ht-internal.h (29185B)
1 /* Copyright 2002 Christopher Clark */ 2 /* Copyright 2005-2012 Nick Mathewson */ 3 /* Copyright 2009-2012 Niels Provos and Nick Mathewson */ 4 /* See license at end. */ 5 6 /* Based on ideas by Christopher Clark and interfaces from Niels Provos. */ 7 8 #ifndef HT_INTERNAL_H_INCLUDED_ 9 #define HT_INTERNAL_H_INCLUDED_ 10 11 #define HT_HEAD(name, type) \ 12 struct name { \ 13 /* The hash table itself. */ \ 14 struct type **hth_table; \ 15 /* How long is the hash table? */ \ 16 unsigned hth_table_length; \ 17 /* How many elements does the table contain? */ \ 18 unsigned hth_n_entries; \ 19 /* How many elements will we allow in the table before resizing it? */ \ 20 unsigned hth_load_limit; \ 21 /* Position of hth_table_length in the primes table. */ \ 22 int hth_prime_idx; \ 23 } 24 25 #define HT_INITIALIZER() \ 26 { NULL, 0, 0, 0, -1 } 27 28 #ifdef HT_NO_CACHE_HASH_VALUES 29 #define HT_ENTRY(type) \ 30 struct { \ 31 struct type *hte_next; \ 32 } 33 #else 34 #define HT_ENTRY(type) \ 35 struct { \ 36 struct type *hte_next; \ 37 unsigned hte_hash; \ 38 } 39 #endif 40 41 #define HT_EMPTY(head) \ 42 ((head)->hth_n_entries == 0) 43 44 /* How many elements in 'head'? */ 45 #define HT_SIZE(head) \ 46 ((head)->hth_n_entries) 47 48 /* Return memory usage for a hashtable (not counting the entries themselves) */ 49 #define HT_MEM_USAGE(head) \ 50 (sizeof(*head) + (head)->hth_table_length * sizeof(void*)) 51 52 #define HT_FIND(name, head, elm) name##_HT_FIND((head), (elm)) 53 #define HT_INSERT(name, head, elm) name##_HT_INSERT((head), (elm)) 54 #define HT_REPLACE(name, head, elm) name##_HT_REPLACE((head), (elm)) 55 #define HT_REMOVE(name, head, elm) name##_HT_REMOVE((head), (elm)) 56 #define HT_START(name, head) name##_HT_START(head) 57 #define HT_NEXT(name, head, elm) name##_HT_NEXT((head), (elm)) 58 #define HT_NEXT_RMV(name, head, elm) name##_HT_NEXT_RMV((head), (elm)) 59 #define HT_CLEAR(name, head) name##_HT_CLEAR(head) 60 #define HT_INIT(name, head) name##_HT_INIT(head) 61 /* Helper: */ 62 static inline unsigned 63 ht_improve_hash_(unsigned h) 64 { 65 /* Aim to protect against poor hash functions by adding logic here 66 * - logic taken from java 1.4 hashtable source */ 67 h += ~(h << 9); 68 h ^= ((h >> 14) | (h << 18)); /* >>> */ 69 h += (h << 4); 70 h ^= ((h >> 10) | (h << 22)); /* >>> */ 71 return h; 72 } 73 74 #if 0 75 /** Basic string hash function, from Java standard String.hashCode(). */ 76 static inline unsigned 77 ht_string_hash_(const char *s) 78 { 79 unsigned h = 0; 80 int m = 1; 81 while (*s) { 82 h += ((signed char)*s++)*m; 83 m = (m<<5)-1; /* m *= 31 */ 84 } 85 return h; 86 } 87 #endif 88 89 /** Basic string hash function, from Python's str.__hash__() */ 90 static inline unsigned 91 ht_string_hash_(const char *s) 92 { 93 unsigned h; 94 const unsigned char *cp = (const unsigned char *)s; 95 h = *cp << 7; 96 while (*cp) { 97 h = (1000003*h) ^ *cp++; 98 } 99 /* This conversion truncates the length of the string, but that's ok. */ 100 h ^= (unsigned)(cp-(const unsigned char*)s); 101 return h; 102 } 103 104 #ifndef HT_NO_CACHE_HASH_VALUES 105 #define HT_SET_HASH_(elm, field, hashfn) \ 106 do { (elm)->field.hte_hash = hashfn(elm); } while (0) 107 #define HT_SET_HASHVAL_(elm, field, val) \ 