1 /* hash - hashing table processing.
3 Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2006, 2007,
4 2009 Free Software Foundation, Inc.
6 Written by Jim Meyering, 1992.
8 This program is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program. If not, see <http://www.gnu.org/licenses/>. */
21 /* A generic hash table package. */
23 /* Define USE_OBSTACK to 1 if you want the allocator to use obstacks instead
24 of malloc. If you change USE_OBSTACK, you have to recompile! */
30 #include "bitrotate.h"
39 # ifndef obstack_chunk_alloc
40 # define obstack_chunk_alloc malloc
42 # ifndef obstack_chunk_free
43 # define obstack_chunk_free free
50 struct hash_entry *next;
55 /* The array of buckets starts at BUCKET and extends to BUCKET_LIMIT-1,
56 for a possibility of N_BUCKETS. Among those, N_BUCKETS_USED buckets
57 are not empty, there are N_ENTRIES active entries in the table. */
58 struct hash_entry *bucket;
59 struct hash_entry const *bucket_limit;
61 size_t n_buckets_used;
64 /* Tuning arguments, kept in a physically separate structure. */
65 const Hash_tuning *tuning;
67 /* Three functions are given to `hash_initialize', see the documentation
68 block for this function. In a word, HASHER randomizes a user entry
69 into a number up from 0 up to some maximum minus 1; COMPARATOR returns
70 true if two user entries compare equally; and DATA_FREER is the cleanup
71 function for a user entry. */
73 Hash_comparator comparator;
74 Hash_data_freer data_freer;
76 /* A linked list of freed struct hash_entry structs. */
77 struct hash_entry *free_entry_list;
80 /* Whenever obstacks are used, it is possible to allocate all overflowed
81 entries into a single stack, so they all can be freed in a single
82 operation. It is not clear if the speedup is worth the trouble. */
83 struct obstack entry_stack;
87 /* A hash table contains many internal entries, each holding a pointer to
88 some user-provided data (also called a user entry). An entry indistinctly
89 refers to both the internal entry and its associated user entry. A user
90 entry contents may be hashed by a randomization function (the hashing
91 function, or just `hasher' for short) into a number (or `slot') between 0
92 and the current table size. At each slot position in the hash table,
93 starts a linked chain of entries for which the user data all hash to this
94 slot. A bucket is the collection of all entries hashing to the same slot.
96 A good `hasher' function will distribute entries rather evenly in buckets.
97 In the ideal case, the length of each bucket is roughly the number of
98 entries divided by the table size. Finding the slot for a data is usually
99 done in constant time by the `hasher', and the later finding of a precise
100 entry is linear in time with the size of the bucket. Consequently, a
101 larger hash table size (that is, a larger number of buckets) is prone to
102 yielding shorter chains, *given* the `hasher' function behaves properly.
