Line data Source code
1 : /*
2 : * $Id: linkhash.c,v 1.4 2006/01/26 02:16:28 mclark Exp $
3 : *
4 : * Copyright (c) 2004, 2005 Metaparadigm Pte. Ltd.
5 : * Michael Clark <michael@metaparadigm.com>
6 : * Copyright (c) 2009 Hewlett-Packard Development Company, L.P.
7 : *
8 : * This library is free software; you can redistribute it and/or modify
9 : * it under the terms of the MIT license. See COPYING for details.
10 : *
11 : */
12 :
13 : #include "config.h"
14 :
15 : #include "cpl_port.h"
16 :
17 : #include <assert.h>
18 : #include <limits.h>
19 : #include <stdarg.h>
20 : #include <stddef.h>
21 : #include <stdio.h>
22 : #include <stdlib.h>
23 : #include <string.h>
24 :
25 : #ifdef HAVE_ENDIAN_H
26 : #include <endian.h> /* attempt to define endianness */
27 : #endif
28 :
29 : #if defined(_MSC_VER) || defined(__MINGW32__)
30 : #define WIN32_LEAN_AND_MEAN
31 : #include <windows.h> /* Get InterlockedCompareExchange */
32 : #endif
33 :
34 : #include "linkhash.h"
35 : #include "random_seed.h"
36 :
37 : /* hash functions */
38 : static unsigned long lh_char_hash(const void *k);
39 : static unsigned long lh_perllike_str_hash(const void *k);
40 : static lh_hash_fn *char_hash_fn = lh_char_hash;
41 :
42 : /* comparison functions */
43 : int lh_char_equal(const void *k1, const void *k2);
44 : int lh_ptr_equal(const void *k1, const void *k2);
45 :
46 0 : int json_global_set_string_hash(const int h)
47 : {
48 0 : switch (h)
49 : {
50 0 : case JSON_C_STR_HASH_DFLT: char_hash_fn = lh_char_hash; break;
51 0 : case JSON_C_STR_HASH_PERLLIKE: char_hash_fn = lh_perllike_str_hash; break;
52 0 : default: return -1;
53 : }
54 0 : return 0;
55 : }
56 :
57 0 : static unsigned long lh_ptr_hash(const void *k)
58 : {
59 : /* CAW: refactored to be 64bit nice */
60 0 : return (unsigned long)((((ptrdiff_t)k * LH_PRIME) >> 4) & ULONG_MAX);
61 : }
62 :
63 0 : int lh_ptr_equal(const void *k1, const void *k2)
64 : {
65 0 : return (k1 == k2);
66 : }
67 :
68 : /*
69 : * hashlittle from lookup3.c, by Bob Jenkins, May 2006, Public Domain.
70 : * http://burtleburtle.net/bob/c/lookup3.c
71 : * minor modifications to make functions static so no symbols are exported
72 : * minor mofifications to compile with -Werror
73 : */
74 :
75 : /*
76 : -------------------------------------------------------------------------------
77 : lookup3.c, by Bob Jenkins, May 2006, Public Domain.
78 :
79 : These are functions for producing 32-bit hashes for hash table lookup.
80 : hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
81 : are externally useful functions. Routines to test the hash are included
82 : if SELF_TEST is defined. You can use this free for any purpose. It's in
83 : the public domain. It has no warranty.
84 :
85 : You probably want to use hashlittle(). hashlittle() and hashbig()
86 : hash byte arrays. hashlittle() is is faster than hashbig() on
87 : little-endian machines. Intel and AMD are little-endian machines.
88 : On second thought, you probably want hashlittle2(), which is identical to
89 : hashlittle() except it returns two 32-bit hashes for the price of one.
90 : You could implement hashbig2() if you wanted but I haven't bothered here.
91 :
92 : If you want to find a hash of, say, exactly 7 integers, do
93 : a = i1; b = i2; c = i3;
94 : mix(a,b,c);
95 : a += i4; b += i5; c += i6;
96 : mix(a,b,c);
97 : a += i7;
98 : final(a,b,c);
99 : then use c as the hash value. If you have a variable length array of
100 : 4-byte integers to hash, use hashword(). If you have a byte array (like
101 : a character string), use hashlittle(). If you have several byte arrays, or
102 : a mix of things, see the comments above hashlittle().
103 :
104 : Why is this so big? I read 12 bytes at a time into 3 4-byte integers,
105 : then mix those integers. This is fast (you can do a lot more thorough
106 : mixing with 12*3 instructions on 3 integers than you can with 3 instructions
107 : on 1 byte), but shoehorning those bytes into integers efficiently is messy.
