Commit | Line | Data |
---|---|---|
e2eaf477 | 1 | /* An expandable hash tables datatype. |
5f9624e3 | 2 | Copyright (C) 1999, 2000, 2001, 2002, 2003 Free Software Foundation, Inc. |
e2eaf477 ILT |
3 | Contributed by Vladimir Makarov (vmakarov@cygnus.com). |
4 | ||
5 | This file is part of the libiberty library. | |
6 | Libiberty is free software; you can redistribute it and/or | |
7 | modify it under the terms of the GNU Library General Public | |
8 | License as published by the Free Software Foundation; either | |
9 | version 2 of the License, or (at your option) any later version. | |
10 | ||
11 | Libiberty is distributed in the hope that it will be useful, | |
12 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | Library General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU Library General Public | |
17 | License along with libiberty; see the file COPYING.LIB. If | |
18 | not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
19 | Boston, MA 02111-1307, USA. */ | |
20 | ||
21 | /* This package implements basic hash table functionality. It is possible | |
22 | to search for an entry, create an entry and destroy an entry. | |
23 | ||
24 | Elements in the table are generic pointers. | |
25 | ||
26 | The size of the table is not fixed; if the occupancy of the table | |
27 | grows too high the hash table will be expanded. | |
28 | ||
29 | The abstract data implementation is based on generalized Algorithm D | |
30 | from Knuth's book "The art of computer programming". Hash table is | |
31 | expanded by creation of new hash table and transferring elements from | |
32 | the old table to the new table. */ | |
33 | ||
34 | #ifdef HAVE_CONFIG_H | |
35 | #include "config.h" | |
36 | #endif | |
37 | ||
38 | #include <sys/types.h> | |
39 | ||
40 | #ifdef HAVE_STDLIB_H | |
41 | #include <stdlib.h> | |
42 | #endif | |
43 | ||
5c82d20a ZW |
44 | #ifdef HAVE_STRING_H |
45 | #include <string.h> | |
46 | #endif | |
47 | ||
e2eaf477 ILT |
48 | #include <stdio.h> |
49 | ||
50 | #include "libiberty.h" | |
51 | #include "hashtab.h" | |
52 | ||
e2eaf477 ILT |
53 | /* This macro defines reserved value for empty table entry. */ |
54 | ||
e0f3df8f | 55 | #define EMPTY_ENTRY ((PTR) 0) |
e2eaf477 ILT |
56 | |
57 | /* This macro defines reserved value for table entry which contained | |
58 | a deleted element. */ | |
59 | ||
e0f3df8f | 60 | #define DELETED_ENTRY ((PTR) 1) |
e2eaf477 | 61 | |
eb383413 L |
62 | static unsigned long higher_prime_number PARAMS ((unsigned long)); |
63 | static hashval_t hash_pointer PARAMS ((const void *)); | |
64 | static int eq_pointer PARAMS ((const void *, const void *)); | |
99a4c1bd | 65 | static int htab_expand PARAMS ((htab_t)); |
e0f3df8f | 66 | static PTR *find_empty_slot_for_expand PARAMS ((htab_t, hashval_t)); |
eb383413 L |
67 | |
68 | /* At some point, we could make these be NULL, and modify the | |
69 | hash-table routines to handle NULL specially; that would avoid | |
70 | function-call overhead for the common case of hashing pointers. */ | |
71 | htab_hash htab_hash_pointer = hash_pointer; | |
72 | htab_eq htab_eq_pointer = eq_pointer; | |
73 | ||
5ca0f83d DD |
74 | /* The following function returns a nearest prime number which is |
75 | greater than N, and near a power of two. */ | |
e2eaf477 ILT |
76 | |
77 | static unsigned long | |
b4fe2683 JM |
78 | higher_prime_number (n) |
79 | unsigned long n; | |
e2eaf477 | 80 | { |
5ca0f83d DD |
81 | /* These are primes that are near, but slightly smaller than, a |
82 | power of two. */ | |
e6450fe5 | 83 | static const unsigned long primes[] = { |
b1e51b3c DD |
84 | (unsigned long) 7, |
85 | (unsigned long) 13, | |
86 | (unsigned long) 31, | |
87 | (unsigned long) 61, | |
88 | (unsigned long) 127, | |
89 | (unsigned long) 251, | |
90 | (unsigned long) 509, | |
91 | (unsigned long) 1021, | |
92 | (unsigned long) 2039, | |
93 | (unsigned long) 4093, | |
94 | (unsigned long) 8191, | |
95 | (unsigned long) 16381, | |
96 | (unsigned long) 32749, | |
97 | (unsigned long) 65521, | |
