]>
git.proxmox.com Git - libgit2.git/blob - src/khash.h
c9b7f131f0a10a708807f07c8daf1cea420bcfb3
3 Copyright (c) 2008, 2009, 2011 by Attractive Chaos <attractor@live.co.uk>
5 Permission is hereby granted, free of charge, to any person obtaining
6 a copy of this software and associated documentation files (the
7 "Software"), to deal in the Software without restriction, including
8 without limitation the rights to use, copy, modify, merge, publish,
9 distribute, sublicense, and/or sell copies of the Software, and to
10 permit persons to whom the Software is furnished to do so, subject to
11 the following conditions:
13 The above copyright notice and this permission notice shall be
14 included in all copies or substantial portions of the Software.
16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
17 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
19 NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
20 BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
21 ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30 KHASH_MAP_INIT_INT(32, char)
34 khash_t(32) *h = kh_init(32);
35 k = kh_put(32, h, 5, &ret);
37 k = kh_get(32, h, 10);
38 is_missing = (k == kh_end(h));
41 for (k = kh_begin(h); k != kh_end(h); ++k)
42 if (kh_exist(h, k)) kh_value(h, k) = 1;
51 * Use quadratic probing. When the capacity is power of 2, stepping function
52 i*(i+1)/2 guarantees to traverse each bucket. It is better than double
53 hashing on cache performance and is more robust than linear probing.
55 In theory, double hashing should be more robust than quadratic probing.
56 However, my implementation is probably not for large hash tables, because
57 the second hash function is closely tied to the first hash function,
58 which reduce the effectiveness of double hashing.
60 Reference: http://research.cs.vt.edu/AVresearch/hashing/quadratic.php
64 * Minor code clean up; no actual effect.
68 * The capacity is a power of 2. This seems to dramatically improve the
69 speed for simple keys. Thank Zilong Tan for the suggestion. Reference:
71 - http://code.google.com/p/ulib/
72 - http://nothings.org/computer/judy/
74 * Allow to optionally use linear probing which usually has better
75 performance for random input. Double hashing is still the default as it
76 is more robust to certain non-random input.
78 * Added Wang's integer hash function (not used by default). This hash
79 function is more robust to certain non-random input.
83 * Allow to declare global functions.
91 * Corrected the example
96 * Improved speed a little in kh_put()
101 * Fixed a compiling error
105 * Changed to token concatenation which increases flexibility.
109 * Fixed a bug in kh_get(), which has not been tested previously.
123 Generic hash table library.
126 #define AC_VERSION_KHASH_H "0.2.8"
132 /* compiler specific configuration */
134 typedef uint32_t khint32_t
;
135 typedef uint64_t khint64_t
;
139 #define kh_inline __inline
140 #elif defined(__GNUC__)
141 #define kh_inline __inline__
145 #endif /* kh_inline */
147 typedef khint32_t khint_t
;
148 typedef khint_t khiter_t
;
150 #define __ac_isempty(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&2)
151 #define __ac_isdel(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&1)
152 #define __ac_iseither(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&3)
153 #define __ac_set_isdel_false(flag, i) (flag[i>>4]&=~(1ul<<((i&0xfU)<<1)))
154 #define __ac_set_isempty_false(flag, i) (flag[i>>4]&=~(2ul<<((i&0xfU)<<1)))
155 #define __ac_set_isboth_false(flag, i) (flag[i>>4]&=~(3ul<<((i&0xfU)<<1)))
156 #define __ac_set_isdel_true(flag, i) (flag[i>>4]|=1ul<<((i&0xfU)<<1))
