dict.c

   1 /* Hash Tables Implementation.
   2  *
   3  * This file implements in memory hash tables with insert/del/replace/find/
   4  * get-random-element operations. Hash tables will auto resize if needed
   5  * tables of power of two in size are used, collisions are handled by
   6  * chaining. See the source code for more information... :)
   7  *
   8  * Copyright (c) 2006-2009, Salvatore Sanfilippo <antirez at gmail dot com>
   9  * All rights reserved.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions are met:
  13  *
  14  *   * Redistributions of source code must retain the above copyright notice,
  15  *     this list of conditions and the following disclaimer.
  16  *   * Redistributions in binary form must reproduce the above copyright
  17  *     notice, this list of conditions and the following disclaimer in the
  18  *     documentation and/or other materials provided with the distribution.
  19  *   * Neither the name of Redis nor the names of its contributors may be used
  20  *     to endorse or promote products derived from this software without
  21  *     specific prior written permission.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  24  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  26  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  27  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  33  * POSSIBILITY OF SUCH DAMAGE.
  34  */
  35
  36 #include "fmacros.h"
  37
  38 #include <stdio.h>
  39 #include <stdlib.h>
  40 #include <string.h>
  41 #include <stdarg.h>
  42 #include <assert.h>
  43 #include <limits.h>
  44
  45 #include "dict.h"
  46 #include "zmalloc.h"
  47
  48 /* ---------------------------- Utility funcitons --------------------------- */
  49
  50 static void _dictPanic(const char *fmt, ...)
  51 {
  52     va_list ap;
  53
  54     va_start(ap, fmt);
  55     fprintf(stderr, "\nDICT LIBRARY PANIC: ");
  56     vfprintf(stderr, fmt, ap);
  57     fprintf(stderr, "\n\n");
  58     va_end(ap);
  59 }
  60
  61 /* ------------------------- Heap Management Wrappers------------------------ */
  62
  63 static void *_dictAlloc(size_t size)
  64 {
  65     void *p = zmalloc(size);
  66     if (p == NULL)
  67         _dictPanic("Out of memory");
  68     return p;
  69 }
  70
  71 static void _dictFree(void *ptr) {
  72     zfree(ptr);
  73 }
  74
  75 /* -------------------------- private prototypes ---------------------------- */
  76
  77 static int _dictExpandIfNeeded(dict *ht);
  78 static unsigned long _dictNextPower(unsigned long size);
  79 static int _dictKeyIndex(dict *ht, const void *key);
  80 static int _dictInit(dict *ht, dictType *type, void *privDataPtr);
  81
  82 /* -------------------------- hash functions -------------------------------- */
  83
  84 /* Thomas Wang's 32 bit Mix Function */
  85 unsigned int dictIntHashFunction(unsigned int key)
  86 {
  87     key += ~(key << 15);
  88     key ^=  (key >> 10);
  89     key +=  (key << 3);
  90     key ^=  (key >> 6);
  91     key += ~(key << 11);
  92     key ^=  (key >> 16);
  93     return key;
  94 }
  95
  96 /* Identity hash function for integer keys */
  97 unsigned int dictIdentityHashFunction(unsigned int key)
  98 {
  99     return key;
 100 }
 101
 102 /* Generic hash function (a popular one from Bernstein).
 103  * I tested a few and this was the best. */
 104 unsigned int dictGenHashFunction(const unsigned char *buf, int len) {
 105     unsigned int hash = 5381;
 106
 107     while (len--)
 108         hash = ((hash << 5) + hash) + (*buf++); /* hash * 33 + c */
 109     return hash;
 110 }
 111
 112 /* ----------------------------- API implementation ------------------------- */
 113
 114 /* Reset an hashtable already initialized with ht_init().