98 | (unsigned long) 131071, | |
99 | (unsigned long) 262139, | |
100 | (unsigned long) 524287, | |
101 | (unsigned long) 1048573, | |
102 | (unsigned long) 2097143, | |
103 | (unsigned long) 4194301, | |
104 | (unsigned long) 8388593, | |
105 | (unsigned long) 16777213, | |
106 | (unsigned long) 33554393, | |
107 | (unsigned long) 67108859, | |
108 | (unsigned long) 134217689, | |
109 | (unsigned long) 268435399, | |
110 | (unsigned long) 536870909, | |
111 | (unsigned long) 1073741789, | |
112 | (unsigned long) 2147483647, | |
113 | /* 4294967291L */ | |
06b0287c | 114 | ((unsigned long) 2147483647) + ((unsigned long) 2147483644), |
5ca0f83d DD |
115 | }; |
116 | ||
e6450fe5 DD |
117 | const unsigned long *low = &primes[0]; |
118 | const unsigned long *high = &primes[sizeof(primes) / sizeof(primes[0])]; | |
5ca0f83d DD |
119 | |
120 | while (low != high) | |
121 | { | |
e6450fe5 | 122 | const unsigned long *mid = low + (high - low) / 2; |
5ca0f83d DD |
123 | if (n > *mid) |
124 | low = mid + 1; | |
125 | else | |
126 | high = mid; | |
127 | } | |
128 | ||
129 | /* If we've run out of primes, abort. */ | |
130 | if (n > *low) | |
131 | { | |
132 | fprintf (stderr, "Cannot find prime bigger than %lu\n", n); | |
133 | abort (); | |
134 | } | |
135 | ||
136 | return *low; | |
e2eaf477 ILT |
137 | } |
138 | ||
eb383413 L |
139 | /* Returns a hash code for P. */ |
140 | ||
141 | static hashval_t | |
142 | hash_pointer (p) | |
e0f3df8f | 143 | const PTR p; |
eb383413 L |
144 | { |
145 | return (hashval_t) ((long)p >> 3); | |
146 | } | |
147 | ||
148 | /* Returns non-zero if P1 and P2 are equal. */ | |
149 | ||
150 | static int | |
151 | eq_pointer (p1, p2) | |
e0f3df8f HPN |
152 | const PTR p1; |
153 | const PTR p2; | |
eb383413 L |
154 | { |
155 | return p1 == p2; | |
156 | } | |
157 | ||
e2eaf477 ILT |
158 | /* This function creates table with length slightly longer than given |
159 | source length. Created hash table is initiated as empty (all the | |
160 | hash table entries are EMPTY_ENTRY). The function returns the | |
18893690 | 161 | created hash table, or NULL if memory allocation fails. */ |
e2eaf477 | 162 | |
b4fe2683 | 163 | htab_t |
18893690 | 164 | htab_create_alloc (size, hash_f, eq_f, del_f, alloc_f, free_f) |
e2eaf477 | 165 | size_t size; |
b4fe2683 JM |
166 | htab_hash hash_f; |
167 | htab_eq eq_f; | |
168 | htab_del del_f; | |
18893690 DD |
169 | htab_alloc alloc_f; |
170 | htab_free free_f; | |
e2eaf477 | 171 | { |
b4fe2683 | 172 | htab_t result; |
e2eaf477 ILT |
173 | |
174 | size = higher_prime_number (size); | |
18893690 DD |
175 | result = (htab_t) (*alloc_f) (1, sizeof (struct htab)); |
176 | if (result == NULL) | |
177 | return NULL; | |
178 | result->entries = (PTR *) (*alloc_f) (size, sizeof (PTR)); | |
179 | if (result->entries == NULL) | |
180 | { | |
181 | if (free_f != NULL) | |
182 | (*free_f) (result); | |
183 | return NULL; | |
184 | } | |
e2eaf477 | 185 | result->size = size; |
b4fe2683 JM |
186 | result->hash_f = hash_f; |
187 | result->eq_f = eq_f; | |
188 | result->del_f = del_f; | |
18893690 DD |
189 | result->alloc_f = alloc_f; |
190 | result->free_f = free_f; | |
99a4c1bd HPN |
191 | return result; |
192 | } | |
193 | ||
5f9624e3 DJ |
194 | /* As above, but use the variants of alloc_f and free_f which accept |
195 | an extra argument. */ | |
196 | ||
197 | htab_t | |
198 | htab_create_alloc_ex (size, hash_f, eq_f, del_f, alloc_arg, alloc_f, | |
199 | free_f) | |
200 | size_t size; | |
201 | htab_hash hash_f; | |
202 | htab_eq eq_f; | |
203 | htab_del del_f; | |
204 | PTR alloc_arg; | |
205 | htab_alloc_with_arg alloc_f; | |
206 | htab_free_with_arg free_f; | |
207 | { | |
208 | htab_t result; | |
209 | ||
210 | size = higher_prime_number (size); | |
211 | result = (htab_t) (*alloc_f) (alloc_arg, 1, sizeof (struct htab)); | |
212 | if (result == NULL) | |
213 | return NULL; | |
214 | result->entries = (PTR *) (*alloc_f) (alloc_arg, size, sizeof (PTR)); | |
215 | if (result->entries == NULL) | |
216 | { | |
217 | if (free_f != NULL) | |