158 #define __ac_fsize(m) ((m) < 16? 1 : (m)>>4)
161 #define kroundup32(x) (--(x), (x)|=(x)>>1, (x)|=(x)>>2, (x)|=(x)>>4, (x)|=(x)>>8, (x)|=(x)>>16, ++(x))
165 #define kcalloc(N,Z) calloc(N,Z)
168 #define kmalloc(Z) malloc(Z)
171 #define krealloc(P,Z) realloc(P,Z)
173 #ifndef kreallocarray
174 #define kreallocarray(P,N,Z) ((SIZE_MAX - N < Z) ? NULL : krealloc(P, (N*Z)))
177 #define kfree(P) free(P)
180 static const double __ac_HASH_UPPER
= 0.77;
182 #define __KHASH_TYPE(name, khkey_t, khval_t) \
183 typedef struct kh_##name##_s { \
184 khint_t n_buckets, size, n_occupied, upper_bound; \
190 #define __KHASH_PROTOTYPES(name, khkey_t, khval_t) \
191 extern kh_##name##_t *kh_init_##name(void); \
192 extern void kh_destroy_##name(kh_##name##_t *h); \
193 extern void kh_clear_##name(kh_##name##_t *h); \
194 extern khint_t kh_get_##name(const kh_##name##_t *h, khkey_t key); \
195 extern int kh_resize_##name(kh_##name##_t *h, khint_t new_n_buckets); \
196 extern khint_t kh_put_##name(kh_##name##_t *h, khkey_t key, int *ret); \
197 extern void kh_del_##name(kh_##name##_t *h, khint_t x);
199 #define __KHASH_IMPL(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \
200 SCOPE kh_##name##_t *kh_init_##name(void) { \
201 return (kh_##name##_t*)kcalloc(1, sizeof(kh_##name##_t)); \
203 SCOPE void kh_destroy_##name(kh_##name##_t *h) \
206 kfree((void *)h->keys); kfree(h->flags); \
207 kfree((void *)h->vals); \
211 SCOPE void kh_clear_##name(kh_##name##_t *h) \
213 if (h && h->flags) { \
214 memset(h->flags, 0xaa, __ac_fsize(h->n_buckets) * sizeof(khint32_t)); \
215 h->size = h->n_occupied = 0; \
218 SCOPE khint_t kh_get_##name(const kh_##name##_t *h, khkey_t key) \
220 if (h->n_buckets) { \
221 khint_t k, i, last, mask, step = 0; \
222 mask = h->n_buckets - 1; \
223 k = __hash_func(key); i = k & mask; \
225 while (!__ac_isempty(h->flags, i) && (__ac_isdel(h->flags, i) || !__hash_equal(h->keys[i], key))) { \
226 i = (i + (++step)) & mask; \
227 if (i == last) return h->n_buckets; \
229 return __ac_iseither(h->flags, i)? h->n_buckets : i; \
232 SCOPE int kh_resize_##name(kh_##name##_t *h, khint_t new_n_buckets) \
233 { /* This function uses 0.25*n_buckets bytes of working space instead of [sizeof(key_t+val_t)+.25]*n_buckets. */ \
234 khint32_t *new_flags = 0; \
237 kroundup32(new_n_buckets); \
238 if (new_n_buckets < 4) new_n_buckets = 4; \
239 if (h->size >= (khint_t)(new_n_buckets * __ac_HASH_UPPER + 0.5)) j = 0; /* requested size is too small */ \
240 else { /* hash table size to be changed (shrink or expand); rehash */ \
241 new_flags = (khint32_t*)kreallocarray(NULL, __ac_fsize(new_n_buckets), sizeof(khint32_t)); \
242 if (!new_flags) return -1; \
243 memset(new_flags, 0xaa, __ac_fsize(new_n_buckets) * sizeof(khint32_t)); \
244 if (h->n_buckets < new_n_buckets) { /* expand */ \
245 khkey_t *new_keys = (khkey_t*)kreallocarray((void *)h->keys, new_n_buckets, sizeof(khkey_t)); \
246 if (!new_keys) { kfree(new_flags); return -1; } \
247 h->keys = new_keys; \
249 khval_t *new_vals = (khval_t*)kreallocarray((void *)h->vals, new_n_buckets, sizeof(khval_t)); \
250 if (!new_vals) { kfree(new_flags); return -1; } \
251 h->vals = new_vals; \
253 } /* otherwise shrink */ \
256 if (j) { /* rehashing is needed */ \
257 for (j = 0; j != h->n_buckets; ++j) { \
258 if (__ac_iseither(h->flags, j) == 0) { \
259 khkey_t key = h->keys[j]; \
262 new_mask = new_n_buckets - 1; \
263 if (kh_is_map) val = h->vals[j]; \
264 __ac_set_isdel_true(h->flags, j); \
265 while (1) { /* kick-out process; sort of like in Cuckoo hashing */ \
266 khint_t k, i, step = 0; \
267 k = __hash_func(key); \