108 : -------------------------------------------------------------------------------
109 : */
110 :
111 : /*
112 : * My best guess at if you are big-endian or little-endian. This may
113 : * need adjustment.
114 : */
115 : #if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && __BYTE_ORDER == __LITTLE_ENDIAN) || \
116 : (defined(i386) || defined(__i386__) || defined(__i486__) || defined(__i586__) || \
117 : defined(__i686__) || defined(vax) || defined(MIPSEL))
118 : #define HASH_LITTLE_ENDIAN 1
119 : #define HASH_BIG_ENDIAN 0
120 : #elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && __BYTE_ORDER == __BIG_ENDIAN) || \
121 : (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel))
122 : #define HASH_LITTLE_ENDIAN 0
123 : #define HASH_BIG_ENDIAN 1
124 : #else
125 : #define HASH_LITTLE_ENDIAN 0
126 : #define HASH_BIG_ENDIAN 0
127 : #endif
128 :
129 : #define hashsize(n) ((uint32_t)1 << (n))
130 : #define hashmask(n) (hashsize(n) - 1)
131 : #define rot(x, k) (((x) << (k)) | ((x) >> (32 - (k))))
132 :
133 : /*
134 : -------------------------------------------------------------------------------
135 : mix -- mix 3 32-bit values reversibly.
136 :
137 : This is reversible, so any information in (a,b,c) before mix() is
138 : still in (a,b,c) after mix().
139 :
140 : If four pairs of (a,b,c) inputs are run through mix(), or through
141 : mix() in reverse, there are at least 32 bits of the output that
142 : are sometimes the same for one pair and different for another pair.
143 : This was tested for:
144 : * pairs that differed by one bit, by two bits, in any combination
145 : of top bits of (a,b,c), or in any combination of bottom bits of
146 : (a,b,c).
147 : * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
148 : the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
149 : is commonly produced by subtraction) look like a single 1-bit
150 : difference.
151 : * the base values were pseudorandom, all zero but one bit set, or
152 : all zero plus a counter that starts at zero.
153 :
154 : Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
155 : satisfy this are
156 : 4 6 8 16 19 4
157 : 9 15 3 18 27 15
158 : 14 9 3 7 17 3
159 : Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
160 : for "differ" defined as + with a one-bit base and a two-bit delta. I
161 : used http://burtleburtle.net/bob/hash/avalanche.html to choose
162 : the operations, constants, and arrangements of the variables.
163 :
164 : This does not achieve avalanche. There are input bits of (a,b,c)
165 : that fail to affect some output bits of (a,b,c), especially of a. The
166 : most thoroughly mixed value is c, but it doesn't really even achieve
167 : avalanche in c.
168 :
169 : This allows some parallelism. Read-after-writes are good at doubling
170 : the number of bits affected, so the goal of mixing pulls in the opposite
171 : direction as the goal of parallelism. I did what I could. Rotates
172 : seem to cost as much as shifts on every machine I could lay my hands
173 : on, and rotates are much kinder to the top and bottom bits, so I used
174 : rotates.
175 : -------------------------------------------------------------------------------
176 : */
177 : /* clang-format off */
178 : #define mix(a,b,c) \
179 : { \
180 : a -= c; a ^= rot(c, 4); c += b; \
181 : b -= a; b ^= rot(a, 6); a += c; \
182 : c -= b; c ^= rot(b, 8); b += a; \
183 : a -= c; a ^= rot(c,16); c += b; \
184 : b -= a; b ^= rot(a,19); a += c; \
185 : c -= b; c ^= rot(b, 4); b += a; \
186 : }
187 : /* clang-format on */
188 :
189 : /*
190 : -------------------------------------------------------------------------------
191 : final -- final mixing of 3 32-bit values (a,b,c) into c
192 :
193 : Pairs of (a,b,c) values differing in only a few bits will usually
194 : produce values of c that look totally different. This was tested for
195 : * pairs that differed by one bit, by two bits, in any combination
196 : of top bits of (a,b,c), or in any combination of bottom bits of
197 : (a,b,c).
198 : * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
199 : the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
200 : is commonly produced by subtraction) look like a single 1-bit
201 : difference.
202 : * the base values were pseudorandom, all zero but one bit set, or
203 : all zero plus a counter that starts at zero.