218 | (*free_f) (alloc_arg, result); | |
219 | return NULL; | |
220 | } | |
221 | result->size = size; | |
222 | result->hash_f = hash_f; | |
223 | result->eq_f = eq_f; | |
224 | result->del_f = del_f; | |
225 | result->alloc_arg = alloc_arg; | |
226 | result->alloc_with_arg_f = alloc_f; | |
227 | result->free_with_arg_f = free_f; | |
228 | return result; | |
229 | } | |
230 | ||
231 | /* Update the function pointers and allocation parameter in the htab_t. */ | |
232 | ||
233 | void | |
234 | htab_set_functions_ex (htab, hash_f, eq_f, del_f, alloc_arg, alloc_f, free_f) | |
235 | htab_t htab; | |
236 | htab_hash hash_f; | |
237 | htab_eq eq_f; | |
238 | htab_del del_f; | |
239 | PTR alloc_arg; | |
240 | htab_alloc_with_arg alloc_f; | |
241 | htab_free_with_arg free_f; | |
242 | { | |
243 | htab->hash_f = hash_f; | |
244 | htab->eq_f = eq_f; | |
245 | htab->del_f = del_f; | |
246 | htab->alloc_arg = alloc_arg; | |
247 | htab->alloc_with_arg_f = alloc_f; | |
248 | htab->free_with_arg_f = free_f; | |
249 | } | |
250 | ||
18893690 | 251 | /* These functions exist solely for backward compatibility. */ |
99a4c1bd | 252 | |
18893690 | 253 | #undef htab_create |
99a4c1bd | 254 | htab_t |
18893690 | 255 | htab_create (size, hash_f, eq_f, del_f) |
99a4c1bd HPN |
256 | size_t size; |
257 | htab_hash hash_f; | |
258 | htab_eq eq_f; | |
259 | htab_del del_f; | |
260 | { | |
18893690 DD |
261 | return htab_create_alloc (size, hash_f, eq_f, del_f, xcalloc, free); |
262 | } | |
99a4c1bd | 263 | |
18893690 DD |
264 | htab_t |
265 | htab_try_create (size, hash_f, eq_f, del_f) | |
266 | size_t size; | |
267 | htab_hash hash_f; | |
268 | htab_eq eq_f; | |
269 | htab_del del_f; | |
270 | { | |
271 | return htab_create_alloc (size, hash_f, eq_f, del_f, calloc, free); | |
e2eaf477 ILT |
272 | } |
273 | ||
274 | /* This function frees all memory allocated for given hash table. | |
275 | Naturally the hash table must already exist. */ | |
276 | ||
277 | void | |
b4fe2683 JM |
278 | htab_delete (htab) |
279 | htab_t htab; | |
e2eaf477 | 280 | { |
b4fe2683 | 281 | int i; |
eb383413 | 282 | |
b4fe2683 JM |
283 | if (htab->del_f) |
284 | for (i = htab->size - 1; i >= 0; i--) | |
eb383413 L |
285 | if (htab->entries[i] != EMPTY_ENTRY |
286 | && htab->entries[i] != DELETED_ENTRY) | |
287 | (*htab->del_f) (htab->entries[i]); | |
b4fe2683 | 288 | |
18893690 DD |
289 | if (htab->free_f != NULL) |
290 | { | |
291 | (*htab->free_f) (htab->entries); | |
292 | (*htab->free_f) (htab); | |
293 | } | |
5f9624e3 DJ |
294 | else if (htab->free_with_arg_f != NULL) |
295 | { | |
296 | (*htab->free_with_arg_f) (htab->alloc_arg, htab->entries); | |
297 | (*htab->free_with_arg_f) (htab->alloc_arg, htab); | |
298 | } | |
e2eaf477 ILT |
299 | } |
300 | ||
301 | /* This function clears all entries in the given hash table. */ | |
302 | ||
303 | void | |
b4fe2683 JM |
304 | htab_empty (htab) |
305 | htab_t htab; | |
306 | { | |
307 | int i; | |
eb383413 | 308 | |
b4fe2683 JM |
309 | if (htab->del_f) |
310 | for (i = htab->size - 1; i >= 0; i--) | |
eb383413 L |
311 | if (htab->entries[i] != EMPTY_ENTRY |
312 | && htab->entries[i] != DELETED_ENTRY) | |
313 | (*htab->del_f) (htab->entries[i]); | |
b4fe2683 | 314 | |
e0f3df8f | 315 | memset (htab->entries, 0, htab->size * sizeof (PTR)); |
b4fe2683 JM |
316 | } |
317 | ||
318 | /* Similar to htab_find_slot, but without several unwanted side effects: | |
319 | - Does not call htab->eq_f when it finds an existing entry. | |
320 | - Does not change the count of elements/searches/collisions in the | |
321 | hash table. | |
322 | This function also assumes there are no deleted entries in the table. | |
323 | HASH is the hash value for the element to be inserted. */ | |
eb383413 | 324 | |
e0f3df8f | 325 | static PTR * |
b4fe2683 JM |
326 | find_empty_slot_for_expand (htab, hash) |
327 | htab_t htab; | |
eb383413 | 328 | hashval_t hash; |
e2eaf477 | 329 | { |
b4fe2683 | 330 | size_t size = htab->size; |
b4fe2683 | 331 | unsigned int index = hash % size; |
b1c933fc RH |