104 Long buckets slow down the lookup algorithm. One might use big hash table
105 sizes in hope to reduce the average length of buckets, but this might
106 become inordinate, as unused slots in the hash table take some space. The
107 best bet is to make sure you are using a good `hasher' function (beware
108 that those are not that easy to write! :-), and to use a table size
109 larger than the actual number of entries. */
111 /* If an insertion makes the ratio of nonempty buckets to table size larger
112 than the growth threshold (a number between 0.0 and 1.0), then increase
113 the table size by multiplying by the growth factor (a number greater than
114 1.0). The growth threshold defaults to 0.8, and the growth factor
115 defaults to 1.414, meaning that the table will have doubled its size
116 every second time 80% of the buckets get used. */
117 #define DEFAULT_GROWTH_THRESHOLD 0.8
118 #define DEFAULT_GROWTH_FACTOR 1.414
120 /* If a deletion empties a bucket and causes the ratio of used buckets to
121 table size to become smaller than the shrink threshold (a number between
122 0.0 and 1.0), then shrink the table by multiplying by the shrink factor (a
123 number greater than the shrink threshold but smaller than 1.0). The shrink
124 threshold and factor default to 0.0 and 1.0, meaning that the table never
126 #define DEFAULT_SHRINK_THRESHOLD 0.0
127 #define DEFAULT_SHRINK_FACTOR 1.0
129 /* Use this to initialize or reset a TUNING structure to
130 some sensible values. */
131 static const Hash_tuning default_tuning =
133 DEFAULT_SHRINK_THRESHOLD,
134 DEFAULT_SHRINK_FACTOR,
135 DEFAULT_GROWTH_THRESHOLD,
136 DEFAULT_GROWTH_FACTOR,
140 /* Information and lookup. */
142 /* The following few functions provide information about the overall hash
143 table organization: the number of entries, number of buckets and maximum
144 length of buckets. */
146 /* Return the number of buckets in the hash table. The table size, the total
147 number of buckets (used plus unused), or the maximum number of slots, are
148 the same quantity. */
151 hash_get_n_buckets (const Hash_table *table)
153 return table->n_buckets;
156 /* Return the number of slots in use (non-empty buckets). */
159 hash_get_n_buckets_used (const Hash_table *table)
161 return table->n_buckets_used;
164 /* Return the number of active entries. */
167 hash_get_n_entries (const Hash_table *table)
169 return table->n_entries;
172 /* Return the length of the longest chain (bucket). */
175 hash_get_max_bucket_length (const Hash_table *table)
177 struct hash_entry const *bucket;
178 size_t max_bucket_length = 0;
180 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
184 struct hash_entry const *cursor = bucket;
185 size_t bucket_length = 1;
187 while (cursor = cursor->next, cursor)
190 if (bucket_length > max_bucket_length)
191 max_bucket_length = bucket_length;
195 return max_bucket_length;
198 /* Do a mild validation of a hash table, by traversing it and checking two
202 hash_table_ok (const Hash_table *table)
204 struct hash_entry const *bucket;
205 size_t n_buckets_used = 0;
206 size_t n_entries = 0;
208 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
212 struct hash_entry const *cursor = bucket;
214 /* Count bucket head. */
218 /* Count bucket overflow. */
219 while (cursor = cursor->next, cursor)
224 if (n_buckets_used == table->n_buckets_used && n_entries == table->n_entries)
231 hash_print_statistics (const Hash_table *table, FILE *stream)
233 size_t n_entries = hash_get_n_entries (table);
234 size_t n_buckets = hash_get_n_buckets (table);
235 size_t n_buckets_used = hash_get_n_buckets_used (table);
236 size_t max_bucket_length = hash_get_max_bucket_length (table);
238 fprintf (stream, "# entries: %lu\n", (unsigned long int) n_entries);
239 fprintf (stream, "# buckets: %lu\n", (unsigned long int) n_buckets);
240 fprintf (stream, "# buckets used: %lu (%.2f%%)\n",
241 (unsigned long int) n_buckets_used,
242 (100.0 * n_buckets_used) / n_buckets);
243 fprintf (stream, "max bucket length: %lu\n",
244 (unsigned long int) max_bucket_length);
247 /* If ENTRY matches an entry already in the hash table, return the
248 entry from the table. Otherwise, return NULL. */
251 hash_lookup (const Hash_table *table, const void *entry)
253 struct hash_entry const *bucket
254 = table->bucket + table->hasher (entry, table->n_buckets);
255 struct hash_entry const *cursor;
257 if (! (bucket < table->bucket_limit))
260 if (bucket->data == NULL)
263 for (cursor = bucket; cursor; cursor = cursor->next)
264 if (entry == cursor->data || table->comparator (entry, cursor->data))
272 /* The functions in this page traverse the hash table and process the
273 contained entries. For the traversal to work properly, the hash table
274 should not be resized nor modified while any particular entry is being
275 processed. In particular, entries should not be added, and an entry
276 may be removed only if there is no shrink threshold and the entry being
277 removed has already been passed to hash_get_next. */
279 /* Return the first data in the table, or NULL if the table is empty. */
282 hash_get_first (const Hash_table *table)
284 struct hash_entry const *bucket;
286 if (table->n_entries == 0)
289 for (bucket = table->bucket; ; bucket++)
290 if (! (bucket < table->bucket_limit))
292 else if (bucket->data)
296 /* Return the user data for the entry following ENTRY, where ENTRY has been
297 returned by a previous call to either `hash_get_first' or `hash_get_next'.