204 :
205 : These constants passed:
206 : 14 11 25 16 4 14 24
207 : 12 14 25 16 4 14 24
208 : and these came close:
209 : 4 8 15 26 3 22 24
210 : 10 8 15 26 3 22 24
211 : 11 8 15 26 3 22 24
212 : -------------------------------------------------------------------------------
213 : */
214 : /* clang-format off */
215 : #define final(a,b,c) \
216 : { \
217 : c ^= b; c -= rot(b,14); \
218 : a ^= c; a -= rot(c,11); \
219 : b ^= a; b -= rot(a,25); \
220 : c ^= b; c -= rot(b,16); \
221 : a ^= c; a -= rot(c,4); \
222 : b ^= a; b -= rot(a,14); \
223 : c ^= b; c -= rot(b,24); \
224 : }
225 : /* clang-format on */
226 :
227 : /*
228 : -------------------------------------------------------------------------------
229 : hashlittle() -- hash a variable-length key into a 32-bit value
230 : k : the key (the unaligned variable-length array of bytes)
231 : length : the length of the key, counting by bytes
232 : initval : can be any 4-byte value
233 : Returns a 32-bit value. Every bit of the key affects every bit of
234 : the return value. Two keys differing by one or two bits will have
235 : totally different hash values.
236 :
237 : The best hash table sizes are powers of 2. There is no need to do
238 : mod a prime (mod is sooo slow!). If you need less than 32 bits,
239 : use a bitmask. For example, if you need only 10 bits, do
240 : h = (h & hashmask(10));
241 : In which case, the hash table should have hashsize(10) elements.
242 :
243 : If you are hashing n strings (uint8_t **)k, do it like this:
244 : for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
245 :
246 : By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this
247 : code any way you wish, private, educational, or commercial. It's free.
248 :
249 : Use for hash table lookup, or anything where one collision in 2^^32 is
250 : acceptable. Do NOT use for cryptographic purposes.
251 : -------------------------------------------------------------------------------
252 : */
253 :
254 : /* clang-format off */
255 :
256 : CPL_NOSANITIZE_UNSIGNED_INT_OVERFLOW
257 2530610 : static uint32_t hashlittle(const void *key, size_t length, uint32_t initval)
258 : {
259 : uint32_t a,b,c; /* internal state */
260 : union
261 : {
262 : const void *ptr;
263 : size_t i;
264 : } u; /* needed for Mac Powerbook G4 */
265 :
266 : /* Set up the internal state */
267 2530610 : a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
268 :
269 2530610 : u.ptr = key;
270 2530610 : if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
271 2491880 : const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
272 :
273 : /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
274 2634210 : while (length > 12)
275 : {
276 142329 : a += k[0];
277 142329 : b += k[1];
278 142329 : c += k[2];
279 142329 : mix(a,b,c);
280 142329 : length -= 12;
281 142329 : k += 3;
282 : }
283 :
284 : /*----------------------------- handle the last (probably partial) block */
285 : /*
286 : * "k[2]&0xffffff" actually reads beyond the end of the string, but
287 : * then masks off the part it's not allowed to read. Because the
288 : * string is aligned, the masked-off tail is in the same word as the
289 : * rest of the string. Every machine with memory protection I've seen
290 : * does it on word boundaries, so is OK with this. But VALGRIND will
291 : * still catch it and complain. The masking trick does make the hash
292 : * noticeably faster for short strings (like English words).