332 | PTR *slot = htab->entries + index; |
333 | hashval_t hash2; | |
334 | ||
335 | if (*slot == EMPTY_ENTRY) | |
336 | return slot; | |
337 | else if (*slot == DELETED_ENTRY) | |
338 | abort (); | |
b4fe2683 | 339 | |
b1c933fc | 340 | hash2 = 1 + hash % (size - 2); |
b4fe2683 JM |
341 | for (;;) |
342 | { | |
b1c933fc RH |
343 | index += hash2; |
344 | if (index >= size) | |
345 | index -= size; | |
eb383413 | 346 | |
b1c933fc | 347 | slot = htab->entries + index; |
b4fe2683 JM |
348 | if (*slot == EMPTY_ENTRY) |
349 | return slot; | |
eb383413 | 350 | else if (*slot == DELETED_ENTRY) |
b4fe2683 | 351 | abort (); |
b4fe2683 | 352 | } |
e2eaf477 ILT |
353 | } |
354 | ||
355 | /* The following function changes size of memory allocated for the | |
356 | entries and repeatedly inserts the table elements. The occupancy | |
357 | of the table after the call will be about 50%. Naturally the hash | |
358 | table must already exist. Remember also that the place of the | |
99a4c1bd HPN |
359 | table entries is changed. If memory allocation failures are allowed, |
360 | this function will return zero, indicating that the table could not be | |
361 | expanded. If all goes well, it will return a non-zero value. */ | |
e2eaf477 | 362 | |
99a4c1bd | 363 | static int |
b4fe2683 JM |
364 | htab_expand (htab) |
365 | htab_t htab; | |
e2eaf477 | 366 | { |
e0f3df8f HPN |
367 | PTR *oentries; |
368 | PTR *olimit; | |
369 | PTR *p; | |
18893690 | 370 | PTR *nentries; |
eed2b28c | 371 | size_t nsize; |
b4fe2683 JM |
372 | |
373 | oentries = htab->entries; | |
374 | olimit = oentries + htab->size; | |
375 | ||
c4d8feb2 DD |
376 | /* Resize only when table after removal of unused elements is either |
377 | too full or too empty. */ | |
378 | if ((htab->n_elements - htab->n_deleted) * 2 > htab->size | |
2336e177 DD |
379 | || ((htab->n_elements - htab->n_deleted) * 8 < htab->size |
380 | && htab->size > 32)) | |
c4d8feb2 DD |
381 | nsize = higher_prime_number ((htab->n_elements - htab->n_deleted) * 2); |
382 | else | |
383 | nsize = htab->size; | |
99a4c1bd | 384 | |
5f9624e3 DJ |
385 | if (htab->alloc_with_arg_f != NULL) |
386 | nentries = (PTR *) (*htab->alloc_with_arg_f) (htab->alloc_arg, nsize, | |
387 | sizeof (PTR *)); | |
388 | else | |
389 | nentries = (PTR *) (*htab->alloc_f) (nsize, sizeof (PTR *)); | |
18893690 DD |
390 | if (nentries == NULL) |
391 | return 0; | |
392 | htab->entries = nentries; | |
eed2b28c | 393 | htab->size = nsize; |
b4fe2683 JM |
394 | |
395 | htab->n_elements -= htab->n_deleted; | |
396 | htab->n_deleted = 0; | |
397 | ||
398 | p = oentries; | |
399 | do | |
400 | { | |
e0f3df8f | 401 | PTR x = *p; |
eb383413 | 402 | |
b4fe2683 JM |
403 | if (x != EMPTY_ENTRY && x != DELETED_ENTRY) |
404 | { | |
e0f3df8f | 405 | PTR *q = find_empty_slot_for_expand (htab, (*htab->hash_f) (x)); |
eb383413 | 406 | |
b4fe2683 JM |
407 | *q = x; |
408 | } | |
eb383413 | 409 | |
b4fe2683 JM |
410 | p++; |
411 | } | |
412 | while (p < olimit); | |
eb383413 | 413 | |
18893690 DD |
414 | if (htab->free_f != NULL) |
415 | (*htab->free_f) (oentries); | |
5f9624e3 DJ |
416 | else if (htab->free_with_arg_f != NULL) |
417 | (*htab->free_with_arg_f) (htab->alloc_arg, oentries); | |
99a4c1bd | 418 | return 1; |
e2eaf477 ILT |
419 | } |
420 | ||
b4fe2683 JM |
421 | /* This function searches for a hash table entry equal to the given |
422 | element. It cannot be used to insert or delete an element. */ | |
423 | ||
e0f3df8f | 424 | PTR |
b4fe2683 JM |
425 | htab_find_with_hash (htab, element, hash) |
426 | htab_t htab; | |
e0f3df8f | 427 | const PTR element; |
eb383413 | 428 | hashval_t hash; |
e2eaf477 | 429 | { |
eb383413 L |
430 | unsigned int index; |
431 | hashval_t hash2; | |
b4fe2683 | 432 | size_t size; |
e0f3df8f | 433 | PTR entry; |
e2eaf477 | 434 | |
b4fe2683 JM |
435 | htab->searches++; |
436 | size = htab->size; | |
b4fe2683 JM |
437 | index = hash % size; |
438 | ||
eb383413 L |
439 | entry = htab->entries[index]; |
440 | if (entry == EMPTY_ENTRY | |