298 Return NULL if there are no more entries. */
301 hash_get_next (const Hash_table *table, const void *entry)
303 struct hash_entry const *bucket
304 = table->bucket + table->hasher (entry, table->n_buckets);
305 struct hash_entry const *cursor;
307 if (! (bucket < table->bucket_limit))
310 /* Find next entry in the same bucket. */
311 for (cursor = bucket; cursor; cursor = cursor->next)
312 if (cursor->data == entry && cursor->next)
313 return cursor->next->data;
315 /* Find first entry in any subsequent bucket. */
316 while (++bucket < table->bucket_limit)
324 /* Fill BUFFER with pointers to active user entries in the hash table, then
325 return the number of pointers copied. Do not copy more than BUFFER_SIZE
329 hash_get_entries (const Hash_table *table, void **buffer,
333 struct hash_entry const *bucket;
334 struct hash_entry const *cursor;
336 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
340 for (cursor = bucket; cursor; cursor = cursor->next)
342 if (counter >= buffer_size)
344 buffer[counter++] = cursor->data;
352 /* Call a PROCESSOR function for each entry of a hash table, and return the
353 number of entries for which the processor function returned success. A
354 pointer to some PROCESSOR_DATA which will be made available to each call to
355 the processor function. The PROCESSOR accepts two arguments: the first is
356 the user entry being walked into, the second is the value of PROCESSOR_DATA
357 as received. The walking continue for as long as the PROCESSOR function
358 returns nonzero. When it returns zero, the walking is interrupted. */
361 hash_do_for_each (const Hash_table *table, Hash_processor processor,
362 void *processor_data)
365 struct hash_entry const *bucket;
366 struct hash_entry const *cursor;
368 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
372 for (cursor = bucket; cursor; cursor = cursor->next)
374 if (! processor (cursor->data, processor_data))
384 /* Allocation and clean-up. */
386 /* Return a hash index for a NUL-terminated STRING between 0 and N_BUCKETS-1.
387 This is a convenience routine for constructing other hashing functions. */
391 /* About hashings, Paul Eggert writes to me (FP), on 1994-01-01: "Please see
392 B. J. McKenzie, R. Harries & T. Bell, Selecting a hashing algorithm,
393 Software--practice & experience 20, 2 (Feb 1990), 209-224. Good hash
394 algorithms tend to be domain-specific, so what's good for [diffutils'] io.c
395 may not be good for your application." */
398 hash_string (const char *string, size_t n_buckets)
400 # define HASH_ONE_CHAR(Value, Byte) \
401 ((Byte) + rotl_sz (Value, 7))
406 for (; (ch = *string); string++)
407 value = HASH_ONE_CHAR (value, ch);
408 return value % n_buckets;
410 # undef HASH_ONE_CHAR
413 #else /* not USE_DIFF_HASH */
415 /* This one comes from `recode', and performs a bit better than the above as
416 per a few experiments. It is inspired from a hashing routine found in the
417 very old Cyber `snoop', itself written in typical Greg Mansfield style.