269 while (!__ac_isempty(new_flags, i)) i = (i + (++step)) & new_mask; \
270 __ac_set_isempty_false(new_flags, i); \
271 if (i < h->n_buckets && __ac_iseither(h->flags, i) == 0) { /* kick out the existing element */ \
272 { khkey_t tmp = h->keys[i]; h->keys[i] = key; key = tmp; } \
273 if (kh_is_map) { khval_t tmp = h->vals[i]; h->vals[i] = val; val = tmp; } \
274 __ac_set_isdel_true(h->flags, i); /* mark it as deleted in the old hash table */ \
275 } else { /* write the element and jump out of the loop */ \
277 if (kh_is_map) h->vals[i] = val; \
283 if (h->n_buckets > new_n_buckets) { /* shrink the hash table */ \
284 h->keys = (khkey_t*)kreallocarray((void *)h->keys, new_n_buckets, sizeof(khkey_t)); \
285 if (kh_is_map) h->vals = (khval_t*)kreallocarray((void *)h->vals, new_n_buckets, sizeof(khval_t)); \
287 kfree(h->flags); /* free the working space */ \
288 h->flags = new_flags; \
289 h->n_buckets = new_n_buckets; \
290 h->n_occupied = h->size; \
291 h->upper_bound = (khint_t)(h->n_buckets * __ac_HASH_UPPER + 0.5); \
295 SCOPE khint_t kh_put_##name(kh_##name##_t *h, khkey_t key, int *ret) \
298 if (h->n_occupied >= h->upper_bound) { /* update the hash table */ \
299 if (h->n_buckets > (h->size<<1)) { \
300 if (kh_resize_##name(h, h->n_buckets - 1) < 0) { /* clear "deleted" elements */ \
301 *ret = -1; return h->n_buckets; \
303 } else if (kh_resize_##name(h, h->n_buckets + 1) < 0) { /* expand the hash table */ \
304 *ret = -1; return h->n_buckets; \
306 } /* TODO: to implement automatically shrinking; resize() already support shrinking */ \
308 khint_t k, i, site, last, mask = h->n_buckets - 1, step = 0; \
309 x = site = h->n_buckets; k = __hash_func(key); i = k & mask; \
310 if (__ac_isempty(h->flags, i)) x = i; /* for speed up */ \
313 while (!__ac_isempty(h->flags, i) && (__ac_isdel(h->flags, i) || !__hash_equal(h->keys[i], key))) { \
314 if (__ac_isdel(h->flags, i)) site = i; \
315 i = (i + (++step)) & mask; \
316 if (i == last) { x = site; break; } \
318 if (x == h->n_buckets) { \
319 if (__ac_isempty(h->flags, i) && site != h->n_buckets) x = site; \
324 if (__ac_isempty(h->flags, x)) { /* not present at all */ \
326 __ac_set_isboth_false(h->flags, x); \
327 ++h->size; ++h->n_occupied; \
329 } else if (__ac_isdel(h->flags, x)) { /* deleted */ \
331 __ac_set_isboth_false(h->flags, x); \
334 } else *ret = 0; /* Don't touch h->keys[x] if present and not deleted */ \
337 SCOPE void kh_del_##name(kh_##name##_t *h, khint_t x) \
339 if (x != h->n_buckets && !__ac_iseither(h->flags, x)) { \
340 __ac_set_isdel_true(h->flags, x); \
345 #define KHASH_DECLARE(name, khkey_t, khval_t) \
346 __KHASH_TYPE(name, khkey_t, khval_t) \
347 __KHASH_PROTOTYPES(name, khkey_t, khval_t)
349 #define KHASH_INIT2(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \
350 __KHASH_TYPE(name, khkey_t, khval_t) \
351 __KHASH_IMPL(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal)
353 #define KHASH_INIT(name, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \
354 KHASH_INIT2(name, static kh_inline, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal)
356 /* --- BEGIN OF HASH FUNCTIONS --- */
359 @abstract Integer hash function
360 @param key The integer [khint32_t]
361 @return The hash value [khint_t]
363 #define kh_int_hash_func(key) (khint32_t)(key)
365 @abstract Integer comparison function
367 #define kh_int_hash_equal(a, b) ((a) == (b))
369 @abstract 64-bit integer hash function
370 @param key The integer [khint64_t]
371 @return The hash value [khint_t]
373 #define kh_int64_hash_func(key) (khint32_t)((key)>>33^(key)^(key)<<11)
375 @abstract 64-bit integer comparison function