293 : * AddressSanitizer is similarly picky about overrunning
294 : * the buffer. (http://clang.llvm.org/docs/AddressSanitizer.html
295 : */
296 : #ifdef VALGRIND
297 : #define PRECISE_MEMORY_ACCESS 1
298 : #elif defined(__SANITIZE_ADDRESS__) /* GCC's ASAN */
299 : #define PRECISE_MEMORY_ACCESS 1
300 : #elif defined(__SANITIZE_HWADDRESS__) /* GCC's HWASAN */
301 : #define PRECISE_MEMORY_ACCESS 1
302 : #elif defined(__has_feature)
303 : #if __has_feature(address_sanitizer) /* Clang's ASAN */
304 : #define PRECISE_MEMORY_ACCESS 1
305 : #elif __has_feature(hwaddress_sanitizer) /* Clang's HWASAN */
306 : #define PRECISE_MEMORY_ACCESS 1
307 : #endif
308 : #endif
309 : #ifndef PRECISE_MEMORY_ACCESS
310 :
311 2491880 : switch(length)
312 : {
313 35232 : case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
314 55025 : case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
315 90124 : case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
316 498260 : case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
317 72506 : case 8 : b+=k[1]; a+=k[0]; break;
318 64156 : case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
319 61609 : case 6 : b+=k[1]&0xffff; a+=k[0]; break;
320 368544 : case 5 : b+=k[1]&0xff; a+=k[0]; break;
321 667358 : case 4 : a+=k[0]; break;
322 427214 : case 3 : a+=k[0]&0xffffff; break;
323 45037 : case 2 : a+=k[0]&0xffff; break;
324 106676 : case 1 : a+=k[0]&0xff; break;
325 138 : case 0 : return c; /* zero length strings require no mixing */
326 : }
327 :
328 : #else /* make valgrind happy */
329 :
330 : const uint8_t *k8 = (const uint8_t *)k;
331 : switch(length)
332 : {
333 : case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
334 : case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
335 : case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
336 : case 9 : c+=k8[8]; /* fall through */
337 : case 8 : b+=k[1]; a+=k[0]; break;
338 : case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
339 : case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
340 : case 5 : b+=k8[4]; /* fall through */
341 : case 4 : a+=k[0]; break;
342 : case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
343 : case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
344 : case 1 : a+=k8[0]; break;
345 : case 0 : return c;
346 : }
347 :
348 : #endif /* !valgrind */
349 :
350 2491740 : }
351 38730 : else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0))
352 : {
353 16198 : const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */
354 : const uint8_t *k8;
355 :
356 : /*--------------- all but last block: aligned reads and different mixing */
357 16466 : while (length > 12)
358 : {
359 268 : a += k[0] + (((uint32_t)k[1])<<16);
360 268 : b += k[2] + (((uint32_t)k[3])<<16);
361 268 : c += k[4] + (((uint32_t)k[5])<<16);
362 268 : mix(a,b,c);
363 268 : length -= 12;
364 268 : k += 6;
365 : }
366 :
367 : /*----------------------------- handle the last (probably partial) block */
368 16198 : k8 = (const uint8_t *)k;
369 16198 : switch(length)
370 : {
371 0 : case 12: c+=k[4]+(((uint32_t)k[5])<<16);
372 0 : b+=k[2]+(((uint32_t)k[3])<<16);
373 0 : a+=k[0]+(((uint32_t)k[1])<<16);
374 0 : break;
375 7272 : case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
376 7839 : case 10: c+=k[4];
377 7839 : b+=k[2]+(((uint32_t)k[3])<<16);
378 7839 : a+=k[0]+(((uint32_t)k[1])<<16);
379 7839 : break;
380 14 : case 9 : c+=k8[8]; /* fall through */
381 6606 : case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
382 6606 : a+=k[0]+(((uint32_t)k[1])<<16);
383 6606 : break;
384 230 : case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
385 416 : case 6 : b+=k[2];
386 416 : a+=k[0]+(((uint32_t)k[1])<<16);
387 416 : break;
388 261 : case 5 : b+=k8[4]; /* fall through */
389 1030 : case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
390 1030 : break;
391 182 : case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