441 | || (entry != DELETED_ENTRY && (*htab->eq_f) (entry, element))) | |
442 | return entry; | |
443 | ||
444 | hash2 = 1 + hash % (size - 2); | |
445 | ||
b4fe2683 | 446 | for (;;) |
e2eaf477 | 447 | { |
b4fe2683 JM |
448 | htab->collisions++; |
449 | index += hash2; | |
450 | if (index >= size) | |
451 | index -= size; | |
eb383413 L |
452 | |
453 | entry = htab->entries[index]; | |
454 | if (entry == EMPTY_ENTRY | |
455 | || (entry != DELETED_ENTRY && (*htab->eq_f) (entry, element))) | |
456 | return entry; | |
e2eaf477 | 457 | } |
b4fe2683 JM |
458 | } |
459 | ||
460 | /* Like htab_find_slot_with_hash, but compute the hash value from the | |
461 | element. */ | |
eb383413 | 462 | |
e0f3df8f | 463 | PTR |
b4fe2683 JM |
464 | htab_find (htab, element) |
465 | htab_t htab; | |
e0f3df8f | 466 | const PTR element; |
b4fe2683 JM |
467 | { |
468 | return htab_find_with_hash (htab, element, (*htab->hash_f) (element)); | |
469 | } | |
470 | ||
471 | /* This function searches for a hash table slot containing an entry | |
472 | equal to the given element. To delete an entry, call this with | |
473 | INSERT = 0, then call htab_clear_slot on the slot returned (possibly | |
474 | after doing some checks). To insert an entry, call this with | |
99a4c1bd HPN |
475 | INSERT = 1, then write the value you want into the returned slot. |
476 | When inserting an entry, NULL may be returned if memory allocation | |
477 | fails. */ | |
b4fe2683 | 478 | |
e0f3df8f | 479 | PTR * |
b4fe2683 JM |
480 | htab_find_slot_with_hash (htab, element, hash, insert) |
481 | htab_t htab; | |
e0f3df8f | 482 | const PTR element; |
eb383413 L |
483 | hashval_t hash; |
484 | enum insert_option insert; | |
b4fe2683 | 485 | { |
e0f3df8f | 486 | PTR *first_deleted_slot; |
eb383413 L |
487 | unsigned int index; |
488 | hashval_t hash2; | |
b4fe2683 | 489 | size_t size; |
b1c933fc | 490 | PTR entry; |
b4fe2683 | 491 | |
99a4c1bd HPN |
492 | if (insert == INSERT && htab->size * 3 <= htab->n_elements * 4 |
493 | && htab_expand (htab) == 0) | |
494 | return NULL; | |
b4fe2683 JM |
495 | |
496 | size = htab->size; | |
b4fe2683 JM |
497 | index = hash % size; |
498 | ||
e2eaf477 | 499 | htab->searches++; |
b4fe2683 JM |
500 | first_deleted_slot = NULL; |
501 | ||
b1c933fc RH |
502 | entry = htab->entries[index]; |
503 | if (entry == EMPTY_ENTRY) | |
504 | goto empty_entry; | |
505 | else if (entry == DELETED_ENTRY) | |
506 | first_deleted_slot = &htab->entries[index]; | |
507 | else if ((*htab->eq_f) (entry, element)) | |
508 | return &htab->entries[index]; | |
509 | ||
510 | hash2 = 1 + hash % (size - 2); | |
b4fe2683 | 511 | for (;;) |
e2eaf477 | 512 | { |
b1c933fc RH |
513 | htab->collisions++; |
514 | index += hash2; | |
515 | if (index >= size) | |
516 | index -= size; | |
517 | ||
518 | entry = htab->entries[index]; | |
b4fe2683 | 519 | if (entry == EMPTY_ENTRY) |
b1c933fc RH |
520 | goto empty_entry; |
521 | else if (entry == DELETED_ENTRY) | |
b4fe2683 JM |
522 | { |
523 | if (!first_deleted_slot) | |
524 | first_deleted_slot = &htab->entries[index]; | |
525 | } | |
b1c933fc | 526 | else if ((*htab->eq_f) (entry, element)) |
eb383413 | 527 | return &htab->entries[index]; |
e2eaf477 | 528 | } |
b1c933fc RH |
529 | |
530 | empty_entry: | |
531 | if (insert == NO_INSERT) | |
532 | return NULL; | |
533 | ||
534 | htab->n_elements++; | |
535 | ||
536 | if (first_deleted_slot) | |
537 | { | |
538 | *first_deleted_slot = EMPTY_ENTRY; | |
539 | return first_deleted_slot; | |
540 | } | |
541 | ||
542 | return &htab->entries[index]; | |
e2eaf477 ILT |
543 | } |
544 | ||
b4fe2683 JM |
545 | /* Like htab_find_slot_with_hash, but compute the hash value from the |
546 | element. */ | |
eb383413 | 547 | |
e0f3df8f | 548 | PTR * |
b4fe2683 JM |
549 | htab_find_slot (htab, element, insert) |
550 | htab_t htab; | |
e0f3df8f | 551 | const PTR element; |
eb383413 | 552 | enum insert_option insert; |
b4fe2683 JM |
553 | { |
554 | return htab_find_slot_with_hash (htab, element, (*htab->hash_f) (element), | |
555 | insert); | |
556 | } | |
557 | ||