418 (By the way, what happened to this excellent man? Is he still alive?) */
421 hash_string (const char *string, size_t n_buckets)
426 for (; (ch = *string); string++)
427 value = (value * 31 + ch) % n_buckets;
431 #endif /* not USE_DIFF_HASH */
433 /* Return true if CANDIDATE is a prime number. CANDIDATE should be an odd
434 number at least equal to 11. */
437 is_prime (size_t candidate)
440 size_t square = divisor * divisor;
442 while (square < candidate && (candidate % divisor))
445 square += 4 * divisor;
449 return (candidate % divisor ? true : false);
452 /* Round a given CANDIDATE number up to the nearest prime, and return that
453 prime. Primes lower than 10 are merely skipped. */
456 next_prime (size_t candidate)
458 /* Skip small primes. */
462 /* Make it definitely odd. */
465 while (SIZE_MAX != candidate && !is_prime (candidate))
472 hash_reset_tuning (Hash_tuning *tuning)
474 *tuning = default_tuning;
477 /* If the user passes a NULL hasher, we hash the raw pointer. */
479 raw_hasher (const void *data, size_t n)
481 /* When hashing unique pointers, it is often the case that they were
482 generated by malloc and thus have the property that the low-order
483 bits are 0. As this tends to give poorer performance with small
484 tables, we rotate the pointer value before performing division,
485 in an attempt to improve hash quality. */
486 size_t val = rotr_sz ((size_t) data, 3);
490 /* If the user passes a NULL comparator, we use pointer comparison. */
492 raw_comparator (const void *a, const void *b)
498 /* For the given hash TABLE, check the user supplied tuning structure for
499 reasonable values, and return true if there is no gross error with it.
500 Otherwise, definitively reset the TUNING field to some acceptable default
501 in the hash table (that is, the user loses the right of further modifying
502 tuning arguments), and return false. */
505 check_tuning (Hash_table *table)
507 const Hash_tuning *tuning = table->tuning;
509 if (tuning == &default_tuning)
512 /* Be a bit stricter than mathematics would require, so that
513 rounding errors in size calculations do not cause allocations to
514 fail to grow or shrink as they should. The smallest allocation
515 is 11 (due to next_prime's algorithm), so an epsilon of 0.1
516 should be good enough. */
519 if (epsilon < tuning->growth_threshold
520 && tuning->growth_threshold < 1 - epsilon
521 && 1 + epsilon < tuning->growth_factor
522 && 0 <= tuning->shrink_threshold
523 && tuning->shrink_threshold + epsilon < tuning->shrink_factor
524 && tuning->shrink_factor <= 1
525 && tuning->shrink_threshold + epsilon < tuning->growth_threshold)
528 table->tuning = &default_tuning;
532 /* Compute the size of the bucket array for the given CANDIDATE and
533 TUNING, or return 0 if there is no possible way to allocate that
537 compute_bucket_size (size_t candidate, const Hash_tuning *tuning)
539 if (!tuning->is_n_buckets)
541 float new_candidate = candidate / tuning->growth_threshold;
542 if (SIZE_MAX <= new_candidate)
544 candidate = new_candidate;
546 candidate = next_prime (candidate);
547 if (xalloc_oversized (candidate, sizeof (struct hash_entry *)))
552 /* Allocate and return a new hash table, or NULL upon failure. The initial
553 number of buckets is automatically selected so as to _guarantee_ that you
554 may insert at least CANDIDATE different user entries before any growth of
555 the hash table size occurs. So, if have a reasonably tight a-priori upper
556 bound on the number of entries you intend to insert in the hash table, you
557 may save some table memory and insertion time, by specifying it here. If
558 the IS_N_BUCKETS field of the TUNING structure is true, the CANDIDATE
559 argument has its meaning changed to the wanted number of buckets.
561 TUNING points to a structure of user-supplied values, in case some fine
562 tuning is wanted over the default behavior of the hasher. If TUNING is
563 NULL, the default tuning parameters are used instead. If TUNING is
564 provided but the values requested are out of bounds or might cause
565 rounding errors, return NULL.
567 The user-supplied HASHER function, when not NULL, accepts two
568 arguments ENTRY and TABLE_SIZE. It computes, by hashing ENTRY contents, a
569 slot number for that entry which should be in the range 0..TABLE_SIZE-1.