 115  * NOTE: This function should only called by ht_destroy(). */
 116 static void _dictReset(dict *ht)
 117 {
 118     ht->table = NULL;
 119     ht->size = 0;
 120     ht->sizemask = 0;
 121     ht->used = 0;
 122 }
 123
 124 /* Create a new hash table */
 125 dict *dictCreate(dictType *type,
 126         void *privDataPtr)
 127 {
 128     dict *ht = _dictAlloc(sizeof(*ht));
 129
 130     _dictInit(ht,type,privDataPtr);
 131     return ht;
 132 }
 133
 134 /* Initialize the hash table */
 135 int _dictInit(dict *ht, dictType *type,
 136         void *privDataPtr)
 137 {
 138     _dictReset(ht);
 139     ht->type = type;
 140     ht->privdata = privDataPtr;
 141     return DICT_OK;
 142 }
 143
 144 /* Resize the table to the minimal size that contains all the elements,
 145  * but with the invariant of a USER/BUCKETS ration near to <= 1 */
 146 int dictResize(dict *ht)
 147 {
 148     int minimal = ht->used;
 149
 150     if (minimal < DICT_HT_INITIAL_SIZE)
 151         minimal = DICT_HT_INITIAL_SIZE;
 152     return dictExpand(ht, minimal);
 153 }
 154
 155 /* Expand or create the hashtable */
 156 int dictExpand(dict *ht, unsigned long size)
 157 {
 158     dict n; /* the new hashtable */
 159     unsigned long realsize = _dictNextPower(size), i;
 160
 161     /* the size is invalid if it is smaller than the number of
 162      * elements already inside the hashtable */
 163     if (ht->used > size)
 164         return DICT_ERR;
 165
 166     _dictInit(&n, ht->type, ht->privdata);
 167     n.size = realsize;
 168     n.sizemask = realsize-1;
 169     n.table = _dictAlloc(realsize*sizeof(dictEntry*));
 170
 171     /* Initialize all the pointers to NULL */
 172     memset(n.table, 0, realsize*sizeof(dictEntry*));
 173
 174     /* Copy all the elements from the old to the new table:
 175      * note that if the old hash table is empty ht->size is zero,
 176      * so dictExpand just creates an hash table. */
 177     n.used = ht->used;
 178     for (i = 0; i < ht->size && ht->used > 0; i++) {
 179         dictEntry *he, *nextHe;
 180
 181         if (ht->table[i] == NULL) continue;
 182
 183         /* For each hash entry on this slot... */
 184         he = ht->table[i];
 185         while(he) {
 186             unsigned int h;
 187
 188             nextHe = he->next;
 189             /* Get the new element index */
 190             h = dictHashKey(ht, he->key) & n.sizemask;
 191             he->next = n.table[h];
 192             n.table[h] = he;
 193             ht->used--;
 194             /* Pass to the next element */
 195             he = nextHe;
 196         }
 197     }
 198     assert(ht->used == 0);
 199     _dictFree(ht->table);
 200
 201     /* Remap the new hashtable in the old */
 202     *ht = n;
 203     return DICT_OK;
 204 }
 205
 206 /* Add an element to the target hash table */
 207 int dictAdd(dict *ht, void *key, void *val)
 208 {
 209     int index;
 210     dictEntry *entry;
 211
 212     /* Get the index of the new element, or -1 if
 213      * the element already exists. */
 214     if ((index = _dictKeyIndex(ht, key)) == -1)
 215         return DICT_ERR;
 216
 217     /* Allocates the memory and stores key */
 218     entry = _dictAlloc(sizeof(*entry));
 219     entry->next = ht->table[index];
 220     ht->table[index] = entry;
 221
 222     /* Set the hash entry fields. */
 223     dictSetHashKey(ht, entry, key);
 224     dictSetHashVal(ht, entry, val);
 225     ht->used++;
 226     return DICT_OK;
 227 }
 228
 229 /* Add an element, discarding the old if the key already exists */
 230 int dictReplace(dict *ht, void *key, void *val)
 231 {
 232     dictEntry *entry;
 233
 234     /* Try to add the element. If the key
 235      * does not exists dictAdd will suceed. */
 236     if (dictAdd(ht, key, val) == DICT_OK)
 237         return DICT_OK;
 238     /* It already exists, get the entry */
 239     entry = dictFind(ht, key);
 240     /* Free the old value and set the new one */
 241     dictFreeEntryVal(ht, entry);
 242     dictSetHashVal(ht, entry, val);
 243     return DICT_OK;
 244 }
 245
