1 |
|
|
/* $OpenBSD: ohash.c,v 1.1 2014/06/02 18:52:03 deraadt Exp $ */ |
2 |
|
|
|
3 |
|
|
/* Copyright (c) 1999, 2004 Marc Espie <espie@openbsd.org> |
4 |
|
|
* |
5 |
|
|
* Permission to use, copy, modify, and distribute this software for any |
6 |
|
|
* purpose with or without fee is hereby granted, provided that the above |
7 |
|
|
* copyright notice and this permission notice appear in all copies. |
8 |
|
|
* |
9 |
|
|
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES |
10 |
|
|
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF |
11 |
|
|
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR |
12 |
|
|
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES |
13 |
|
|
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN |
14 |
|
|
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF |
15 |
|
|
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. |
16 |
|
|
*/ |
17 |
|
|
|
18 |
|
|
#include <stddef.h> |
19 |
|
|
#include <stdint.h> |
20 |
|
|
#include <stdlib.h> |
21 |
|
|
#include <string.h> |
22 |
|
|
#include <limits.h> |
23 |
|
|
#include "ohash.h" |
24 |
|
|
|
25 |
|
|
struct _ohash_record { |
26 |
|
|
uint32_t hv; |
27 |
|
|
const char *p; |
28 |
|
|
}; |
29 |
|
|
|
30 |
|
|
#define DELETED ((const char *)h) |
31 |
|
|
#define NONE (h->size) |
32 |
|
|
|
33 |
|
|
/* Don't bother changing the hash table if the change is small enough. */ |
34 |
|
|
#define MINSIZE (1UL << 4) |
35 |
|
|
#define MINDELETED 4 |
36 |
|
|
|
37 |
|
|
static void ohash_resize(struct ohash *); |
38 |
|
|
|
39 |
|
|
|
40 |
|
|
/* This handles the common case of variable length keys, where the |
41 |
|
|
* key is stored at the end of the record. |
42 |
|
|
*/ |
43 |
|
|
void * |
44 |
|
|
ohash_create_entry(struct ohash_info *i, const char *start, const char **end) |
45 |
|
|
{ |
46 |
|
|
char *p; |
47 |
|
|
|
48 |
|
|
if (!*end) |
49 |
|
|
*end = start + strlen(start); |
50 |
|
|
p = (i->alloc)(i->key_offset + (*end - start) + 1, i->data); |
51 |
|
|
if (p) { |
52 |
|
|
memcpy(p+i->key_offset, start, *end-start); |
53 |
|
|
p[i->key_offset + (*end - start)] = '\0'; |
54 |
|
|
} |
55 |
|
|
return (void *)p; |
56 |
|
|
} |
57 |
|
|
|
58 |
|
|
/* hash_delete only frees the hash structure. Use hash_first/hash_next |
59 |
|
|
* to free entries as well. */ |
60 |
|
|
void |
61 |
|
|
ohash_delete(struct ohash *h) |
62 |
|
|
{ |
63 |
|
|
(h->info.free)(h->t, h->info.data); |
64 |
|
|
#ifndef NDEBUG |
65 |
|
|
h->t = NULL; |
66 |
|
|
#endif |
67 |
|
|
} |
68 |
|
|
|
69 |
|
|
static void |
70 |
|
|