570 This slot number is then returned.
572 The user-supplied COMPARATOR function, when not NULL, accepts two
573 arguments pointing to user data, it then returns true for a pair of entries
574 that compare equal, or false otherwise. This function is internally called
575 on entries which are already known to hash to the same bucket index,
576 but which are distinct pointers.
578 The user-supplied DATA_FREER function, when not NULL, may be later called
579 with the user data as an argument, just before the entry containing the
580 data gets freed. This happens from within `hash_free' or `hash_clear'.
581 You should specify this function only if you want these functions to free
582 all of your `data' data. This is typically the case when your data is
583 simply an auxiliary struct that you have malloc'd to aggregate several
587 hash_initialize (size_t candidate, const Hash_tuning *tuning,
588 Hash_hasher hasher, Hash_comparator comparator,
589 Hash_data_freer data_freer)
595 if (comparator == NULL)
596 comparator = raw_comparator;
598 table = malloc (sizeof *table);
603 tuning = &default_tuning;
604 table->tuning = tuning;
605 if (!check_tuning (table))
607 /* Fail if the tuning options are invalid. This is the only occasion
608 when the user gets some feedback about it. Once the table is created,
609 if the user provides invalid tuning options, we silently revert to
610 using the defaults, and ignore further request to change the tuning
615 table->n_buckets = compute_bucket_size (candidate, tuning);
616 if (!table->n_buckets)
619 table->bucket = calloc (table->n_buckets, sizeof *table->bucket);
620 if (table->bucket == NULL)
622 table->bucket_limit = table->bucket + table->n_buckets;
623 table->n_buckets_used = 0;
624 table->n_entries = 0;
626 table->hasher = hasher;
627 table->comparator = comparator;
628 table->data_freer = data_freer;
630 table->free_entry_list = NULL;
632 obstack_init (&table->entry_stack);
641 /* Make all buckets empty, placing any chained entries on the free list.
642 Apply the user-specified function data_freer (if any) to the datas of any
646 hash_clear (Hash_table *table)
648 struct hash_entry *bucket;
650 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
654 struct hash_entry *cursor;
655 struct hash_entry *next;
657 /* Free the bucket overflow. */
658 for (cursor = bucket->next; cursor; cursor = next)
660 if (table->data_freer)
661 table->data_freer (cursor->data);
665 /* Relinking is done one entry at a time, as it is to be expected
666 that overflows are either rare or short. */
667 cursor->next = table->free_entry_list;
668 table->free_entry_list = cursor;
671 /* Free the bucket head. */
672 if (table->data_freer)
673 table->data_freer (bucket->data);
679 table->n_buckets_used = 0;
680 table->n_entries = 0;
683 /* Reclaim all storage associated with a hash table. If a data_freer
684 function has been supplied by the user when the hash table was created,
685 this function applies it to the data of each entry before freeing that
689 hash_free (Hash_table *table)
691 struct hash_entry *bucket;
692 struct hash_entry *cursor;
693 struct hash_entry *next;
695 /* Call the user data_freer function. */
696 if (table->data_freer && table->n_entries)
698 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
702 for (cursor = bucket; cursor; cursor = cursor->next)
703 table->data_freer (cursor->data);
710 obstack_free (&table->entry_stack, NULL);
714 /* Free all bucket overflowed entries. */
715 for (bucket = table->bucket; bucket < table->bucket_limit; bucket++)
717 for (cursor = bucket->next; cursor; cursor = next)
724 /* Also reclaim the internal list of previously freed entries. */
725 for (cursor = table->free_entry_list; cursor; cursor = next)
733 /* Free the remainder of the hash table structure. */
734 free (table->bucket);
738 /* Insertion and deletion. */
740 /* Get a new hash entry for a bucket overflow, possibly by recycling a
741 previously freed one. If this is not possible, allocate a new one. */
743 static struct hash_entry *
744 allocate_entry (Hash_table *table)
746 struct hash_entry *new;
748 if (table->free_entry_list)
750 new = table->free_entry_list;
751 table->free_entry_list = new->next;
756 new = obstack_alloc (&table->entry_stack, sizeof *new);
758 new = malloc (sizeof *new);
765 /* Free a hash entry which was part of some bucket overflow,
766 saving it for later recycling. */
769 free_entry (Hash_table *table, struct hash_entry *entry)
772 entry->next = table->free_entry_list;
773 table->free_entry_list = entry;
776 /* This private function is used to help with insertion and deletion. When
777 ENTRY matches an entry in the table, return a pointer to the corresponding
778 user data and set *BUCKET_HEAD to the head of the selected bucket.