 246 /* Search and remove an element */
 247 static int dictGenericDelete(dict *ht, const void *key, int nofree)
 248 {
 249     unsigned int h;
 250     dictEntry *he, *prevHe;
 251
 252     if (ht->size == 0)
 253         return DICT_ERR;
 254     h = dictHashKey(ht, key) & ht->sizemask;
 255     he = ht->table[h];
 256
 257     prevHe = NULL;
 258     while(he) {
 259         if (dictCompareHashKeys(ht, key, he->key)) {
 260             /* Unlink the element from the list */
 261             if (prevHe)
 262                 prevHe->next = he->next;
 263             else
 264                 ht->table[h] = he->next;
 265             if (!nofree) {
 266                 dictFreeEntryKey(ht, he);
 267                 dictFreeEntryVal(ht, he);
 268             }
 269             _dictFree(he);
 270             ht->used--;
 271             return DICT_OK;
 272         }
 273         prevHe = he;
 274         he = he->next;
 275     }
 276     return DICT_ERR; /* not found */
 277 }
 278
 279 int dictDelete(dict *ht, const void *key) {
 280     return dictGenericDelete(ht,key,0);
 281 }
 282
 283 int dictDeleteNoFree(dict *ht, const void *key) {
 284     return dictGenericDelete(ht,key,1);
 285 }
 286
 287 /* Destroy an entire hash table */
 288 int _dictClear(dict *ht)
 289 {
 290     unsigned long i;
 291
 292     /* Free all the elements */
 293     for (i = 0; i < ht->size && ht->used > 0; i++) {
 294         dictEntry *he, *nextHe;
 295
 296         if ((he = ht->table[i]) == NULL) continue;
 297         while(he) {
 298             nextHe = he->next;
 299             dictFreeEntryKey(ht, he);
 300             dictFreeEntryVal(ht, he);
 301             _dictFree(he);
 302             ht->used--;
 303             he = nextHe;
 304         }
 305     }
 306     /* Free the table and the allocated cache structure */
 307     _dictFree(ht->table);
 308     /* Re-initialize the table */
 309     _dictReset(ht);
 310     return DICT_OK; /* never fails */
 311 }
 312
 313 /* Clear & Release the hash table */
 314 void dictRelease(dict *ht)
 315 {
 316     _dictClear(ht);
 317     _dictFree(ht);
 318 }
 319
 320 dictEntry *dictFind(dict *ht, const void *key)
 321 {
 322     dictEntry *he;
 323     unsigned int h;
 324
 325     if (ht->size == 0) return NULL;
 326     h = dictHashKey(ht, key) & ht->sizemask;
 327     he = ht->table[h];
 328     while(he) {
 329         if (dictCompareHashKeys(ht, key, he->key))
 330             return he;
 331         he = he->next;
 332     }
 333     return NULL;
 334 }
 335
 336 dictIterator *dictGetIterator(dict *ht)
 337 {
 338     dictIterator *iter = _dictAlloc(sizeof(*iter));
 339
 340     iter->ht = ht;
 341     iter->index = -1;
 342     iter->entry = NULL;
 343     iter->nextEntry = NULL;
 344     return iter;
 345 }
 346
 347 dictEntry *dictNext(dictIterator *iter)
 348 {
 349     while (1) {
 350         if (iter->entry == NULL) {
 351             iter->index++;
 352             if (iter->index >=
 353                     (signed)iter->ht->size) break;
 354             iter->entry = iter->ht->table[iter->index];
 355         } else {
 356             iter->entry = iter->nextEntry;
 357         }
 358         if (iter->entry) {
 359             /* We need to save the 'next' here, the iterator user
 360              * may delete the entry we are returning. */
 361             iter->nextEntry = iter->entry->next;
 362             return iter->entry;
 363         }
 364     }
 365     return NULL;
 366 }
 367
 368 void dictReleaseIterator(dictIterator *iter)
 369 {
 370     _dictFree(iter);
 371 }
 372
 373 /* Return a random entry from the hash table. Useful to
 374  * implement randomized algorithms */
 375 dictEntry *dictGetRandomKey(dict *ht)
 376 {
 377     dictEntry *he;
 378     unsigned int h;
 379     int listlen, listele;
 380
 381     if (ht->used == 0) return NULL;
 382     do {
 383         h = random() & ht->sizemask;
 384         he = ht->table[h];
 385     } while(he == NULL);
 386
 387     /* Now we found a non empty bucket, but it is a linked
 388      * list and we need to get a random element from the list.