ohash_resize(struct ohash *h) |
71 |
|
|
{ |
72 |
|
|
struct _ohash_record *n; |
73 |
|
|
size_t ns; |
74 |
|
|
unsigned int j; |
75 |
|
|
unsigned int i, incr; |
76 |
|
|
|
77 |
|
|
if (4 * h->deleted < h->total) { |
78 |
|
|
if (h->size >= (UINT_MAX >> 1U)) |
79 |
|
|
ns = UINT_MAX; |
80 |
|
|
else |
81 |
|
|
ns = h->size << 1U; |
82 |
|
|
} else if (3 * h->deleted > 2 * h->total) |
83 |
|
|
ns = h->size >> 1U; |
84 |
|
|
else |
85 |
|
|
ns = h->size; |
86 |
|
|
if (ns < MINSIZE) |
87 |
|
|
ns = MINSIZE; |
88 |
|
|
#ifdef STATS_HASH |
89 |
|
|
STAT_HASH_EXPAND++; |
90 |
|
|
STAT_HASH_SIZE += ns - h->size; |
91 |
|
|
#endif |
92 |
|
|
|
93 |
|
|
n = (h->info.calloc)(ns, sizeof(struct _ohash_record), h->info.data); |
94 |
|
|
if (!n) |
95 |
|
|
return; |
96 |
|
|
|
97 |
|
|
for (j = 0; j < h->size; j++) { |
98 |
|
|
if (h->t[j].p != NULL && h->t[j].p != DELETED) { |
99 |
|
|
i = h->t[j].hv % ns; |
100 |
|
|
incr = ((h->t[j].hv % (ns - 2)) & ~1) + 1; |
101 |
|
|
while (n[i].p != NULL) { |
102 |
|
|
i += incr; |
103 |
|
|
if (i >= ns) |
104 |
|
|
i -= ns; |
105 |
|
|
} |
106 |
|
|
n[i].hv = h->t[j].hv; |
107 |
|
|
n[i].p = h->t[j].p; |
108 |
|
|
} |
109 |
|
|
} |
110 |
|
|
(h->info.free)(h->t, h->info.data); |
111 |
|
|
h->t = n; |
112 |
|
|
h->size = ns; |
113 |
|
|
h->total -= h->deleted; |
114 |
|
|
h->deleted = 0; |
115 |
|
|
} |
116 |
|
|
|
117 |
|
|
void * |
118 |
|
|
ohash_remove(struct ohash *h, unsigned int i) |
119 |
|
|
{ |
120 |
|
|
void *result = (void *)h->t[i].p; |
121 |
|
|
|
122 |
|
|
if (result == NULL || result == DELETED) |
123 |
|
|
return NULL; |
124 |
|
|
|
125 |
|
|
#ifdef STATS_HASH |
126 |
|
|
STAT_HASH_ENTRIES--; |
127 |
|
|
#endif |
128 |
|
|
h->t[i].p = DELETED; |
129 |
|
|
h->deleted++; |
130 |
|
|
if (h->deleted >= MINDELETED && 4 * h->deleted > h->total) |
131 |
|
|
ohash_resize(h); |
132 |
|
|
return result; |
133 |
|
|
} |
134 |
|
|
|
135 |
|
|
void * |
136 |
|
|
ohash_find(struct ohash *h, unsigned int i) |
137 |
|
|
{ |
138 |
|
|
if (h->t[i].p == DELETED) |
139 |
|
|
return NULL; |
140 |
|
|
else |
141 |
|
|
return (void *)h->t[i].p; |
142 |
|
|
} |
143 |
|
|
|
144 |
|
|
void * |
145 |
|
|
ohash_insert(struct ohash *h, unsigned int i, void *p) |
146 |
|
|
{ |
147 |
|
|
#ifdef STATS_HASH |
148 |
|
|
STAT_HASH_ENTRIES++; |
149 |
|
|
#endif |
150 |
|
|
if (h->t[i].p == DELETED) { |
151 |
|
|
h->deleted--; |
152 |
|
|
h->t[i].p = p; |
153 |
|
|
} else { |
154 |
|
|
h->t[i].p = p; |
155 |
|
|
/* Arbitrary resize boundary. Tweak if not efficient enough. */ |
156 |
|
|