377 #define kh_int64_hash_equal(a, b) ((a) == (b))
379 @abstract const char* hash function
380 @param s Pointer to a null terminated string
381 @return The hash value
383 static kh_inline khint_t
__ac_X31_hash_string(const char *s
)
385 khint_t h
= (khint_t
)*s
;
386 if (h
) for (++s
; *s
; ++s
) h
= (h
<< 5) - h
+ (khint_t
)*s
;
390 @abstract Another interface to const char* hash function
391 @param key Pointer to a null terminated string [const char*]
392 @return The hash value [khint_t]
394 #define kh_str_hash_func(key) __ac_X31_hash_string(key)
396 @abstract Const char* comparison function
398 #define kh_str_hash_equal(a, b) (strcmp(a, b) == 0)
400 static kh_inline khint_t
__ac_Wang_hash(khint_t key
)
410 #define kh_int_hash_func2(k) __ac_Wang_hash((khint_t)key)
412 /* --- END OF HASH FUNCTIONS --- */
414 /* Other convenient macros... */
417 @abstract Type of the hash table.
418 @param name Name of the hash table [symbol]
420 #define khash_t(name) kh_##name##_t
423 @abstract Initiate a hash table.
424 @param name Name of the hash table [symbol]
425 @return Pointer to the hash table [khash_t(name)*]
427 #define kh_init(name) kh_init_##name()
430 @abstract Destroy a hash table.
431 @param name Name of the hash table [symbol]
432 @param h Pointer to the hash table [khash_t(name)*]
434 #define kh_destroy(name, h) kh_destroy_##name(h)
437 @abstract Reset a hash table without deallocating memory.
438 @param name Name of the hash table [symbol]
439 @param h Pointer to the hash table [khash_t(name)*]
441 #define kh_clear(name, h) kh_clear_##name(h)
444 @abstract Resize a hash table.
445 @param name Name of the hash table [symbol]
446 @param h Pointer to the hash table [khash_t(name)*]
447 @param s New size [khint_t]
449 #define kh_resize(name, h, s) kh_resize_##name(h, s)
452 @abstract Insert a key to the hash table.
453 @param name Name of the hash table [symbol]
454 @param h Pointer to the hash table [khash_t(name)*]
455 @param k Key [type of keys]
456 @param r Extra return code: -1 if the operation failed;
457 0 if the key is present in the hash table;
458 1 if the bucket is empty (never used); 2 if the element in
459 the bucket has been deleted [int*]
460 @return Iterator to the inserted element [khint_t]
462 #define kh_put(name, h, k, r) kh_put_##name(h, k, r)
465 @abstract Retrieve a key from the hash table.
466 @param name Name of the hash table [symbol]
467 @param h Pointer to the hash table [khash_t(name)*]
468 @param k Key [type of keys]
469 @return Iterator to the found element, or kh_end(h) if the element is absent [khint_t]
471 #define kh_get(name, h, k) kh_get_##name(h, k)
474 @abstract Remove a key from the hash table.
475 @param name Name of the hash table [symbol]
476 @param h Pointer to the hash table [khash_t(name)*]
477 @param k Iterator to the element to be deleted [khint_t]
479 #define kh_del(name, h, k) kh_del_##name(h, k)
482 @abstract Test whether a bucket contains data.
483 @param h Pointer to the hash table [khash_t(name)*]
484 @param x Iterator to the bucket [khint_t]
485 @return 1 if containing data; 0 otherwise [int]
487 #define kh_exist(h, x) (!__ac_iseither((h)->flags, (x)))
490 @abstract Get key given an iterator
491 @param h Pointer to the hash table [khash_t(name)*]
492 @param x Iterator to the bucket [khint_t]
493 @return Key [type of keys]
495 #define kh_key(h, x) ((h)->keys[x])
498 @abstract Get value given an iterator
499 @param h Pointer to the hash table [khash_t(name)*]
500 @param x Iterator to the bucket [khint_t]
501 @return Value [type of values]
502 @discussion For hash sets, calling this results in segfault.
