2 Copyright (C) 2000, 2001 Free Software Foundation, Inc.
4 This file is part of GNU Wget.
6 GNU Wget is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or (at
9 your option) any later version.
11 GNU Wget 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
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with Wget; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 In addition, as a special exception, the Free Software Foundation
21 gives permission to link the code of its release of Wget with the
22 OpenSSL project's "OpenSSL" library (or with modified versions of it
23 that use the same license as the "OpenSSL" library), and distribute
24 the linked executables. You must obey the GNU General Public License
25 in all respects for all of the code used other than "OpenSSL". If you
26 modify this file, you may extend this exception to your version of the
27 file, but you are not obligated to do so. If you do not wish to do
28 so, delete this exception statement from your version. */
38 #endif /* HAVE_STRING_H */
52 # define xmalloc malloc
53 # define xrealloc realloc
57 # define TOLOWER(x) ('A' <= (x) && (x) <= 'Z' ? (x) - 32 : (x))
62 Hash tables are a technique used to implement mapping between
63 objects with near-constant-time access and storage. The table
64 associates keys to values, and a value can be very quickly
65 retrieved by providing the key. Fast lookup tables are typically
66 implemented as hash tables.
69 hash_table_new -- creates the table.
70 hash_table_destroy -- destroys the table.
71 hash_table_put -- establishes or updates key->value mapping.
72 hash_table_get -- retrieves value of key.
73 hash_table_get_pair -- get key/value pair for key.
74 hash_table_contains -- test whether the table contains key.
75 hash_table_remove -- remove the key->value mapping for key.
76 hash_table_map -- iterate through table mappings.
77 hash_table_clear -- clear hash table contents.
78 hash_table_count -- return the number of entries in the table.
80 The hash table grows internally as new entries are added and is not
81 limited in size, except by available memory. The table doubles
82 with each resize, which ensures that the amortized time per
83 operation remains constant.
85 By default, tables created by hash_table_new consider the keys to
86 be equal if their pointer values are the same. You can use
87 make_string_hash_table to create tables whose keys are considered
88 equal if their string contents are the same. In the general case,
89 the criterion of equality used to compare keys is specified at
90 table creation time with two callback functions, "hash" and "test".
91 The hash function transforms the key into an arbitrary number that
92 must be the same for two equal keys. The test function accepts two
93 keys and returns non-zero if they are to be considered equal.
95 Note that neither keys nor values are copied when inserted into the
96 hash table, so they must exist for the lifetime of the table. This
97 means that e.g. the use of static strings is OK, but objects with a
98 shorter life-time need to be copied (with strdup() or the like in
99 the case of strings) before being inserted. */
103 The hash table is implemented as an open-addressed table with
104 linear probing collision resolution.
106 For those not up to CS parlance, it means that all the hash entries
107 (pairs of pointers key and value) are stored in a contiguous array.
108 The position of each mapping is determined by the hash value of its
109 key and the size of the table: location := hash(key) % size. If
110 two different keys end up on the same position (collide), the one
111 that came second is placed at the next empty position following the
112 occupied place. This collision resolution technique is called
115 There are more advanced collision resolution methods (quadratic
116 probing, double hashing), but we don't use them because they incur
117 more non-sequential access to the array, which results in worse CPU
118 cache behavior. Linear probing works well as long as the
119 count/size ratio (fullness) is kept below 75%. We make sure to
120 grow and rehash the table whenever this threshold is exceeded.
122 Collisions make deletion tricky because clearing a position
123 followed by a colliding entry would make the position seem empty
124 and the colliding entry not found. One solution is to leave a
125 "tombstone" instead of clearing the entry, and another is to
126 carefully rehash the entries immediately following the deleted one.
