2 * Bacula hash table routines
4 * htable is a hash table of items (pointers). This code is
5 * adapted and enhanced from code I wrote in 1982 for a
6 * relocatable linker. At that time, the hash table size
7 * was fixed and a primary number, which essentially provides
8 * the randomness. In this program, the hash table can grow when
9 * it gets too full, so the table size here is a binary number. The
10 * hashing is provided using an idea from Tcl where the initial
11 * hash code is then "randomized" using a simple calculation from
12 * a random number generator that multiplies by a big number
13 * (I multiply by a prime number, while Tcl did not)
14 * then shifts the results down and does the binary division
15 * by masking. Increasing the size of the hash table is simple.
16 * Just create a new larger table, walk the old table and
17 * re-hash insert each entry into the new table.
20 * Kern Sibbald, July MMIII
26 Copyright (C) 2003-2004 Kern Sibbald and John Walker
28 This program is free software; you can redistribute it and/or
29 modify it under the terms of the GNU General Public License as
30 published by the Free Software Foundation; either version 2 of
31 the License, or (at your option) any later version.
33 This program is distributed in the hope that it will be useful,
34 but WITHOUT ANY WARRANTY; without even the implied warranty of
35 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
36 General Public License for more details.
38 You should have received a copy of the GNU General Public
39 License along with this program; if not, write to the Free
40 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
49 /* ===================================================================
54 * Create hash of key, stored in hash then
55 * create and return the pseudo random bucket index
57 void htable::hash_index(char *key)
60 for (char *p=key; *p; p++) {
61 hash += (hash << 3) + (uint32_t)*p;
63 /* Multiply by large prime number, take top bits, mask for remainder */
64 index = ((hash * 1103515249) >> rshift) & mask;
65 Dmsg2(100, "Leave hash_index hash=0x%x index=%d\n", hash, index);
69 htable::htable(void *item, void *link, int tsize)
71 init(item, link, tsize);
74 void htable::init(void *item, void *link, int tsize)
78 for (pwr=0; tsize; pwr++) {
81 loffset = (char *)link - (char *)item;
82 mask = ~((~0)<<pwr); /* 3 bits => table size = 8 */
83 rshift = 30 - pwr; /* start using bits 28, 29, 30 */
84 num_items = 0; /* number of entries in table */
85 buckets = 1<<pwr; /* hash table size -- power of two */
86 max_items = buckets * 4; /* allow average 4 entries per chain */
87 table = (hlink **)malloc(buckets * sizeof(hlink *));
88 memset(table, 0, buckets * sizeof(hlink *));
94 void * htable::operator new(size_t)
96 return malloc(sizeof(htable));
99 void htable::operator delete(void *tbl)
101 ((htable *)tbl)->destroy();
106 uint32_t htable::size()
117 printf("\n\nNumItems=%d\nBuckets=%d\n", num_items, buckets);
118 for (i=0; i<10; i++) {
121 for (i=0; i<(int)buckets; i++) {
125 p = (hlink *)(p->next);
135 for (i=0; i<10; i++) {
136 printf("%2d: %d\n",i, count[i]);
138 printf("max = %d\n", max);
141 void htable::grow_table()
143 Dmsg1(100, "Grow called old size = %d\n", buckets);
144 /* Setup a bigger table */
145 htable *big = (htable *)malloc(sizeof(htable));
146 big->loffset = loffset;
147 big->mask = mask<<1 | 1;
148 big->rshift = rshift - 1;
150 big->buckets = buckets * 2;
151 big->max_items = big->buckets * 4;
152 big->table = (hlink **)malloc(big->buckets * sizeof(hlink *));
153 memset(big->table, 0, big->buckets * sizeof(hlink *));
156 /* Insert all the items in the new hash table */
157 Dmsg1(100, "Before copy num_items=%d\n", num_items);
159 * We walk through the old smaller tree getting items,
160 * but since we are overwriting the colision links, we must
161 * explicitly save the item->next pointer and walk each
162 * colision chain ourselves. We do use next() for getting
163 * to the next bucket.