558 | /* This function deletes an element with the given value from hash | |
559 | table. If there is no matching element in the hash table, this | |
560 | function does nothing. */ | |
e2eaf477 ILT |
561 | |
562 | void | |
b4fe2683 JM |
563 | htab_remove_elt (htab, element) |
564 | htab_t htab; | |
e0f3df8f | 565 | PTR element; |
e2eaf477 | 566 | { |
e0f3df8f | 567 | PTR *slot; |
b4fe2683 | 568 | |
eb383413 | 569 | slot = htab_find_slot (htab, element, NO_INSERT); |
b4fe2683 JM |
570 | if (*slot == EMPTY_ENTRY) |
571 | return; | |
572 | ||
573 | if (htab->del_f) | |
574 | (*htab->del_f) (*slot); | |
e2eaf477 | 575 | |
b4fe2683 JM |
576 | *slot = DELETED_ENTRY; |
577 | htab->n_deleted++; | |
e2eaf477 ILT |
578 | } |
579 | ||
b4fe2683 JM |
580 | /* This function clears a specified slot in a hash table. It is |
581 | useful when you've already done the lookup and don't want to do it | |
582 | again. */ | |
e2eaf477 ILT |
583 | |
584 | void | |
b4fe2683 JM |
585 | htab_clear_slot (htab, slot) |
586 | htab_t htab; | |
e0f3df8f | 587 | PTR *slot; |
e2eaf477 ILT |
588 | { |
589 | if (slot < htab->entries || slot >= htab->entries + htab->size | |
590 | || *slot == EMPTY_ENTRY || *slot == DELETED_ENTRY) | |
591 | abort (); | |
eb383413 | 592 | |
b4fe2683 JM |
593 | if (htab->del_f) |
594 | (*htab->del_f) (*slot); | |
eb383413 | 595 | |
e2eaf477 | 596 | *slot = DELETED_ENTRY; |
b4fe2683 | 597 | htab->n_deleted++; |
e2eaf477 ILT |
598 | } |
599 | ||
600 | /* This function scans over the entire hash table calling | |
601 | CALLBACK for each live entry. If CALLBACK returns false, | |
602 | the iteration stops. INFO is passed as CALLBACK's second | |
603 | argument. */ | |
604 | ||
605 | void | |
f77ed96c | 606 | htab_traverse_noresize (htab, callback, info) |
b4fe2683 JM |
607 | htab_t htab; |
608 | htab_trav callback; | |
e0f3df8f | 609 | PTR info; |
e2eaf477 | 610 | { |
c4d8feb2 DD |
611 | PTR *slot; |
612 | PTR *limit; | |
613 | ||
c4d8feb2 DD |
614 | slot = htab->entries; |
615 | limit = slot + htab->size; | |
eb383413 | 616 | |
b4fe2683 JM |
617 | do |
618 | { | |
e0f3df8f | 619 | PTR x = *slot; |
eb383413 | 620 | |
b4fe2683 JM |
621 | if (x != EMPTY_ENTRY && x != DELETED_ENTRY) |
622 | if (!(*callback) (slot, info)) | |
623 | break; | |
624 | } | |
625 | while (++slot < limit); | |
e2eaf477 ILT |
626 | } |
627 | ||
f77ed96c DD |
628 | /* Like htab_traverse_noresize, but does resize the table when it is |
629 | too empty to improve effectivity of subsequent calls. */ | |
630 | ||
631 | void | |
632 | htab_traverse (htab, callback, info) | |
633 | htab_t htab; | |
634 | htab_trav callback; | |
635 | PTR info; | |
636 | { | |
f77ed96c DD |
637 | if ((htab->n_elements - htab->n_deleted) * 8 < htab->size) |
638 | htab_expand (htab); | |
639 | ||
640 | htab_traverse_noresize (htab, callback, info); | |
641 | } | |
642 | ||
eb383413 | 643 | /* Return the current size of given hash table. */ |
e2eaf477 ILT |
644 | |
645 | size_t | |
b4fe2683 JM |
646 | htab_size (htab) |
647 | htab_t htab; | |
e2eaf477 ILT |
648 | { |
649 | return htab->size; | |
650 | } | |
651 | ||
eb383413 | 652 | /* Return the current number of elements in given hash table. */ |
e2eaf477 ILT |
653 | |
654 | size_t | |
b4fe2683 JM |
655 | htab_elements (htab) |
656 | htab_t htab; | |
e2eaf477 | 657 | { |
b4fe2683 | 658 | return htab->n_elements - htab->n_deleted; |
e2eaf477 ILT |
659 | } |
660 | ||
eb383413 L |
661 | /* Return the fraction of fixed collisions during all work with given |
662 | hash table. */ | |
e2eaf477 | 663 | |
b4fe2683 JM |
664 | double |
665 | htab_collisions (htab) | |
666 | htab_t htab; | |
e2eaf477 | 667 | { |
eb383413 | 668 | if (htab->searches == 0) |
b4fe2683 | 669 | return 0.0; |
eb383413 L |
670 | |
671 | return (double) htab->collisions / (double) htab->searches; | |
e2eaf477 | 672 | } |
8fc34799 | 673 | |
68a41de7 DD |
674 | /* Hash P as a null-terminated string. |
675 | ||
676 | Copied from gcc/hashtable.c. Zack had the following to say with respect | |