127 We use the latter method because it results in less bookkeeping and
128 faster retrieval at the (slight) expense of deletion. */
130 /* Maximum allowed fullness: when hash table's fullness exceeds this
131 value, the table is resized. */
132 #define HASH_MAX_FULLNESS 0.75
134 /* The hash table size is multiplied by this factor (and then rounded
135 to the next prime) with each resize. This guarantees infrequent
137 #define HASH_RESIZE_FACTOR 2
145 unsigned long (*hash_function) PARAMS ((const void *));
146 int (*test_function) PARAMS ((const void *, const void *));
148 int size; /* size of the array. */
149 int count; /* number of non-empty entries. */
151 int resize_threshold; /* after size exceeds this number of
152 entries, resize the table. */
153 int prime_offset; /* the offset of the current prime in
156 struct mapping *mappings; /* the array of mapping pairs. */
159 /* We use all-bit-set marker to mean that a mapping is empty. It is
160 (hopefully) illegal as a pointer, and it allows the users to use
161 NULL (as well as any non-negative integer) as key. */
162 #define NON_EMPTY(mp) (mp->key != (void *)~(unsigned long)0)
164 /* "Next" mapping is the mapping after MP, but wrapping back to
165 MAPPINGS when MP would reach MAPPINGS+SIZE. */
166 #define NEXT_MAPPING(mp, mappings, size) (mp != mappings + (size - 1) \
169 /* Loop over non-empty mappings starting at MP. */
170 #define LOOP_NON_EMPTY(mp, mappings, size) \
171 for (; NON_EMPTY (mp); mp = NEXT_MAPPING (mp, mappings, size))
173 /* #### Some implementations multiply the hash with the "golden ratio"
174 of the table to get better spread for keys that do not come from a
175 good hashing source. I'm not sure if that is necessary for the
176 hash functions we use. */
178 #define HASH_POSITION(ht, key) (ht->hash_function (key) % ht->size)
180 /* Find a prime near, but greather than or equal to SIZE. Of course,
181 the primes are not calculated, but looked up from a table. The
182 table does not contain all primes in range, just a selection useful
185 PRIME_OFFSET is a minor optimization: if specified, it starts the
186 search for the prime number beginning with the specific offset in
187 the prime number table. The final offset is stored in the same
191 prime_size (int size, int *prime_offset)
193 static const unsigned long primes [] = {
194 13, 19, 29, 41, 59, 79, 107, 149, 197, 263, 347, 457, 599, 787, 1031,
195 1361, 1777, 2333, 3037, 3967, 5167, 6719, 8737, 11369, 14783,
196 19219, 24989, 32491, 42257, 54941, 71429, 92861, 120721, 156941,
197 204047, 265271, 344857, 448321, 582821, 757693, 985003, 1280519,
198 1664681, 2164111, 2813353, 3657361, 4754591, 6180989, 8035301,
199 10445899, 13579681, 17653589, 22949669, 29834603, 38784989,
200 50420551, 65546729, 85210757, 110774011, 144006217, 187208107,
201 243370577, 316381771, 411296309, 534685237, 695090819, 903618083,
202 1174703521, 1527114613, 1985248999,
203 (unsigned long)0x99d43ea5, (unsigned long)0xc7fa5177
205 int i = *prime_offset;
207 for (; i < countof (primes); i++)
208 if (primes[i] >= size)
210 /* Set the offset to the next prime. That is safe because,
211 next time we are called, it will be with a larger SIZE,
212 which means we could never return the same prime anyway.
213 (If that is not the case, the caller can simply reset
215 *prime_offset = i + 1;
223 static unsigned long ptrhash PARAMS ((const void *));
224 static int ptrcmp PARAMS ((const void *, const void *));
226 /* Create a hash table with hash function HASH_FUNCTION and test
227 function TEST_FUNCTION. The table is empty (its count is 0), but
228 pre-allocated to store at least ITEMS items.
230 ITEMS is the number of items that the table can accept without
231 needing to resize. It is useful when creating a table that is to
232 be immediately filled with a known number of items. In that case,
233 the regrows are a waste of time, and specifying ITEMS correctly
234 will avoid them altogether.
236 Note that hash tables grow dynamically regardless of ITEMS. The
237 only use of ITEMS is to preallocate the table and avoid unnecessary
238 dynamic regrows. Don't bother making ITEMS prime because it's not
239 used as size unchanged. To start with a small table that grows as
240 needed, simply specify zero ITEMS.
242 If HASH_FUNCTION is not provided, identity table is assumed,
243 i.e. key pointers are compared as keys. If you want strings with
244 equal contents to hash the same, use make_string_hash_table. */
247 hash_table_new (int items,
248 unsigned long (*hash_function) (const void *),
249 int (*test_function) (const void *, const void *))
252 struct hash_table *ht = xnew (struct hash_table);
254 ht->hash_function = hash_function ? hash_function : ptrhash;
255 ht->test_function = test_function ? test_function : ptrcmp;
257 /* If the size of struct hash_table ever becomes a concern, this
258 field can go. (Wget doesn't create many hashes.) */
259 ht->prime_offset = 0;
261 /* Calculate the size that ensures that the table will store at
262 least ITEMS keys without the need to resize. */
263 size = 1 + items / HASH_MAX_FULLNESS;
264 size = prime_size (size, &ht->prime_offset);
266 ht->resize_threshold = size * HASH_MAX_FULLNESS;
267 /*assert (ht->resize_threshold >= items);*/
269 ht->mappings = xnew_array (struct mapping, ht->size);
270 /* Mark mappings as empty. We use 0xff rather than 0 to mark empty
271 keys because it allows us to store NULL keys to the table. */
272 memset (ht->mappings, 255, size * sizeof (struct mapping));
279 /* Free the data associated with hash table HT. */
282 hash_table_destroy (struct hash_table *ht)
284 xfree (ht->mappings);
288 /* The heart of most functions in this file -- find the mapping whose
289 KEY is equal to key, using linear probing. Returns the mapping
290 that matches KEY, or the first empty mapping if none matches. */
292 static inline struct mapping *
293 find_mapping (const struct hash_table *ht, const void *key)
295 struct mapping *mappings = ht->mappings;
297 struct mapping *mp = mappings + HASH_POSITION (ht, key);
298 int (*equals) PARAMS ((const void *, const void *)) = ht->test_function;
300 LOOP_NON_EMPTY (mp, mappings, size)
301 if (equals (key, mp->key))
306 /* Get the value that corresponds to the key KEY in the hash table HT.