108 do { (elm)->field.hte_hash = (val); } while (0) 109 #define HT_ELT_HASH_(elm, field, hashfn) \ 110 ((elm)->field.hte_hash) 111 #else 112 #define HT_SET_HASH_(elm, field, hashfn) \ 113 ((void)0) 114 #define HT_ELT_HASH_(elm, field, hashfn) \ 115 (hashfn(elm)) 116 #define HT_SET_HASHVAL_(elm, field, val) \ 117 ((void)0) 118 #endif 119 120 /* Helper: alias for the bucket containing 'elm'. */ 121 #define HT_BUCKET_(head, field, elm, hashfn) \ 122 ((head)->hth_table[HT_ELT_HASH_(elm,field,hashfn) % head->hth_table_length]) 123 124 #define HT_FOREACH(x, name, head) \ 125 for ((x) = HT_START(name, head); \ 126 (x) != NULL; \ 127 (x) = HT_NEXT(name, head, x)) 128 129 #define HT_PROTOTYPE(name, type, field, hashfn, eqfn) \ 130 int name##_HT_GROW(struct name *ht, unsigned min_capacity); \ 131 void name##_HT_CLEAR(struct name *ht); \ 132 int name##_HT_REP_IS_BAD_(const struct name *ht); \ 133 static inline void \ 134 name##_HT_INIT(struct name *head) { \ 135 head->hth_table_length = 0; \ 136 head->hth_table = NULL; \ 137 head->hth_n_entries = 0; \ 138 head->hth_load_limit = 0; \ 139 head->hth_prime_idx = -1; \ 140 } \ 141 /* Helper: returns a pointer to the right location in the table \ 142 * 'head' to find or insert the element 'elm'. */ \ 143 static inline struct type ** \ 144 name##_HT_FIND_P_(struct name *head, struct type *elm) \ 145 { \ 146 struct type **p; \ 147 if (!head->hth_table) \ 148 return NULL; \ 149 p = &HT_BUCKET_(head, field, elm, hashfn); \ 150 while (*p) { \ 151 if (eqfn(*p, elm)) \ 152 return p; \ 153 p = &(*p)->field.hte_next; \ 154 } \ 155 return p; \ 156 } \ 157 /* Return a pointer to the element in the table 'head' matching 'elm', \ 158 * or NULL if no such element exists */ \ 159 static inline struct type * \ 160 name##_HT_FIND(const struct name *head, struct type *elm) \ 161 { \ 162 struct type **p; \ 163 struct name *h = (struct name *) head; \ 164 HT_SET_HASH_(elm, field, hashfn); \ 165 p = name##_HT_FIND_P_(h, elm); \ 166 return p ? *p : NULL; \ 167 } \ 168 /* Insert the element 'elm' into the table 'head'. Do not call this \ 169 * function if the table might already contain a matching element. */ \ 170 static inline void \ 171 name##_HT_INSERT(struct name *head, struct type *elm) \ 172 { \ 173 struct type **p; \ 174 if (!head->hth_table || head->hth_n_entries >= head->hth_load_limit) \ 175 name##_HT_GROW(head, head->hth_n_entries+1); \ 176 ++head->hth_n_entries; \ 177 HT_SET_HASH_(elm, field, hashfn); \ 178 p = &HT_BUCKET_(head, field, elm, hashfn); \ 179 elm->field.hte_next = *p; \ 180 *p = elm; \ 181 } \ 182 /* Insert the element 'elm' into the table 'head'. If there already \ 183 * a matching element in the table, replace that element and return \ 184 * it. */ \ 185 static inline struct type * \ 186 name##_HT_REPLACE(struct name *head, struct type *elm) \ 187 { \ 188 struct type **p, *r; \ 189 if (!head->hth_table || head->hth_n_entries >= head->hth_load_limit) \ 190 name##_HT_GROW(head, head->hth_n_entries+1); \ 191 HT_SET_HASH_(elm, field, hashfn); \ 192 p = name##_HT_FIND_P_(head, elm); \ 193 r = *p; \ 194 *p = elm; \ 195 if (r && (r!