165 for (void *item=first(); item; ) {
166 void *ni = ((hlink *)((char *)item+loffset))->next; /* save link overwritten by insert */
167 Dmsg1(100, "Grow insert: %s\n", ((hlink *)((char *)item+loffset))->key);
168 big->insert(((hlink *)((char *)item+loffset))->key, item);
170 item = (void *)((char *)ni-loffset);
176 Dmsg1(100, "After copy new num_items=%d\n", big->num_items);
177 if (num_items != big->num_items) {
178 Dmsg0(000, "****** Big problems num_items mismatch ******\n");
181 memcpy(this, big, sizeof(htable)); /* move everything across */
183 Dmsg0(100, "Exit grow.\n");
186 bool htable::insert(char *key, void *item)
190 return false; /* already exists */
192 sm_check(__FILE__, __LINE__, false);
193 ASSERT(index < buckets);
194 Dmsg2(100, "Insert: hash=0x%x index=%d\n", (unsigned)hash, index);
195 hp = (hlink *)(((char *)item)+loffset);
196 Dmsg4(100, "Insert hp=0x%x index=%d item=0x%x offset=%u\n", (unsigned)hp,
197 index, (unsigned)item, loffset);
198 hp->next = table[index];
202 Dmsg3(100, "Insert hp->next=0x%x hp->hash=0x%x hp->key=%s\n",
203 (unsigned)hp->next, hp->hash, hp->key);
205 if (++num_items >= max_items) {
206 Dmsg2(100, "num_items=%d max_items=%d\n", num_items, max_items);
209 sm_check(__FILE__, __LINE__, false);
210 Dmsg3(100, "Leave insert index=%d num_items=%d key=%s\n", index, num_items, key);
214 void *htable::lookup(char *key)
217 for (hlink *hp=table[index]; hp; hp=(hlink *)hp->next) {
218 // Dmsg2(100, "hp=0x%x key=%s\n", (long)hp, hp->key);
219 if (hash == hp->hash && strcmp(key, hp->key) == 0) {
220 Dmsg1(100, "lookup return %x\n", ((char *)hp)-loffset);
221 return ((char *)hp)-loffset;
229 Dmsg1(100, "Enter next: walkptr=0x%x\n", (unsigned)walkptr);
231 walkptr = (hlink *)(walkptr->next);
233 while (!walkptr && walk_index < buckets) {
234 walkptr = table[walk_index++];
236 Dmsg3(100, "new walkptr=0x%x next=0x%x inx=%d\n", (unsigned)walkptr,
237 (unsigned)(walkptr->next), walk_index-1);
241 Dmsg2(100, "next: rtn 0x%x walk_index=%d\n",
242 (unsigned)(((char *)walkptr)-loffset), walk_index);
243 return ((char *)walkptr)-loffset;
245 Dmsg0(100, "next: return NULL\n");
249 void *htable::first()
251 Dmsg0(100, "Enter first\n");
252 walkptr = table[0]; /* get first bucket */
253 walk_index = 1; /* Point to next index */
254 while (!walkptr && walk_index < buckets) {
255 walkptr = table[walk_index++]; /* go to next bucket */
257 Dmsg3(100, "first new walkptr=0x%x next=0x%x inx=%d\n", (unsigned)walkptr,
258 (unsigned)(walkptr->next), walk_index-1);
262 Dmsg1(100, "Leave first walkptr=0x%x\n", (unsigned)walkptr);
263 return ((char *)walkptr)-loffset;
265 Dmsg0(100, "Leave first walkptr=NULL\n");
269 /* Destroy the table and its contents */
270 void htable::destroy()
282 Dmsg0(100, "Done destroy.\n");
300 MYJCR *save_jcr = NULL, *item;
304 jcrtbl = (htable *)malloc(sizeof(htable));
305 jcrtbl->init(jcr, &jcr->link, NITEMS);
307 Dmsg1(000, "Inserting %d items\n", NITEMS);
308 for (int i=0; i<NITEMS; i++) {
309 sprintf(mkey, "This is htable item %d", i);
310 jcr = (MYJCR *)malloc(sizeof(MYJCR));
311 Dmsg2(100, "link=0x%x jcr=0x%x\n", (unsigned)&jcr->link, (unsigned)jcr);
312 jcr->key = bstrdup(mkey);
314 jcrtbl->insert(jcr->key, jcr);
319 if (!(item = (MYJCR *)jcrtbl->lookup(save_jcr->key))) {
320 printf("Bad news: %s not found.\n", save_jcr->key);
322 printf("Item 10's key is: %s\n", item->key);
326 printf("Walk the hash table:\n");
327 foreach_htable (jcr, jcrtbl) {
328 // printf("htable item = %s\n", jcr->key);
332 printf("Got %d items -- %s\n", count, count==NITEMS?"OK":"***ERROR***");
333 printf("Calling destroy\n");
337 printf("Freed jcrtbl\n");