504 #define kh_val(h, x) ((h)->vals[x])
507 @abstract Alias of kh_val()
509 #define kh_value(h, x) ((h)->vals[x])
512 @abstract Get the start iterator
513 @param h Pointer to the hash table [khash_t(name)*]
514 @return The start iterator [khint_t]
516 #define kh_begin(h) (khint_t)(0)
519 @abstract Get the end iterator
520 @param h Pointer to the hash table [khash_t(name)*]
521 @return The end iterator [khint_t]
523 #define kh_end(h) ((h)->n_buckets)
526 @abstract Get the number of elements in the hash table
527 @param h Pointer to the hash table [khash_t(name)*]
528 @return Number of elements in the hash table [khint_t]
530 #define kh_size(h) ((h)->size)
533 @abstract Get the number of buckets in the hash table
534 @param h Pointer to the hash table [khash_t(name)*]
535 @return Number of buckets in the hash table [khint_t]
537 #define kh_n_buckets(h) ((h)->n_buckets)
540 @abstract Iterate over the entries in the hash table
541 @param h Pointer to the hash table [khash_t(name)*]
542 @param kvar Variable to which key will be assigned
543 @param vvar Variable to which value will be assigned
544 @param code Block of code to execute
546 #define kh_foreach(h, kvar, vvar, code) { khint_t __i; \
547 for (__i = kh_begin(h); __i != kh_end(h); ++__i) { \
548 if (!kh_exist(h,__i)) continue; \
549 (kvar) = kh_key(h,__i); \
550 (vvar) = kh_val(h,__i); \
555 @abstract Iterate over the values in the hash table
556 @param h Pointer to the hash table [khash_t(name)*]
557 @param vvar Variable to which value will be assigned
558 @param code Block of code to execute
560 #define kh_foreach_value(h, vvar, code) { khint_t __i; \
561 for (__i = kh_begin(h); __i != kh_end(h); ++__i) { \
562 if (!kh_exist(h,__i)) continue; \
563 (vvar) = kh_val(h,__i); \
567 /* More convenient interfaces */
570 @abstract Instantiate a hash set containing integer keys
571 @param name Name of the hash table [symbol]
573 #define KHASH_SET_INIT_INT(name) \
574 KHASH_INIT(name, khint32_t, char, 0, kh_int_hash_func, kh_int_hash_equal)
577 @abstract Instantiate a hash map containing integer keys
578 @param name Name of the hash table [symbol]
579 @param khval_t Type of values [type]
581 #define KHASH_MAP_INIT_INT(name, khval_t) \
582 KHASH_INIT(name, khint32_t, khval_t, 1, kh_int_hash_func, kh_int_hash_equal)
585 @abstract Instantiate a hash map containing 64-bit integer keys
586 @param name Name of the hash table [symbol]
588 #define KHASH_SET_INIT_INT64(name) \
589 KHASH_INIT(name, khint64_t, char, 0, kh_int64_hash_func, kh_int64_hash_equal)
592 @abstract Instantiate a hash map containing 64-bit integer keys
593 @param name Name of the hash table [symbol]
594 @param khval_t Type of values [type]
596 #define KHASH_MAP_INIT_INT64(name, khval_t) \
597 KHASH_INIT(name, khint64_t, khval_t, 1, kh_int64_hash_func, kh_int64_hash_equal)
599 typedef const char *kh_cstr_t
;
601 @abstract Instantiate a hash map containing const char* keys
602 @param name Name of the hash table [symbol]
604 #define KHASH_SET_INIT_STR(name) \
605 KHASH_INIT(name, kh_cstr_t, char, 0, kh_str_hash_func, kh_str_hash_equal)
608 @abstract Instantiate a hash map containing const char* keys
609 @param name Name of the hash table [symbol]
610 @param khval_t Type of values [type]
612 #define KHASH_MAP_INIT_STR(name, khval_t) \
613 KHASH_INIT(name, kh_cstr_t, khval_t, 1, kh_str_hash_func, kh_str_hash_equal)
615 #endif /* __AC_KHASH_H */