392 215 : case 2 : a+=k[0];
393 215 : break;
394 42 : case 1 : a+=k8[0];
395 42 : break;
396 50 : case 0 : return c; /* zero length requires no mixing */
397 : }
398 :
399 16148 : }
400 : else
401 : {
402 : /* need to read the key one byte at a time */
403 22532 : const uint8_t *k = (const uint8_t *)key;
404 :
405 : /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
406 23972 : while (length > 12)
407 : {
408 1440 : a += k[0];
409 1440 : a += ((uint32_t)k[1])<<8;
410 1440 : a += ((uint32_t)k[2])<<16;
411 1440 : a += ((uint32_t)k[3])<<24;
412 1440 : b += k[4];
413 1440 : b += ((uint32_t)k[5])<<8;
414 1440 : b += ((uint32_t)k[6])<<16;
415 1440 : b += ((uint32_t)k[7])<<24;
416 1440 : c += k[8];
417 1440 : c += ((uint32_t)k[9])<<8;
418 1440 : c += ((uint32_t)k[10])<<16;
419 1440 : c += ((uint32_t)k[11])<<24;
420 1440 : mix(a,b,c);
421 1440 : length -= 12;
422 1440 : k += 12;
423 : }
424 :
425 : /*-------------------------------- last block: affect all 32 bits of (c) */
426 22532 : switch(length) /* all the case statements fall through */
427 : {
428 130 : case 12: c+=((uint32_t)k[11])<<24; /* FALLTHRU */
429 1175 : case 11: c+=((uint32_t)k[10])<<16; /* FALLTHRU */
430 4147 : case 10: c+=((uint32_t)k[9])<<8; /* FALLTHRU */
431 4929 : case 9 : c+=k[8]; /* FALLTHRU */
432 5273 : case 8 : b+=((uint32_t)k[7])<<24; /* FALLTHRU */
433 5318 : case 7 : b+=((uint32_t)k[6])<<16; /* FALLTHRU */
434 5569 : case 6 : b+=((uint32_t)k[5])<<8; /* FALLTHRU */
435 7131 : case 5 : b+=k[4]; /* FALLTHRU */
436 13264 : case 4 : a+=((uint32_t)k[3])<<24; /* FALLTHRU */
437 13407 : case 3 : a+=((uint32_t)k[2])<<16; /* FALLTHRU */
438 22228 : case 2 : a+=((uint32_t)k[1])<<8; /* FALLTHRU */
439 22532 : case 1 : a+=k[0];
440 22532 : break;
441 0 : case 0 : return c;
442 : }
443 2530420 : }
444 :
445 2530420 : final(a,b,c);
446 2530420 : return c;
447 : }
448 : /* clang-format on */
449 :
450 : /* a simple hash function similar to what perl does for strings.
451 : * for good results, the string should not be excessivly large.
452 : */
453 0 : static unsigned long lh_perllike_str_hash(const void *k)
454 : {
455 0 : const char *rkey = (const char *)k;
456 0 : unsigned hashval = 1;
457 :
458 0 : while (*rkey)
459 0 : hashval = hashval * 33 + *rkey++;
460 :
461 0 : return hashval;
462 : }
463 :
464 2530610 : static unsigned long lh_char_hash(const void *k)
465 : {
466 : #if defined _MSC_VER || defined __MINGW32__
467 : #define RANDOM_SEED_TYPE LONG
468 : #else
469 : #define RANDOM_SEED_TYPE int
470 : #endif
471 : static volatile RANDOM_SEED_TYPE random_seed = -1;
472 :
473 2530610 : if (random_seed == -1)
474 : {
475 : RANDOM_SEED_TYPE seed;
476 : /* we can't use -1 as it is the uninitialized sentinel */
477 143 : while ((seed = json_c_get_random_seed()) == -1) {}
478 : #if SIZEOF_INT == 8 && defined __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8
479 : #define USE_SYNC_COMPARE_AND_SWAP 1
480 : #endif
481 : #if SIZEOF_INT == 4 && defined __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4
482 : #define USE_SYNC_COMPARE_AND_SWAP 1
483 : #endif
484 : #if SIZEOF_INT == 2 && defined __GCC_HAVE_SYNC_COMPARE_AND_SWAP_2
485 : #define USE_SYNC_COMPARE_AND_SWAP 1
486 : #endif
487 : #if defined USE_SYNC_COMPARE_AND_SWAP
488 143 : (void)__sync_val_compare_and_swap(&random_seed, -1, seed);
489 : #elif defined _MSC_VER || defined __MINGW32__
490 : InterlockedCompareExchange(&random_seed, seed, -1);
491 : #else
492 : //#warning "racy random seed initializtion if used by multiple threads"
493 : random_seed = seed; /* potentially racy */
494 : #endif
495 : }
496 :
497 2530610 : return hashlittle((const char *)k, strlen((const char *)k), random_seed);
498 : }
499 :
500 1147410 : int lh_char_equal(const void *k1, const void *k2)
501 : {
502 1147410 : return (strcmp((const char *)k1, (const char *)k2) == 0);
503 : }