if (++h->total * 4 > h->size * 3) |
157 |
|
|
ohash_resize(h); |
158 |
|
|
} |
159 |
|
|
return p; |
160 |
|
|
} |
161 |
|
|
|
162 |
|
|
unsigned int |
163 |
|
|
ohash_entries(struct ohash *h) |
164 |
|
|
{ |
165 |
|
|
return h->total - h->deleted; |
166 |
|
|
} |
167 |
|
|
|
168 |
|
|
void * |
169 |
|
|
ohash_first(struct ohash *h, unsigned int *pos) |
170 |
|
|
{ |
171 |
|
|
*pos = 0; |
172 |
|
|
return ohash_next(h, pos); |
173 |
|
|
} |
174 |
|
|
|
175 |
|
|
void * |
176 |
|
|
ohash_next(struct ohash *h, unsigned int *pos) |
177 |
|
|
{ |
178 |
|
|
for (; *pos < h->size; (*pos)++) |
179 |
|
|
if (h->t[*pos].p != DELETED && h->t[*pos].p != NULL) |
180 |
|
|
return (void *)h->t[(*pos)++].p; |
181 |
|
|
return NULL; |
182 |
|
|
} |
183 |
|
|
|
184 |
|
|
void |
185 |
|
|
ohash_init(struct ohash *h, unsigned int size, struct ohash_info *info) |
186 |
|
|
{ |
187 |
|
|
h->size = 1UL << size; |
188 |
|
|
if (h->size < MINSIZE) |
189 |
|
|
h->size = MINSIZE; |
190 |
|
|
#ifdef STATS_HASH |
191 |
|
|
STAT_HASH_CREATION++; |
192 |
|
|
STAT_HASH_SIZE += h->size; |
193 |
|
|
#endif |
194 |
|
|
/* Copy info so that caller may free it. */ |
195 |
|
|
h->info.key_offset = info->key_offset; |
196 |
|
|
h->info.calloc = info->calloc; |
197 |
|
|
h->info.free = info->free; |
198 |
|
|
h->info.alloc = info->alloc; |
199 |
|
|
h->info.data = info->data; |
200 |
|
|
h->t = (h->info.calloc)(h->size, sizeof(struct _ohash_record), |
201 |
|
|
h->info.data); |
202 |
|
|
h->total = h->deleted = 0; |
203 |
|
|
} |
204 |
|
|
|
205 |
|
|
uint32_t |
206 |
|
|
ohash_interval(const char *s, const char **e) |
207 |
|
|
{ |
208 |
|
|
uint32_t k; |
209 |
|
|
|
210 |
|
|
if (!*e) |
211 |
|
|
*e = s + strlen(s); |
212 |
|
|
if (s == *e) |
213 |
|
|
k = 0; |
214 |
|
|
else |
215 |
|
|
k = *s++; |
216 |
|
|
while (s != *e) |
217 |
|
|
k = ((k << 2) | (k >> 30)) ^ *s++; |
218 |
|
|
return k; |
219 |
|
|
} |
220 |
|
|
|
221 |
|
|
unsigned int |
222 |
|
|
ohash_lookup_interval(struct ohash *h, const char *start, const char *end, |
223 |
|
|
uint32_t hv) |
224 |
|
|
{ |
225 |
|
|
unsigned int i, incr; |
226 |
|
|
unsigned int empty; |
227 |
|
|
|
228 |
|
|
#ifdef STATS_HASH |
229 |
|
|
STAT_HASH_LOOKUP++; |
230 |
|
|
#endif |
231 |
|
|
empty = NONE; |
232 |
|
|
i = hv % h->size; |
233 |
|
|
incr = ((hv % (h->size-2)) & ~1) + 1; |
234 |
|
|
while (h->t[i].p != NULL) { |
235 |
|
|
#ifdef STATS_HASH |
236 |
|
|
STAT_HASH_LENGTH++; |
237 |
|
|
#endif |
238 |
|
|
if (h->t[i].p == DELETED) { |
239 |
|
|
if (empty == NONE) |
240 |
|
|
empty = i; |
241 |
|
|