 389      * The only sane way to do so is to count the element and
 390      * select a random index. */
 391     listlen = 0;
 392     while(he) {
 393         he = he->next;
 394         listlen++;
 395     }
 396     listele = random() % listlen;
 397     he = ht->table[h];
 398     while(listele--) he = he->next;
 399     return he;
 400 }
 401
 402 /* ------------------------- private functions ------------------------------ */
 403
 404 /* Expand the hash table if needed */
 405 static int _dictExpandIfNeeded(dict *ht)
 406 {
 407     /* If the hash table is empty expand it to the intial size,
 408      * if the table is "full" dobule its size. */
 409     if (ht->size == 0)
 410         return dictExpand(ht, DICT_HT_INITIAL_SIZE);
 411     if (ht->used == ht->size)
 412         return dictExpand(ht, ht->size*2);
 413     return DICT_OK;
 414 }
 415
 416 /* Our hash table capability is a power of two */
 417 static unsigned long _dictNextPower(unsigned long size)
 418 {
 419     unsigned long i = DICT_HT_INITIAL_SIZE;
 420
 421     if (size >= LONG_MAX) return LONG_MAX;
 422     while(1) {
 423         if (i >= size)
 424             return i;
 425         i *= 2;
 426     }
 427 }
 428
 429 /* Returns the index of a free slot that can be populated with
 430  * an hash entry for the given 'key'.
 431  * If the key already exists, -1 is returned. */
 432 static int _dictKeyIndex(dict *ht, const void *key)
 433 {
 434     unsigned int h;
 435     dictEntry *he;
 436
 437     /* Expand the hashtable if needed */
 438     if (_dictExpandIfNeeded(ht) == DICT_ERR)
 439         return -1;
 440     /* Compute the key hash value */
 441     h = dictHashKey(ht, key) & ht->sizemask;
 442     /* Search if this slot does not already contain the given key */
 443     he = ht->table[h];
 444     while(he) {
 445         if (dictCompareHashKeys(ht, key, he->key))
 446             return -1;
 447         he = he->next;
 448     }
 449     return h;
 450 }
 451
 452 void dictEmpty(dict *ht) {
 453     _dictClear(ht);
 454 }
 455
 456 #define DICT_STATS_VECTLEN 50
 457 void dictPrintStats(dict *ht) {
 458     unsigned long i, slots = 0, chainlen, maxchainlen = 0;
 459     unsigned long totchainlen = 0;
 460     unsigned long clvector[DICT_STATS_VECTLEN];
 461
 462     if (ht->used == 0) {
 463         printf("No stats available for empty dictionaries\n");
 464         return;
 465     }
 466
 467     for (i = 0; i < DICT_STATS_VECTLEN; i++) clvector[i] = 0;
 468     for (i = 0; i < ht->size; i++) {
 469         dictEntry *he;
 470
 471         if (ht->table[i] == NULL) {
 472             clvector[0]++;
 473             continue;
 474         }
 475         slots++;
 476         /* For each hash entry on this slot... */
 477         chainlen = 0;
 478         he = ht->table[i];
 479         while(he) {
 480             chainlen++;
 481             he = he->next;
 482         }
 483         clvector[(chainlen < DICT_STATS_VECTLEN) ? chainlen : (DICT_STATS_VECTLEN-1)]++;
 484         if (chainlen > maxchainlen) maxchainlen = chainlen;
 485         totchainlen += chainlen;
 486     }
 487     printf("Hash table stats:\n");
 488     printf(" table size: %ld\n", ht->size);
 489     printf(" number of elements: %ld\n", ht->used);
 490     printf(" different slots: %ld\n", slots);
 491     printf(" max chain length: %ld\n", maxchainlen);