=elm)) { \ 196 elm->field.hte_next = r->field.hte_next; \ 197 r->field.hte_next = NULL; \ 198 return r; \ 199 } else { \ 200 ++head->hth_n_entries; \ 201 return NULL; \ 202 } \ 203 } \ 204 /* Remove any element matching 'elm' from the table 'head'. If such \ 205 * an element is found, return it; otherwise return NULL. */ \ 206 static inline struct type * \ 207 name##_HT_REMOVE(struct name *head, struct type *elm) \ 208 { \ 209 struct type **p, *r; \ 210 HT_SET_HASH_(elm, field, hashfn); \ 211 p = name##_HT_FIND_P_(head,elm); \ 212 if (!p || !*p) \ 213 return NULL; \ 214 r = *p; \ 215 *p = r->field.hte_next; \ 216 r->field.hte_next = NULL; \ 217 --head->hth_n_entries; \ 218 return r; \ 219 } \ 220 /* Invoke the function 'fn' on every element of the table 'head', \ 221 * using 'data' as its second argument. If the function returns \ 222 * nonzero, remove the most recently examined element before invoking \ 223 * the function again. */ \ 224 static inline void \ 225 name##_HT_FOREACH_FN(struct name *head, \ 226 int (*fn)(struct type *, void *), \ 227 void *data) \ 228 { \ 229 unsigned idx; \ 230 struct type **p, **nextp, *next; \ 231 if (!head->hth_table) \ 232 return; \ 233 for (idx=0; idx < head->hth_table_length; ++idx) { \ 234 p = &head->hth_table[idx]; \ 235 while (*p) { \ 236 nextp = &(*p)->field.hte_next; \ 237 next = *nextp; \ 238 if (fn(*p, data)) { \ 239 --head->hth_n_entries; \ 240 *p = next; \ 241 } else { \ 242 p = nextp; \ 243 } \ 244 } \ 245 } \ 246 } \ 247 /* Return a pointer to the first element in the table 'head', under \ 248 * an arbitrary order. This order is stable under remove operations, \ 249 * but not under others. If the table is empty, return NULL. */ \ 250 static inline struct type ** \ 251 name##_HT_START(struct name *head) \ 252 { \ 253 unsigned b = 0; \ 254 while (b < head->hth_table_length) { \ 255 if (head->hth_table[b]) \ 256 return &head->hth_table[b]; \ 257 ++b; \ 258 } \ 259 return NULL; \ 260 } \ 261 /* Return the next element in 'head' after 'elm', under the arbitrary \ 262 * order used by HT_START. If there are no more elements, return \ 263 * NULL. If 'elm' is to be removed from the table, you must call \ 264 * this function for the next value before you remove it. \ 265 */ \ 266 static inline struct type ** \ 267 name##_HT_NEXT(struct name *head, struct type **elm) \ 268 { \ 269 if ((*elm)->field.hte_next) { \ 270 return &(*elm)->field.hte_next; \ 271 } else { \ 272 unsigned b = (HT_ELT_HASH_(*elm, field, hashfn) % head->hth_table_length)+1; \ 273 while (b < head->hth_table_length) { \ 274 if (head->hth_table[b]) \ 275 return &head->hth_table[b]; \ 276 ++b; \ 277 } \ 278 return NULL; \ 279 } \ 280 } \ 281 static inline struct type ** \ 282 name##_HT_NEXT_RMV(struct name *head, struct type **elm) \ 283 { \ 284 unsigned h = HT_ELT_HASH_(*elm, field, hashfn); \ 285 *elm = (*elm)->field.hte_next; \ 286 --head->hth_n_entries; \ 287 if (*elm) { \ 288 return elm; \ 289 } else { \ 290 unsigned b = (h % head->hth_table_length)+1; \ 291 while (b < head->hth_table_length) { \ 292 if (head->hth_table[b]) \ 293 return &head->hth_table[b]; \ 294 ++b; \ 295 } \ 296 return NULL; \ 297 } \ 298 } 299 300 #define HT_GENERATE(name, type, field, hashfn, eqfn, load, mallocfn, \ 301 reallocfn, freefn) \ 302 static unsigned name##_PRIMES[] = { \ 303 53, 97, 193, 389, \ 304 769, 1543, 3079, 6151, \ 305 12289, 24593, 49157, 