779 Otherwise, return NULL. When DELETE is true and ENTRY matches an entry in
780 the table, unlink the matching entry. */
783 hash_find_entry (Hash_table *table, const void *entry,
784 struct hash_entry **bucket_head, bool delete)
786 struct hash_entry *bucket
787 = table->bucket + table->hasher (entry, table->n_buckets);
788 struct hash_entry *cursor;
790 if (! (bucket < table->bucket_limit))
793 *bucket_head = bucket;
795 /* Test for empty bucket. */
796 if (bucket->data == NULL)
799 /* See if the entry is the first in the bucket. */
800 if (entry == bucket->data || table->comparator (entry, bucket->data))
802 void *data = bucket->data;
808 struct hash_entry *next = bucket->next;
810 /* Bump the first overflow entry into the bucket head, then save
811 the previous first overflow entry for later recycling. */
813 free_entry (table, next);
824 /* Scan the bucket overflow. */
825 for (cursor = bucket; cursor->next; cursor = cursor->next)
827 if (entry == cursor->next->data
828 || table->comparator (entry, cursor->next->data))
830 void *data = cursor->next->data;
834 struct hash_entry *next = cursor->next;
836 /* Unlink the entry to delete, then save the freed entry for later
838 cursor->next = next->next;
839 free_entry (table, next);
846 /* No entry found. */
850 /* Internal helper, to move entries from SRC to DST. Both tables must
851 share the same free entry list. If SAFE, only move overflow
852 entries, saving bucket heads for later, so that no allocations will
853 occur. Return false if the free entry list is exhausted and an
857 transfer_entries (Hash_table *dst, Hash_table *src, bool safe)
859 struct hash_entry *bucket;
860 struct hash_entry *cursor;
861 struct hash_entry *next;
862 for (bucket = src->bucket; bucket < src->bucket_limit; bucket++)
866 struct hash_entry *new_bucket;
868 /* Within each bucket, transfer overflow entries first and
869 then the bucket head, to minimize memory pressure. After
870 all, the only time we might allocate is when moving the
871 bucket head, but moving overflow entries first may create
872 free entries that can be recycled by the time we finally
873 get to the bucket head. */
874 for (cursor = bucket->next; cursor; cursor = next)
877 new_bucket = (dst->bucket + dst->hasher (data, dst->n_buckets));
879 if (! (new_bucket < dst->bucket_limit))
884 if (new_bucket->data)
886 /* Merely relink an existing entry, when moving from a
887 bucket overflow into a bucket overflow. */
888 cursor->next = new_bucket->next;
889 new_bucket->next = cursor;
893 /* Free an existing entry, when moving from a bucket
894 overflow into a bucket header. */
895 new_bucket->data = data;
896 dst->n_buckets_used++;
897 free_entry (dst, cursor);
900 /* Now move the bucket head. Be sure that if we fail due to
901 allocation failure that the src table is in a consistent
907 new_bucket = (dst->bucket + dst->hasher (data, dst->n_buckets));
909 if (! (new_bucket < dst->bucket_limit))
912 if (new_bucket->data)
914 /* Allocate or recycle an entry, when moving from a bucket
915 header into a bucket overflow. */
916 struct hash_entry *new_entry = allocate_entry (dst);
918 if (new_entry == NULL)
921 new_entry->data = data;
922 new_entry->next = new_bucket->next;
923 new_bucket->next = new_entry;
927 /* Move from one bucket header to another. */
928 new_bucket->data = data;
929 dst->n_buckets_used++;
932 src->n_buckets_used--;
937 /* For an already existing hash table, change the number of buckets through
938 specifying CANDIDATE. The contents of the hash table are preserved. The
939 new number of buckets is automatically selected so as to _guarantee_ that
