145 {
    146 	HashEntry **hp, *h, *nexth = NULL;
    147 	int i;
    148 
    149 	for (hp = t->buckets, i = (int)t->bucketsSize; --i >= 0;) {
    150 		for (h = *hp++; h != NULL; h = nexth) {
    151 			nexth = h->next;
    152 			free(h);
    153 		}
    154 	}
    155 	free(t->buckets);
    156 
    157 	/*
    158 	 * Set up the hash table to cause memory faults on any future access
    159 	 * attempts until re-initialization.
    160 	 */
    161 	t->buckets = NULL;
    162 }
    163 
    164 /* Searches the hash table for an entry corresponding to the key.
    165  *
    166  * Input:
    167  *	t		Hash table to search.
    168  *	key		A hash key.
    169  *
    170  * Results:
    171  *	Returns a pointer to the entry for key, or NULL if the table contains
    172  *	no entry for the key.
    173  */
    174 HashEntry *
    175 Hash_FindEntry(HashTable *t, const char *key)
    176 {
    177 	unsigned int h = hash(key, NULL);
    178 	return HashTable_Find(t, h, key);
    179 }
    180 
    181 void *
    182 Hash_FindValue(HashTable *t, const char *key)
    183 {
    184 	HashEntry *he = Hash_FindEntry(t, key);
    185 	return he != NULL ? he->value : NULL;
    186 }
    187 
    188 /* Makes a new hash table that is larger than the old one. The entire hash
    189  * table is moved, so any bucket numbers from the old table become invalid. */
    190 static void
    191 RebuildTable(HashTable *t)
    192 {
    193 	HashEntry *e, *next = NULL, **hp, **xp;
    194 	int i;
    195 	unsigned int mask, oldsize, newsize;
    196 	HashEntry **oldhp;
    197 
    198 	oldhp = t->buckets;
    199 	oldsize = t->bucketsSize;
    200 	newsize = oldsize << 1;
    201 	t->bucketsSize = (unsigned int)newsize;
    202 	t->bucketsMask = mask = newsize - 1;
    203 	t->buckets = hp = bmake_malloc(sizeof(*hp) * newsize);
    204 	i = (int)newsize;
    205 	while (--i >= 0)
    206 		*hp++ = NULL;
    207 	for (hp = oldhp, i = (int)oldsize; --i >= 0;) {
    208 		for (e = *hp++; e != NULL; e = next) {
    209 			next = e->next;
    210 			xp = &t->buckets[e->key_hash & mask];
    211 			e->next = *xp;
    212 			*xp = e;
    213 		}
    214 	}
    215 	free(oldhp);
    216 	DEBUG5(HASH, "%s: %p size=%d entries=%d maxchain=%d\n",
    217 	       __func__, t, t->bucketsSize, t->numEntries, t->maxchain);
    218 	t->maxchain = 0;
    219 }
    220 
    221 /* Searches the hash table for an entry corresponding to the key.
    222  * If no entry is found, then one is created.
    223  *
    224  * Input:
    225  *	t		Hash table to search.
    226  *	key		A hash key.
    227  *	newPtr		Filled with TRUE if new entry created,
    228  *			FALSE otherwise.
    229  */
    230 HashEntry *
    231 Hash_CreateEntry(HashTable *t, const char *key, Boolean *newPtr)
    232 {
    233 	HashEntry *e;
    234 	unsigned h;
    235 	size_t keylen;
    236 	HashEntry **hp;
    237 
    238 	h = hash(key, &keylen);
    239 	e = HashTable_Find(t, h, key);
    240 	if (e) {
    241 		if (newPtr != NULL)
    242 			*newPtr = FALSE;
    243 		return e;
    244 	}
    245 
    246 	/*
    247 	 * The desired entry isn't there.  Before allocating a new entry,
    248 	 * expand the table if necessary (and this changes the resulting
    249 	 * bucket chain).
    250 	 */
    251 	if (t->numEntries >= rebuildLimit * t->bucketsSize)
    252 		RebuildTable(t);
    253 
    254 	e = bmake_malloc(sizeof(*e) + keylen);
    255 	hp = &t->buckets[h & t->bucketsMask];
    256 	e->next = *hp;
    257 	*hp = e;
    258 	Hash_SetValue(e, NULL);
    259 	e->key_hash = h;
    260 	memcpy(e->key, key, keylen + 1);
    261 	t->numEntries++;
    262 
    263 	if (newPtr != NULL)
    264 		*newPtr = TRUE;
    265 	return e;
    266 }
    267 
    268 /* Delete the given hash table entry and free memory associated with it. */
    269 void
    270 Hash_DeleteEntry(HashTable *t, HashEntry *e)
    271 {
    272 	HashEntry **hp, *p;
    273 
    274 	for (hp = &t->buckets[e->key_hash & t->bucketsMask];
    275 	     (p = *hp) != NULL; hp = &p->next) {
    276 		if (p == e) {
    277 			*hp = p->next;
    278 			free(p);
    279 			t->numEntries--;
    280 			return;
    281 		}
    282 	}
    283 	abort();
    284 }
    285 
    286 /* Set things up for iterating over all entries in the hash table. */
    287 void
    288 HashIter_Init(HashIter *hi, HashTable *t)
    289 {
    290 	hi->table = t;
    291 	hi->nextBucket = 0;
    292 	hi->entry = NULL;
    293 }
    294 
    295 /* Return the next entry in the hash table, or NULL if the end of the table
    296  * is reached. */
    297 HashEntry *
    298 HashIter_Next(HashIter *hi)
    299 {
    300 	HashEntry *e;
    301 	HashTable *t = hi->table;
    302 
    303 	/*
    304 	 * The entry field points to the most recently returned
    305 	 * entry, or is NULL if we are starting up.  If not NULL, we have
    306 	 * to start at the next one in the chain.
    307 	 */
    308 	e = hi->entry;
    309 	if (e != NULL)
    310 		e = e->next;
    311 	/*
    312 	 * If the chain ran out, or if we are starting up, we need to
    313 	 * find the next nonempty chain.
    314 	 */
    315 	while (e == NULL) {
    316 		if (hi->nextBucket >= t->bucketsSize)
    317 			return NULL;
    318 		e = t->buckets[hi->nextBucket++];
    319 	}
    320 	hi->entry = e;
    321 	return e;
    322 }
    323 
    324 void
    325 Hash_DebugStats(HashTable *t, const char *name)
    326 {
    327 	DEBUG4(HASH, "HashTable %s: size=%u numEntries=%u maxchain=%u\n",
    328 	       name, t->bucketsSize, t->numEntries, t->maxchain);
    329 }
    330