677 | to applicability, though note that unlike hashtable.c, this hash table | |
678 | implementation re-hashes rather than chain buckets. | |
679 | ||
680 | http://gcc.gnu.org/ml/gcc-patches/2001-08/msg01021.html | |
681 | From: Zack Weinberg <zackw@panix.com> | |
682 | Date: Fri, 17 Aug 2001 02:15:56 -0400 | |
683 | ||
684 | I got it by extracting all the identifiers from all the source code | |
685 | I had lying around in mid-1999, and testing many recurrences of | |
686 | the form "H_n = H_{n-1} * K + c_n * L + M" where K, L, M were either | |
687 | prime numbers or the appropriate identity. This was the best one. | |
688 | I don't remember exactly what constituted "best", except I was | |
689 | looking at bucket-length distributions mostly. | |
690 | ||
691 | So it should be very good at hashing identifiers, but might not be | |
692 | as good at arbitrary strings. | |
693 | ||
694 | I'll add that it thoroughly trounces the hash functions recommended | |
695 | for this use at http://burtleburtle.net/bob/hash/index.html, both | |
696 | on speed and bucket distribution. I haven't tried it against the | |
697 | function they just started using for Perl's hashes. */ | |
8fc34799 DD |
698 | |
699 | hashval_t | |
700 | htab_hash_string (p) | |
701 | const PTR p; | |
702 | { | |
703 | const unsigned char *str = (const unsigned char *) p; | |
704 | hashval_t r = 0; | |
705 | unsigned char c; | |
706 | ||
707 | while ((c = *str++) != 0) | |
708 | r = r * 67 + c - 113; | |
709 | ||
710 | return r; | |
711 | } | |
7108c5dc JM |
712 | |
713 | /* DERIVED FROM: | |
714 | -------------------------------------------------------------------- | |
715 | lookup2.c, by Bob Jenkins, December 1996, Public Domain. | |
716 | hash(), hash2(), hash3, and mix() are externally useful functions. | |
717 | Routines to test the hash are included if SELF_TEST is defined. | |
718 | You can use this free for any purpose. It has no warranty. | |
719 | -------------------------------------------------------------------- | |
720 | */ | |
721 | ||
722 | /* | |
723 | -------------------------------------------------------------------- | |
724 | mix -- mix 3 32-bit values reversibly. | |
725 | For every delta with one or two bit set, and the deltas of all three | |
726 | high bits or all three low bits, whether the original value of a,b,c | |
727 | is almost all zero or is uniformly distributed, | |
728 | * If mix() is run forward or backward, at least 32 bits in a,b,c | |
729 | have at least 1/4 probability of changing. | |
730 | * If mix() is run forward, every bit of c will change between 1/3 and | |
731 | 2/3 of the time. (Well, 22/100 and 78/100 for some 2-bit deltas.) | |
732 | mix() was built out of 36 single-cycle latency instructions in a | |
733 | structure that could supported 2x parallelism, like so: | |
734 | a -= b; | |
735 | a -= c; x = (c>>13); | |
736 | b -= c; a ^= x; | |
737 | b -= a; x = (a<<8); | |
738 | c -= a; b ^= x; | |
739 | c -= b; x = (b>>13); | |
740 | ... | |
741 | Unfortunately, superscalar Pentiums and Sparcs can't take advantage | |
742 | of that parallelism. They've also turned some of those single-cycle | |
743 | latency instructions into multi-cycle latency instructions. Still, | |
744 | this is the fastest good hash I could find. There were about 2^^68 | |
745 | to choose from. I only looked at a billion or so. | |
746 | -------------------------------------------------------------------- | |
747 | */ | |
748 | /* same, but slower, works on systems that might have 8 byte hashval_t's */ | |
749 | #define mix(a,b,c) \ | |
750 | { \ | |
751 | a -= b; a -= c; a ^= (c>>13); \ | |
752 | b -= c; b -= a; b ^= (a<< 8); \ | |
753 | c -= a; c -= b; c ^= ((b&0xffffffff)>>13); \ | |
754 | a -= b; a -= c; a ^= ((c&0xffffffff)>>12); \ | |
755 | b -= c; b -= a; b = (b ^ (a<<16)) & 0xffffffff; \ | |
756 | c -= a; c -= b; c = (c ^ (b>> 5)) & 0xffffffff; \ | |
757 | a -= b; a -= c; a = (a ^ (c>> 3)) & 0xffffffff; \ | |
758 | b -= c; b -= a; b = (b ^ (a<<10)) & 0xffffffff; \ | |
759 | c -= a; c -= b; c = (c ^ (b>>15)) & 0xffffffff; \ | |