} else if (h->t[i].hv == hv && |
242 |
|
|
strncmp(h->t[i].p+h->info.key_offset, start, |
243 |
|
|
end - start) == 0 && |
244 |
|
|
(h->t[i].p+h->info.key_offset)[end-start] == '\0') { |
245 |
|
|
if (empty != NONE) { |
246 |
|
|
h->t[empty].hv = hv; |
247 |
|
|
h->t[empty].p = h->t[i].p; |
248 |
|
|
h->t[i].p = DELETED; |
249 |
|
|
return empty; |
250 |
|
|
} else { |
251 |
|
|
#ifdef STATS_HASH |
252 |
|
|
STAT_HASH_POSITIVE++; |
253 |
|
|
#endif |
254 |
|
|
return i; |
255 |
|
|
} |
256 |
|
|
} |
257 |
|
|
i += incr; |
258 |
|
|
if (i >= h->size) |
259 |
|
|
i -= h->size; |
260 |
|
|
} |
261 |
|
|
|
262 |
|
|
/* Found an empty position. */ |
263 |
|
|
if (empty != NONE) |
264 |
|
|
i = empty; |
265 |
|
|
h->t[i].hv = hv; |
266 |
|
|
return i; |
267 |
|
|
} |
268 |
|
|
|
269 |
|
|
unsigned int |
270 |
|
|
ohash_lookup_memory(struct ohash *h, const char *k, size_t size, uint32_t hv) |
271 |
|
|
{ |
272 |
|
|
unsigned int i, incr; |
273 |
|
|
unsigned int empty; |
274 |
|
|
|
275 |
|
|
#ifdef STATS_HASH |
276 |
|
|
STAT_HASH_LOOKUP++; |
277 |
|
|
#endif |
278 |
|
|
empty = NONE; |
279 |
|
|
i = hv % h->size; |
280 |
|
|
incr = ((hv % (h->size-2)) & ~1) + 1; |
281 |
|
|
while (h->t[i].p != NULL) { |
282 |
|
|
#ifdef STATS_HASH |
283 |
|
|
STAT_HASH_LENGTH++; |
284 |
|
|
#endif |
285 |
|
|
if (h->t[i].p == DELETED) { |
286 |
|
|
if (empty == NONE) |
287 |
|
|
empty = i; |
288 |
|
|
} else if (h->t[i].hv == hv && |
289 |
|
|
memcmp(h->t[i].p+h->info.key_offset, k, size) == 0) { |
290 |
|
|
if (empty != NONE) { |
291 |
|
|
h->t[empty].hv = hv; |
292 |
|
|
h->t[empty].p = h->t[i].p; |
293 |
|
|
h->t[i].p = DELETED; |
294 |
|
|
return empty; |
295 |
|
|
} else { |
296 |
|
|
#ifdef STATS_HASH |
297 |
|
|
STAT_HASH_POSITIVE++; |
298 |
|
|
#endif |
299 |
|
|
} return i; |
300 |
|
|
} |
301 |
|
|
i += incr; |
302 |
|
|
if (i >= h->size) |
303 |
|
|
i -= h->size; |
304 |
|
|
} |
305 |
|
|
|
306 |
|
|
/* Found an empty position. */ |
307 |
|
|
if (empty != NONE) |
308 |
|
|
i = empty; |
309 |
|
|
h->t[i].hv = hv; |
310 |
|
|
return i; |
311 |
|
|
} |
312 |
|
|
|
313 |
|
|
unsigned int |
314 |
|
|
ohash_qlookup(struct ohash *h, const char *s) |
315 |
|
|
{ |
316 |
|
|
const char *e = NULL; |
317 |
|
|
return ohash_qlookupi(h, s, &e); |
318 |
|
|
} |
319 |
|
|
|
320 |
|
|
unsigned int |
321 |
|
|
ohash_qlookupi(struct ohash *h, const char *s, const char **e) |
322 |
|
|
{ |
323 |
|
|
uint32_t hv; |
324 |
|
|
|
325 |
|
|
hv = ohash_interval(s, e); |
326 |
|
|
return ohash_lookup_interval(h, s, *e, hv); |
327 |
|
|
} |