9b6a12859be67a09685f13310f9769df260d558d
[git/git.git] / hashmap.c
1 /*
2 * Generic implementation of hash-based key value mappings.
3 */
4 #include "cache.h"
5 #include "hashmap.h"
6
7 #define FNV32_BASE ((unsigned int) 0x811c9dc5)
8 #define FNV32_PRIME ((unsigned int) 0x01000193)
9
10 unsigned int strhash(const char *str)
11 {
12 unsigned int c, hash = FNV32_BASE;
13 while ((c = (unsigned char) *str++))
14 hash = (hash * FNV32_PRIME) ^ c;
15 return hash;
16 }
17
18 unsigned int strihash(const char *str)
19 {
20 unsigned int c, hash = FNV32_BASE;
21 while ((c = (unsigned char) *str++)) {
22 if (c >= 'a' && c <= 'z')
23 c -= 'a' - 'A';
24 hash = (hash * FNV32_PRIME) ^ c;
25 }
26 return hash;
27 }
28
29 unsigned int memhash(const void *buf, size_t len)
30 {
31 unsigned int hash = FNV32_BASE;
32 unsigned char *ucbuf = (unsigned char *) buf;
33 while (len--) {
34 unsigned int c = *ucbuf++;
35 hash = (hash * FNV32_PRIME) ^ c;
36 }
37 return hash;
38 }
39
40 unsigned int memihash(const void *buf, size_t len)
41 {
42 unsigned int hash = FNV32_BASE;
43 unsigned char *ucbuf = (unsigned char *) buf;
44 while (len--) {
45 unsigned int c = *ucbuf++;
46 if (c >= 'a' && c <= 'z')
47 c -= 'a' - 'A';
48 hash = (hash * FNV32_PRIME) ^ c;
49 }
50 return hash;
51 }
52
53 /*
54 * Incoporate another chunk of data into a memihash
55 * computation.
56 */
57 unsigned int memihash_cont(unsigned int hash_seed, const void *buf, size_t len)
58 {
59 unsigned int hash = hash_seed;
60 unsigned char *ucbuf = (unsigned char *) buf;
61 while (len--) {
62 unsigned int c = *ucbuf++;
63 if (c >= 'a' && c <= 'z')
64 c -= 'a' - 'A';
65 hash = (hash * FNV32_PRIME) ^ c;
66 }
67 return hash;
68 }
69
70 #define HASHMAP_INITIAL_SIZE 64
71 /* grow / shrink by 2^2 */
72 #define HASHMAP_RESIZE_BITS 2
73 /* load factor in percent */
74 #define HASHMAP_LOAD_FACTOR 80
75
76 static void alloc_table(struct hashmap *map, unsigned int size)
77 {
78 map->tablesize = size;
79 map->table = xcalloc(size, sizeof(struct hashmap_entry *));
80
81 /* calculate resize thresholds for new size */
82 map->grow_at = (unsigned int) ((uint64_t) size * HASHMAP_LOAD_FACTOR / 100);
83 if (size <= HASHMAP_INITIAL_SIZE)
84 map->shrink_at = 0;
85 else
86 /*
87 * The shrink-threshold must be slightly smaller than
88 * (grow-threshold / resize-factor) to prevent erratic resizing,
89 * thus we divide by (resize-factor + 1).
90 */
91 map->shrink_at = map->grow_at / ((1 << HASHMAP_RESIZE_BITS) + 1);
92 }
93
94 static inline int entry_equals(const struct hashmap *map,
95 const struct hashmap_entry *e1, const struct hashmap_entry *e2,
96 const void *keydata)
97 {
98 return (e1 == e2) ||
99 (e1->hash == e2->hash &&
100 !map->cmpfn(map->cmpfn_data, e1, e2, keydata));
101 }
102
103 static inline unsigned int bucket(const struct hashmap *map,
104 const struct hashmap_entry *key)
105 {
106 return key->hash & (map->tablesize - 1);
107 }
108
109 int hashmap_bucket(const struct hashmap *map, unsigned int hash)
110 {
111 return hash & (map->tablesize - 1);
112 }
113
114 static void rehash(struct hashmap *map, unsigned int newsize)
115 {
116 unsigned int i, oldsize = map->tablesize;
117 struct hashmap_entry **oldtable = map->table;
118
119 if (map->disallow_rehash)
120 return;
121
122 alloc_table(map, newsize);
123 for (i = 0; i < oldsize; i++) {
124 struct hashmap_entry *e = oldtable[i];
125 while (e) {
126 struct hashmap_entry *next = e->next;
127 unsigned int b = bucket(map, e);
128 e->next = map->table[b];
129 map->table[b] = e;
130 e = next;
131 }
132 }
133 free(oldtable);
134 }
135
136 static inline struct hashmap_entry **find_entry_ptr(const struct hashmap *map,
137 const struct hashmap_entry *key, const void *keydata)
138 {
139 struct hashmap_entry **e = &map->table[bucket(map, key)];
140 while (*e && !entry_equals(map, *e, key, keydata))
141 e = &(*e)->next;
142 return e;
143 }
144
145 static int always_equal(const void *unused_cmp_data,
146 const void *unused1,
147 const void *unused2,