760 | } | |
761 | ||
762 | /* | |
763 | -------------------------------------------------------------------- | |
764 | hash() -- hash a variable-length key into a 32-bit value | |
765 | k : the key (the unaligned variable-length array of bytes) | |
766 | len : the length of the key, counting by bytes | |
767 | level : can be any 4-byte value | |
768 | Returns a 32-bit value. Every bit of the key affects every bit of | |
769 | the return value. Every 1-bit and 2-bit delta achieves avalanche. | |
770 | About 36+6len instructions. | |
771 | ||
772 | The best hash table sizes are powers of 2. There is no need to do | |
773 | mod a prime (mod is sooo slow!). If you need less than 32 bits, | |
774 | use a bitmask. For example, if you need only 10 bits, do | |
775 | h = (h & hashmask(10)); | |
776 | In which case, the hash table should have hashsize(10) elements. | |
777 | ||
778 | If you are hashing n strings (ub1 **)k, do it like this: | |
779 | for (i=0, h=0; i<n; ++i) h = hash( k[i], len[i], h); | |
780 | ||
781 | By Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may use this | |
782 | code any way you wish, private, educational, or commercial. It's free. | |
783 | ||
784 | See http://burtleburtle.net/bob/hash/evahash.html | |
785 | Use for hash table lookup, or anything where one collision in 2^32 is | |
786 | acceptable. Do NOT use for cryptographic purposes. | |
787 | -------------------------------------------------------------------- | |
788 | */ | |
789 | ||
eafaf5eb | 790 | hashval_t iterative_hash (k_in, length, initval) |
7108c5dc JM |
791 | const PTR k_in; /* the key */ |
792 | register size_t length; /* the length of the key */ | |
793 | register hashval_t initval; /* the previous hash, or an arbitrary value */ | |
794 | { | |
795 | register const unsigned char *k = (const unsigned char *)k_in; | |
796 | register hashval_t a,b,c,len; | |
797 | ||
798 | /* Set up the internal state */ | |
799 | len = length; | |
800 | a = b = 0x9e3779b9; /* the golden ratio; an arbitrary value */ | |
801 | c = initval; /* the previous hash value */ | |
802 | ||
803 | /*---------------------------------------- handle most of the key */ | |
804 | #ifndef WORDS_BIGENDIAN | |
805 | /* On a little-endian machine, if the data is 4-byte aligned we can hash | |
806 | by word for better speed. This gives nondeterministic results on | |
807 | big-endian machines. */ | |
808 | if (sizeof (hashval_t) == 4 && (((size_t)k)&3) == 0) | |
809 | while (len >= 12) /* aligned */ | |
810 | { | |
811 | a += *(hashval_t *)(k+0); | |
812 | b += *(hashval_t *)(k+4); | |
813 | c += *(hashval_t *)(k+8); | |
814 | mix(a,b,c); | |
815 | k += 12; len -= 12; | |
816 | } | |
817 | else /* unaligned */ | |
818 | #endif | |
819 | while (len >= 12) | |
820 | { | |
821 | a += (k[0] +((hashval_t)k[1]<<8) +((hashval_t)k[2]<<16) +((hashval_t)k[3]<<24)); | |
822 | b += (k[4] +((hashval_t)k[5]<<8) +((hashval_t)k[6]<<16) +((hashval_t)k[7]<<24)); | |
823 | c += (k[8] +((hashval_t)k[9]<<8) +((hashval_t)k[10]<<16)+((hashval_t)k[11]<<24)); | |
824 | mix(a,b,c); | |
825 | k += 12; len -= 12; | |
826 | } | |
827 | ||
828 | /*------------------------------------- handle the last 11 bytes */ | |
829 | c += length; | |
830 | switch(len) /* all the case statements fall through */ | |
831 | { | |
832 | case 11: c+=((hashval_t)k[10]<<24); | |
833 | case 10: c+=((hashval_t)k[9]<<16); | |
834 | case 9 : c+=((hashval_t)k[8]<<8); | |
835 | /* the first byte of c is reserved for the length */ | |
836 | case 8 : b+=((hashval_t)k[7]<<24); | |
837 | case 7 : b+=((hashval_t)k[6]<<16); | |
838 | case 6 : b+=((hashval_t)k[5]<<8); | |
839 | case 5 : b+=k[4]; | |
840 | case 4 : a+=((hashval_t)k[3]<<24); | |
841 | case 3 : a+=((hashval_t)k[2]<<16); | |
842 | case 2 : a+=((hashval_t)k[1]<<8); | |
843 | case 1 : a+=k[0]; | |
844 | /* case 0: nothing left to add */ | |
845 | } | |
846 | mix(a,b,c); | |
847 | /*-------------------------------------------- report the result */ | |
848 | return c; | |
849 | } |