940 the table may receive at least CANDIDATE different user entries, including
941 those already in the table, before any other growth of the hash table size
942 occurs. If TUNING->IS_N_BUCKETS is true, then CANDIDATE specifies the
943 exact number of buckets desired. Return true iff the rehash succeeded. */
946 hash_rehash (Hash_table *table, size_t candidate)
949 Hash_table *new_table;
950 size_t new_size = compute_bucket_size (candidate, table->tuning);
954 if (new_size == table->n_buckets)
956 new_table = &storage;
957 new_table->bucket = calloc (new_size, sizeof *new_table->bucket);
958 if (new_table->bucket == NULL)
960 new_table->n_buckets = new_size;
961 new_table->bucket_limit = new_table->bucket + new_size;
962 new_table->n_buckets_used = 0;
963 new_table->n_entries = 0;
964 new_table->tuning = table->tuning;
965 new_table->hasher = table->hasher;
966 new_table->comparator = table->comparator;
967 new_table->data_freer = table->data_freer;
969 /* In order for the transfer to successfully complete, we need
970 additional overflow entries when distinct buckets in the old
971 table collide into a common bucket in the new table. The worst
972 case possible is a hasher that gives a good spread with the old
973 size, but returns a constant with the new size; if we were to
974 guarantee table->n_buckets_used-1 free entries in advance, then
975 the transfer would be guaranteed to not allocate memory.
976 However, for large tables, a guarantee of no further allocation
977 introduces a lot of extra memory pressure, all for an unlikely
978 corner case (most rehashes reduce, rather than increase, the
979 number of overflow entries needed). So, we instead ensure that
980 the transfer process can be reversed if we hit a memory
981 allocation failure mid-transfer. */
983 /* Merely reuse the extra old space into the new table. */
985 new_table->entry_stack = table->entry_stack;
987 new_table->free_entry_list = table->free_entry_list;
989 if (transfer_entries (new_table, table, false))
991 /* Entries transferred successfully; tie up the loose ends. */
992 free (table->bucket);
993 table->bucket = new_table->bucket;
994 table->bucket_limit = new_table->bucket_limit;
995 table->n_buckets = new_table->n_buckets;
996 table->n_buckets_used = new_table->n_buckets_used;
997 table->free_entry_list = new_table->free_entry_list;
998 /* table->n_entries and table->entry_stack already hold their value. */
1002 /* We've allocated new_table->bucket (and possibly some entries),
1003 exhausted the free list, and moved some but not all entries into
1004 new_table. We must undo the partial move before returning
1005 failure. The only way to get into this situation is if new_table
1006 uses fewer buckets than the old table, so we will reclaim some
1007 free entries as overflows in the new table are put back into
1008 distinct buckets in the old table.
1010 There are some pathological cases where a single pass through the
1011 table requires more intermediate overflow entries than using two
1012 passes. Two passes give worse cache performance and takes
1013 longer, but at this point, we're already out of memory, so slow
1014 and safe is better than failure. */
1015 table->free_entry_list = new_table->free_entry_list;
1016 if (! (transfer_entries (table, new_table, true)
1017 && transfer_entries (table, new_table, false)))
1019 /* table->n_entries already holds its value. */
1020 free (new_table->bucket);
1024 /* If ENTRY matches an entry already in the hash table, return the pointer
1025 to the entry from the table. Otherwise, insert ENTRY and return ENTRY.