307 If no value is found, return NULL. Note that NULL is a legal value
308 for value; if you are storing NULLs in your hash table, you can use
309 hash_table_contains to be sure that a (possibly NULL) value exists
310 in the table. Or, you can use hash_table_get_pair instead of this
314 hash_table_get (const struct hash_table *ht, const void *key)
316 struct mapping *mp = find_mapping (ht, key);
323 /* Like hash_table_get, but writes out the pointers to both key and
324 value. Returns non-zero on success. */
327 hash_table_get_pair (const struct hash_table *ht, const void *lookup_key,
328 void *orig_key, void *value)
330 struct mapping *mp = find_mapping (ht, lookup_key);
334 *(void **)orig_key = mp->key;
336 *(void **)value = mp->value;
343 /* Return 1 if HT contains KEY, 0 otherwise. */
346 hash_table_contains (const struct hash_table *ht, const void *key)
348 struct mapping *mp = find_mapping (ht, key);
349 return NON_EMPTY (mp);
352 /* Grow hash table HT as necessary, and rehash all the key-value
356 grow_hash_table (struct hash_table *ht)
358 struct mapping *old_mappings = ht->mappings;
359 struct mapping *old_end = ht->mappings + ht->size;
360 struct mapping *mp, *mappings;
363 newsize = prime_size (ht->size * HASH_RESIZE_FACTOR, &ht->prime_offset);
365 printf ("growing from %d to %d; fullness %.2f%% to %.2f%%\n",
367 100.0 * ht->count / ht->size,
368 100.0 * ht->count / newsize);
372 ht->resize_threshold = newsize * HASH_MAX_FULLNESS;
374 mappings = xnew_array (struct mapping, newsize);
375 memset (mappings, 255, newsize * sizeof (struct mapping));
376 ht->mappings = mappings;
378 for (mp = old_mappings; mp < old_end; mp++)
381 struct mapping *new_mp = mappings + HASH_POSITION (ht, mp->key);
382 /* We don't need to test for uniqueness of keys because they
383 come from the hash table and are therefore known to be
385 LOOP_NON_EMPTY (new_mp, mappings, newsize)
390 xfree (old_mappings);
393 /* Put VALUE in the hash table HT under the key KEY. This regrows the
394 table if necessary. */
397 hash_table_put (struct hash_table *ht, const void *key, void *value)
399 struct mapping *mp = find_mapping (ht, key);
402 /* update existing item */
403 mp->key = (void *)key; /* const? */
408 /* If adding the item would make the table exceed max. fullness,
409 grow the table first. */
410 if (ht->count >= ht->resize_threshold)
412 grow_hash_table (ht);
413 mp = find_mapping (ht, key);
418 mp->key = (void *)key; /* const? */
422 /* Remove a mapping that matches KEY from HT. Return 0 if there was
423 no such entry; return 1 if an entry was removed. */
426 hash_table_remove (struct hash_table *ht, const void *key)
428 struct mapping *mp = find_mapping (ht, key);
434 struct mapping *mappings = ht->mappings;
439 /* Rehash all the entries following MP. The alternative
440 approach is to mark the entry as deleted, i.e. create a
441 "tombstone". That makes remove faster, but leaves a lot of
442 garbage and slows down hash_table_get and hash_table_put. */
444 mp = NEXT_MAPPING (mp, mappings, size);
445 LOOP_NON_EMPTY (mp, mappings, size)
447 const void *key2 = mp->key;
448 struct mapping *mp_new = mappings + HASH_POSITION (ht, key2);
450 /* Find the new location for the key. */
452 LOOP_NON_EMPTY (mp_new, mappings, size)
453 if (key2 == mp_new->key)
454 /* The mapping MP (key2) is already where we want it (in
455 MP_NEW's "chain" of keys.) */
468 /* Clear HT of all entries. After calling this function, the count
469 and the fullness of the hash table will be zero. The size will
473 hash_table_clear (struct hash_table *ht)
475 memset (ht->mappings, '\0', ht->size * sizeof (struct mapping));
479 /* Map MAPFUN over all the mappings in hash table HT. MAPFUN is
480 called with three arguments: the key, the value, and MAPARG.