148 const void *unused_keydata)
149 {
150 return 0;
151 }
152
153 void hashmap_init(struct hashmap *map, hashmap_cmp_fn equals_function,
154 const void *cmpfn_data, size_t initial_size)
155 {
156 unsigned int size = HASHMAP_INITIAL_SIZE;
157
158 memset(map, 0, sizeof(*map));
159
160 map->cmpfn = equals_function ? equals_function : always_equal;
161 map->cmpfn_data = cmpfn_data;
162
163 /* calculate initial table size and allocate the table */
164 initial_size = (unsigned int) ((uint64_t) initial_size * 100
165 / HASHMAP_LOAD_FACTOR);
166 while (initial_size > size)
167 size <<= HASHMAP_RESIZE_BITS;
168 alloc_table(map, size);
169 }
170
171 void hashmap_free(struct hashmap *map, int free_entries)
172 {
173 if (!map || !map->table)
174 return;
175 if (free_entries) {
176 struct hashmap_iter iter;
177 struct hashmap_entry *e;
178 hashmap_iter_init(map, &iter);
179 while ((e = hashmap_iter_next(&iter)))
180 free(e);
181 }
182 free(map->table);
183 memset(map, 0, sizeof(*map));
184 }
185
186 void *hashmap_get(const struct hashmap *map, const void *key, const void *keydata)
187 {
188 return *find_entry_ptr(map, key, keydata);
189 }
190
191 void *hashmap_get_next(const struct hashmap *map, const void *entry)
192 {
193 struct hashmap_entry *e = ((struct hashmap_entry *) entry)->next;
194 for (; e; e = e->next)
195 if (entry_equals(map, entry, e, NULL))
196 return e;
197 return NULL;
198 }
199
200 void hashmap_add(struct hashmap *map, void *entry)
201 {
202 unsigned int b = bucket(map, entry);
203
204 /* add entry */
205 ((struct hashmap_entry *) entry)->next = map->table[b];
206 map->table[b] = entry;
207
208 /* fix size and rehash if appropriate */
209 map->size++;
210 if (map->size > map->grow_at)
211 rehash(map, map->tablesize << HASHMAP_RESIZE_BITS);
212 }
213
214 void *hashmap_remove(struct hashmap *map, const void *key, const void *keydata)
215 {
216 struct hashmap_entry *old;
217 struct hashmap_entry **e = find_entry_ptr(map, key, keydata);
218 if (!*e)
219 return NULL;
220
221 /* remove existing entry */
222 old = *e;
223 *e = old->next;
224 old->next = NULL;
225
226 /* fix size and rehash if appropriate */
227 map->size--;
228 if (map->size < map->shrink_at)
229 rehash(map, map->tablesize >> HASHMAP_RESIZE_BITS);
230 return old;
231 }
232
233 void *hashmap_put(struct hashmap *map, void *entry)
234 {
235 struct hashmap_entry *old = hashmap_remove(map, entry, NULL);
236 hashmap_add(map, entry);
237 return old;
238 }
239
240 void hashmap_iter_init(struct hashmap *map, struct hashmap_iter *iter)
241 {
242 iter->map = map;
243 iter->tablepos = 0;
244 iter->next = NULL;
245 }
246
247 void *hashmap_iter_next(struct hashmap_iter *iter)
248 {
249 struct hashmap_entry *current = iter->next;
250 for (;;) {
251 if (current) {
252 iter->next = current->next;
253 return current;
254 }
255
256 if (iter->tablepos >= iter->map->tablesize)
257 return NULL;
258
259 current = iter->map->table[iter->tablepos++];
260 }
261 }
262
263 struct pool_entry {
264 struct hashmap_entry ent;
265 size_t len;
266 unsigned char data[FLEX_ARRAY];
267 };
268
269 static int pool_entry_cmp(const void *unused_cmp_data,
270 const struct pool_entry *e1,
271 const struct pool_entry *e2,
272 const unsigned char *keydata)
273 {
274 return e1->data != keydata &&
275 (e1->len != e2->len || memcmp(e1->data, keydata, e1->len));
276 }
277
278 const void *memintern(const void *data, size_t len)
279 {
280 static struct hashmap map;
281 struct pool_entry key, *e;
282
283 /* initialize string pool hashmap */
284 if (!map.tablesize)
285 hashmap_init(&map, (hashmap_cmp_fn) pool_entry_cmp, NULL, 0);
286
287 /* lookup interned string in pool */
288 hashmap_entry_init(&key, memhash(data, len));
289 key.len = len;
290 e = hashmap_get(&map, &key, data);
291 if (!e) {
292 /* not found: create it */
293 FLEX_ALLOC_MEM(e, data, data, len);
294 hashmap_entry_init(e, key.ent.hash);
295 e->len = len;
296 hashmap_add(&map, e);
297 }
298 return e->data;
299 }