98317, \ 306 196613, 393241, 786433, 1572869, \ 307 3145739, 6291469, 12582917, 25165843, \ 308 50331653, 100663319, 201326611, 402653189, \ 309 805306457, 1610612741 \ 310 }; \ 311 static unsigned name##_N_PRIMES = \ 312 (unsigned)(sizeof(name##_PRIMES)/sizeof(name##_PRIMES[0])); \ 313 /* Expand the internal table of 'head' until it is large enough to \ 314 * hold 'size' elements. Return 0 on success, -1 on allocation \ 315 * failure. */ \ 316 int \ 317 name##_HT_GROW(struct name *head, unsigned size) \ 318 { \ 319 unsigned new_len, new_load_limit; \ 320 int prime_idx; \ 321 struct type **new_table; \ 322 if (head->hth_prime_idx == (int)name##_N_PRIMES - 1) \ 323 return 0; \ 324 if (head->hth_load_limit > size) \ 325 return 0; \ 326 prime_idx = head->hth_prime_idx; \ 327 do { \ 328 new_len = name##_PRIMES[++prime_idx]; \ 329 new_load_limit = (unsigned)(load*new_len); \ 330 } while (new_load_limit <= size && \ 331 prime_idx < (int)name##_N_PRIMES); \ 332 if ((new_table = mallocfn(new_len*sizeof(struct type*)))) { \ 333 unsigned b; \ 334 memset(new_table, 0, new_len*sizeof(struct type*)); \ 335 for (b = 0; b < head->hth_table_length; ++b) { \ 336 struct type *elm, *next; \ 337 unsigned b2; \ 338 elm = head->hth_table[b]; \ 339 while (elm) { \ 340 next = elm->field.hte_next; \ 341 b2 = HT_ELT_HASH_(elm, field, hashfn) % new_len; \ 342 elm->field.hte_next = new_table[b2]; \ 343 new_table[b2] = elm; \ 344 elm = next; \ 345 } \ 346 } \ 347 if (head->hth_table) \ 348 freefn(head->hth_table); \ 349 head->hth_table = new_table; \ 350 } else { \ 351 unsigned b, b2; \ 352 new_table = reallocfn(head->hth_table, new_len*sizeof(struct type*)); \ 353 if (!new_table) return -1; \ 354 memset(new_table + head->hth_table_length, 0, \ 355 (new_len - head->hth_table_length)*sizeof(struct type*)); \ 356 for (b=0; b < head->hth_table_length; ++b) { \ 357 struct type *e, **pE; \ 358 for (pE = &new_table[b], e = *pE; e != NULL; e = *pE) { \ 359 b2 = HT_ELT_HASH_(e, field, hashfn) % new_len; \ 360 if (b2 == b) { \ 361 pE = &e->field.hte_next; \ 362 } else { \ 363 *pE = e->field.hte_next; \ 364 e->field.hte_next = new_table[b2]; \ 365 new_table[b2] = e; \ 366 } \ 367 } \ 368 } \ 369 head->hth_table = new_table; \ 370 } \ 371 head->hth_table_length = new_len; \ 372 head->hth_prime_idx = prime_idx; \ 373 head->hth_load_limit = new_load_limit; \ 374 return 0; \ 375 } \ 376 /* Free all storage held by 'head'. Does not free 'head' itself, or \ 377 * individual elements. */ \ 378 void \ 379 name##_HT_CLEAR(struct name *head) \ 380 { \ 381 if (head->hth_table) \ 382 freefn(head->hth_table); \ 383 name##_HT_INIT(head); \ 384 } \ 385 /* Debugging helper: return false iff the representation of 'head' is \ 386 * internally consistent. */ \ 387 int \ 388 name##_HT_REP_IS_BAD_(const struct name *head) \ 389 { \ 390 unsigned n, i; \ 391 struct type *elm; \ 392 if (!head->hth_table_length) { \ 393 if (!head->hth_table && !head->hth_n_entries && \ 394 !head->hth_load_limit && head->hth_prime_idx == -1) \ 395 return 0; \ 396 else \ 397 return 1; \ 398 } \ 399 if (!head->hth_table || head->hth_prime_idx < 0 || \ 400 !head->hth_load_limit) \ 401 return 2; \ 402 if (head->hth_n_entries > head->hth_load_limit) \ 403 return 3; \ 404 if (head->hth_table_length != name##_PRIMES[head->hth_prime_idx]) \ 405 return 4; \ 406 if (head->hth_load_limit != (unsigned)(load*head->hth_table_length)) \ 407 return 5; \ 408 for (n = i = 0; i < head->hth_table_length; ++i) { \ 409 for (elm = head->hth_table[i]; elm; elm = elm->field.hte_next) { \ 410 if (HT_ELT_HASH_(elm, field, hashfn) != hashfn(elm)) \ 411 return 1000 + i; \ 412 if ((HT_ELT_HASH_(elm, field, hashfn) % head->hth_table_length) != i) \ 413 return 10000 + i; \ 414 ++n; \ 415 } \ 416 } \ 417 if (n != head->hth_n_entries) \ 418 return 6; \ 419 return 0; \ 420 } 421 422 /** Implements an over-optimized "find and insert if absent" block; 423 * not meant for direct usage by typical code, or usage outside the critical 424 * path.*/ 425 #define HT_FIND_OR_INSERT_(name, field, hashfn, head, eltype, elm, var, y, n) \ 426 { \ 427 struct name *var##_head_ = head; \ 428 struct eltype **var; \ 429 if (!var##_head_->hth_table || \ 430 var##_head_->hth_n_entries >= var##_head_->hth_load_limit) \ 431 name##_HT_GROW(var##_head_, var##_head_->hth_n_entries+1); \ 432 HT_SET_HASH_((elm), field, hashfn); \ 433 var = name##_HT_FIND_P_(var##_head_, (elm)); \ 434 if (*var) { \ 435 y; \ 436 } else { \ 437 n; \ 438 } \ 439 } 440 #define HT_FOI_INSERT_(field, head, elm, newent, var) \ 441 { \ 442 HT_SET_HASHVAL_(newent, field, (elm)->field.hte_hash); \ 443 newent->field.hte_next = NULL; \ 444 *var = newent; \ 445 ++((head)->hth_n_entries); \ 446 } 447 448 /* 449 * Copyright 2005, Nick Mathewson. Implementation logic is adapted from code 450 * by Christopher Clark, retrofit to allow drop-in memory management, and to 451 * use the same interface as Niels Provos's tree.h. This is probably still 452 * a derived work, so the original license below still applies. 453 * 454 * Copyright (c) 2002, Christopher Clark 455 * All rights reserved. 456 * 457 * Redistribution and use in source and binary forms, with or without 458 * modification, are permitted provided that the following conditions 459 * are met: 460 * 461 * * Redistributions of source code must retain the above copyright 462 * notice, this list of conditions and the following disclaimer. 463 * 464 * * Redistributions in binary form must reproduce the above copyright 465 * notice, this list of conditions and the following disclaimer in the 466 * documentation and/or other materials provided with the distribution. 467 * 468 * * Neither the name of the original author; nor the names of any contributors 469 * may be used to endorse or promote products derived from this software 470 * without specific prior written permission. 471 * 472 * 473 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 474 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 475 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 476 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER 477 * OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 478 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 479 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 480 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 481 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 482 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 483 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 484 */ 485 486 #endif