482 It is undefined what happens if you add or remove entries in the
483 hash table while hash_table_map is running. The exception is the
484 entry you're currently mapping over; you may remove or change that
488 hash_table_map (struct hash_table *ht,
489 int (*mapfun) (void *, void *, void *),
492 struct mapping *mp = ht->mappings;
493 struct mapping *end = ht->mappings + ht->size;
495 for (; mp < end; mp++)
501 if (mapfun (key, mp->value, maparg))
503 /* hash_table_remove might have moved the adjacent
505 if (mp->key != key && NON_EMPTY (mp))
510 /* Return the number of elements in the hash table. This is not the
511 same as the physical size of the hash table, which is always
512 greater than the number of elements. */
515 hash_table_count (const struct hash_table *ht)
520 /* Functions from this point onward are meant for convenience and
521 don't strictly belong to this file. However, this is as good a
522 place for them as any. */
524 /* Rules for creating custom hash and test functions:
526 - The test function returns non-zero for keys that are considered
527 "equal", zero otherwise.
529 - The hash function returns a number that represents the
530 "distinctness" of the object. In more precise terms, it means
531 that for any two objects that test "equal" under the test
532 function, the hash function MUST produce the same result.
534 This does not mean that all different objects must produce
535 different values (that would be "perfect" hashing), only that
536 non-distinct objects must produce the same values! For instance,
537 a hash function that returns 0 for any given object is a
538 perfectly valid (albeit extremely bad) hash function. A hash
539 function that hashes a string by adding up all its characters is
540 another example of a valid (but quite bad) hash function.
542 It is not hard to make hash and test functions agree about
543 equality. For example, if the test function compares strings
544 case-insensitively, the hash function can lower-case the
545 characters when calculating the hash value. That ensures that
546 two strings differing only in case will hash the same.
548 - If you care about performance, choose a hash function with as
549 good "spreading" as possible. A good hash function will use all
550 the bits of the input when calculating the hash, and will react
551 to even small changes in input with a completely different
552 output. Finally, don't make the hash function itself overly
553 slow, because you'll be incurring a non-negligible overhead to
554 all hash table operations. */
557 * Support for hash tables whose keys are strings.
561 /* 31 bit hash function. Taken from Gnome's glib, modified to use
564 We used to use the popular hash function from the Dragon Book, but
565 this one seems to perform much better. */
568 string_hash (const void *key)
574 for (p += 1; *p != '\0'; p++)
575 h = (h << 5) - h + *p;
580 /* Frontend for strcmp usable for hash tables. */
583 string_cmp (const void *s1, const void *s2)
585 return !strcmp ((const char *)s1, (const char *)s2);
588 /* Return a hash table of preallocated to store at least ITEMS items
589 suitable to use strings as keys. */
592 make_string_hash_table (int items)
594 return hash_table_new (items, string_hash, string_cmp);
598 * Support for hash tables whose keys are strings, but which are
599 * compared case-insensitively.
603 /* Like string_hash, but produce the same hash regardless of the case. */
606 string_hash_nocase (const void *key)
609 unsigned int h = TOLOWER (*p);
612 for (p += 1; *p != '\0'; p++)
613 h = (h << 5) - h + TOLOWER (*p);
618 /* Like string_cmp, but doing case-insensitive compareison. */
621 string_cmp_nocase (const void *s1, const void *s2)
623 return !strcasecmp ((const char *)s1, (const char *)s2);
626 /* Like make_string_hash_table, but uses string_hash_nocase and
627 string_cmp_nocase. */
630 make_nocase_string_hash_table (int items)
632 return hash_table_new (items, string_hash_nocase, string_cmp_nocase);
635 /* Hashing of pointers. Used for hash tables that are keyed by
636 pointer identity. (Common Lisp calls them EQ hash tables, and Java
637 calls them IdentityHashMaps.) */
640 ptrhash (const void *ptr)
642 unsigned long key = (unsigned long)ptr;
665 ptrcmp (const void *ptr1, const void *ptr2)
676 print_hash_table_mapper (void *key, void *value, void *count)
679 printf ("%s: %s\n", (const char *)key, (char *)value);
684 print_hash (struct hash_table *sht)
687 hash_table_map (sht, print_hash_table_mapper, &debug_count);
688 assert (debug_count == sht->count);
694 struct hash_table *ht = make_string_hash_table (0);
696 while ((fgets (line, sizeof (line), stdin)))
698 int len = strlen (line);
702 if (!hash_table_contains (ht, line))
703 hash_table_put (ht, strdup (line), "here I am!");
708 if (hash_table_get_pair (ht, line, &line_copy, NULL))
710 hash_table_remove (ht, line);
720 printf ("%d %d\n", ht->count, ht->size);