 492     printf(" avg chain length (counted): %.02f\n", (float)totchainlen/slots);
 493     printf(" avg chain length (computed): %.02f\n", (float)ht->used/slots);
 494     printf(" Chain length distribution:\n");
 495     for (i = 0; i < DICT_STATS_VECTLEN-1; i++) {
 496         if (clvector[i] == 0) continue;
 497         printf("   %s%ld: %ld (%.02f%%)\n",(i == DICT_STATS_VECTLEN-1)?">= ":"", i, clvector[i], ((float)clvector[i]/ht->size)*100);
 498     }
 499 }
 500
 501 /* ----------------------- StringCopy Hash Table Type ------------------------*/
 502
 503 static unsigned int _dictStringCopyHTHashFunction(const void *key)
 504 {
 505     return dictGenHashFunction(key, strlen(key));
 506 }
 507
 508 static void *_dictStringCopyHTKeyDup(void *privdata, const void *key)
 509 {
 510     int len = strlen(key);
 511     char *copy = _dictAlloc(len+1);
 512     DICT_NOTUSED(privdata);
 513
 514     memcpy(copy, key, len);
 515     copy[len] = '\0';
 516     return copy;
 517 }
 518
 519 static void *_dictStringKeyValCopyHTValDup(void *privdata, const void *val)
 520 {
 521     int len = strlen(val);
 522     char *copy = _dictAlloc(len+1);
 523     DICT_NOTUSED(privdata);
 524
 525     memcpy(copy, val, len);
 526     copy[len] = '\0';
 527     return copy;
 528 }
 529
 530 static int _dictStringCopyHTKeyCompare(void *privdata, const void *key1,
 531         const void *key2)
 532 {
 533     DICT_NOTUSED(privdata);
 534
 535     return strcmp(key1, key2) == 0;
 536 }
 537
 538 static void _dictStringCopyHTKeyDestructor(void *privdata, void *key)
 539 {
 540     DICT_NOTUSED(privdata);
 541
 542     _dictFree((void*)key); /* ATTENTION: const cast */
 543 }
 544
 545 static void _dictStringKeyValCopyHTValDestructor(void *privdata, void *val)
 546 {
 547     DICT_NOTUSED(privdata);
 548
 549     _dictFree((void*)val); /* ATTENTION: const cast */
 550 }
 551
 552 dictType dictTypeHeapStringCopyKey = {
 553     _dictStringCopyHTHashFunction,        /* hash function */
 554     _dictStringCopyHTKeyDup,              /* key dup */
 555     NULL,                               /* val dup */
 556     _dictStringCopyHTKeyCompare,          /* key compare */
 557     _dictStringCopyHTKeyDestructor,       /* key destructor */
 558     NULL                                /* val destructor */
 559 };
 560
 561 /* This is like StringCopy but does not auto-duplicate the key.
 562  * It's used for intepreter's shared strings. */
 563 dictType dictTypeHeapStrings = {
 564     _dictStringCopyHTHashFunction,        /* hash function */
 565     NULL,                               /* key dup */
 566     NULL,                               /* val dup */
 567     _dictStringCopyHTKeyCompare,          /* key compare */
 568     _dictStringCopyHTKeyDestructor,       /* key destructor */
 569     NULL                                /* val destructor */
 570 };
 571
 572 /* This is like StringCopy but also automatically handle dynamic
 573  * allocated C strings as values. */
 574 dictType dictTypeHeapStringCopyKeyValue = {
 575     _dictStringCopyHTHashFunction,        /* hash function */
 576     _dictStringCopyHTKeyDup,              /* key dup */
 577     _dictStringKeyValCopyHTValDup,        /* val dup */
 578     _dictStringCopyHTKeyCompare,          /* key compare */
 579     _dictStringCopyHTKeyDestructor,       /* key destructor */
 580     _dictStringKeyValCopyHTValDestructor, /* val destructor */
 581 };