1026 Return NULL if the storage required for insertion cannot be allocated.
1027 This implementation does not support duplicate entries or insertion of
1031 hash_insert (Hash_table *table, const void *entry)
1034 struct hash_entry *bucket;
1036 /* The caller cannot insert a NULL entry. */
1040 /* If there's a matching entry already in the table, return that. */
1041 if ((data = hash_find_entry (table, entry, &bucket, false)) != NULL)
1044 /* If the growth threshold of the buckets in use has been reached, increase
1045 the table size and rehash. There's no point in checking the number of
1046 entries: if the hashing function is ill-conditioned, rehashing is not
1047 likely to improve it. */
1049 if (table->n_buckets_used
1050 > table->tuning->growth_threshold * table->n_buckets)
1052 /* Check more fully, before starting real work. If tuning arguments
1053 became invalid, the second check will rely on proper defaults. */
1054 check_tuning (table);
1055 if (table->n_buckets_used
1056 > table->tuning->growth_threshold * table->n_buckets)
1058 const Hash_tuning *tuning = table->tuning;
1060 (tuning->is_n_buckets
1061 ? (table->n_buckets * tuning->growth_factor)
1062 : (table->n_buckets * tuning->growth_factor
1063 * tuning->growth_threshold));
1065 if (SIZE_MAX <= candidate)
1068 /* If the rehash fails, arrange to return NULL. */
1069 if (!hash_rehash (table, candidate))
1072 /* Update the bucket we are interested in. */
1073 if (hash_find_entry (table, entry, &bucket, false) != NULL)
1078 /* ENTRY is not matched, it should be inserted. */
1082 struct hash_entry *new_entry = allocate_entry (table);
1084 if (new_entry == NULL)
1087 /* Add ENTRY in the overflow of the bucket. */
1089 new_entry->data = (void *) entry;
1090 new_entry->next = bucket->next;
1091 bucket->next = new_entry;
1093 return (void *) entry;
1096 /* Add ENTRY right in the bucket head. */
1098 bucket->data = (void *) entry;
1100 table->n_buckets_used++;
1102 return (void *) entry;
1105 /* If ENTRY is already in the table, remove it and return the just-deleted
1106 data (the user may want to deallocate its storage). If ENTRY is not in the
1107 table, don't modify the table and return NULL. */
1110 hash_delete (Hash_table *table, const void *entry)
1113 struct hash_entry *bucket;
1115 data = hash_find_entry (table, entry, &bucket, true);
1122 table->n_buckets_used--;
1124 /* If the shrink threshold of the buckets in use has been reached,
1125 rehash into a smaller table. */
1127 if (table->n_buckets_used
1128 < table->tuning->shrink_threshold * table->n_buckets)
1130 /* Check more fully, before starting real work. If tuning arguments
1131 became invalid, the second check will rely on proper defaults. */
1132 check_tuning (table);
1133 if (table->n_buckets_used
1134 < table->tuning->shrink_threshold * table->n_buckets)
1136 const Hash_tuning *tuning = table->tuning;
1138 (tuning->is_n_buckets
1139 ? table->n_buckets * tuning->shrink_factor
1140 : (table->n_buckets * tuning->shrink_factor
1141 * tuning->growth_threshold));
1143 if (!hash_rehash (table, candidate))
1145 /* Failure to allocate memory in an attempt to
1146 shrink the table is not fatal. But since memory
1147 is low, we can at least be kind and free any
1148 spare entries, rather than keeping them tied up
1149 in the free entry list. */
1151 struct hash_entry *cursor = table->free_entry_list;
1152 struct hash_entry *next;
1155 next = cursor->next;
1159 table->free_entry_list = NULL;
1174 hash_print (const Hash_table *table)
1176 struct hash_entry *bucket = (struct hash_entry *) table->bucket;
1178 for ( ; bucket < table->bucket_limit; bucket++)
1180 struct hash_entry *cursor;
1183 printf ("%lu:\n", (unsigned long int) (bucket - table->bucket));
1185 for (cursor = bucket; cursor; cursor = cursor->next)
1187 char const *s = cursor->data;
1190 printf (" %s\n", s);
1195 #endif /* TESTING */