504 :
505 479365 : struct lh_table *lh_table_new(int size, lh_entry_free_fn *free_fn, lh_hash_fn *hash_fn,
506 : lh_equal_fn *equal_fn)
507 : {
508 : int i;
509 : struct lh_table *t;
510 :
511 : /* Allocate space for elements to avoid divisions by zero. */
512 479365 : assert(size > 0);
513 479365 : t = (struct lh_table *)calloc(1, sizeof(struct lh_table));
514 479365 : if (!t)
515 0 : return NULL;
516 :
517 479365 : t->count = 0;
518 479365 : t->size = size;
519 479365 : t->table = (struct lh_entry *)calloc(size, sizeof(struct lh_entry));
520 479365 : if (!t->table)
521 : {
522 0 : free(t);
523 0 : return NULL;
524 : }
525 479365 : t->free_fn = free_fn;
526 479365 : t->hash_fn = hash_fn;
527 479365 : t->equal_fn = equal_fn;
528 8235810 : for (i = 0; i < size; i++)
529 7756450 : t->table[i].k = LH_EMPTY;
530 479365 : return t;
531 : }
532 :
533 477170 : struct lh_table *lh_kchar_table_new(int size, lh_entry_free_fn *free_fn)
534 : {
535 477170 : return lh_table_new(size, free_fn, char_hash_fn, lh_char_equal);
536 : }
537 :
538 0 : struct lh_table *lh_kptr_table_new(int size, lh_entry_free_fn *free_fn)
539 : {
540 0 : return lh_table_new(size, free_fn, lh_ptr_hash, lh_ptr_equal);
541 : }
542 :
543 2195 : int lh_table_resize(struct lh_table *t, int new_size)
544 : {
545 : struct lh_table *new_t;
546 : struct lh_entry *ent;
547 :
548 2195 : new_t = lh_table_new(new_size, NULL, t->hash_fn, t->equal_fn);
549 2195 : if (new_t == NULL)
550 0 : return -1;
551 :
552 43686 : for (ent = t->head; ent != NULL; ent = ent->next)
553 : {
554 41491 : unsigned long h = lh_get_hash(new_t, ent->k);
555 41491 : unsigned int opts = 0;
556 41491 : if (ent->k_is_constant)
557 0 : opts = JSON_C_OBJECT_KEY_IS_CONSTANT;
558 41491 : if (lh_table_insert_w_hash(new_t, ent->k, ent->v, h, opts) != 0)
559 : {
560 0 : lh_table_free(new_t);
561 0 : return -1;
562 : }
563 : }
564 2195 : free(t->table);
565 2195 : t->table = new_t->table;
566 2195 : t->size = new_size;
567 2195 : t->head = new_t->head;
568 2195 : t->tail = new_t->tail;
569 2195 : free(new_t);
570 :
571 2195 : return 0;
572 : }
573 :
574 477170 : void lh_table_free(struct lh_table *t)
575 : {
576 : struct lh_entry *c;
577 477170 : if (t->free_fn)
578 : {
579 1965650 : for (c = t->head; c != NULL; c = c->next)
580 1488480 : t->free_fn(c);
581 : }
582 477170 : free(t->table);
583 477170 : free(t);
584 477170 : }
585 :
586 1531250 : int lh_table_insert_w_hash(struct lh_table *t, const void *k, const void *v, const unsigned long h,
587 : const unsigned opts)
588 : {
589 : unsigned long n;
590 :
591 1531250 : if (t->count >= t->size * LH_LOAD_FACTOR)
592 : {
593 : /* Avoid signed integer overflow with large tables. */
594 2195 : int new_size = (t->size > INT_MAX / 2) ? INT_MAX : (t->size * 2);
595 2195 : if (t->size == INT_MAX || lh_table_resize(t, new_size) != 0)
596 0 : return -1;
597 : }
598 :
599 1531250 : n = h % t->size;
600 :
601 : while (1)
602 : {
603 1690600 : if (t->table[n].k == LH_EMPTY || t->table[n].k == LH_FREED)
604 : break;
605 159357 : if ((int)++n == t->size)
606 22421 : n = 0;
607 : }
608 :
609 1531250 : t->table[n].k = k;
610 1531250 : t->table[n].k_is_constant = (opts & JSON_C_OBJECT_KEY_IS_CONSTANT);
611 1531250 : t->table[n].v = v;
612 1531250 : t->count++;
613 :
614 1531250 : if (t->head == NULL)
615 : {
616 380339 : t->head = t->tail = &t->table[n];
617 380339 : t->table[n].next = t->table[n].prev = NULL;
618 : }
619 : else
620 : {
621 1150910 : t->tail->next = &t->table[n];
622 1150910 : t->table[n].prev = t->tail;
623 1150910 : t->table[n].next = NULL;
624 1150910 : t->tail = &t->table[n];
625 : }
626 :
627 1531250 : return 0;
628 : }
629 0 : int lh_table_insert(struct lh_table *t, const void *k, const void *v)
630 : {
631 0 : return lh_table_insert_w_hash(t, k, v, lh_get_hash(t, k), 0);
632 : }
633 :
634 2489120 : struct lh_entry *lh_table_lookup_entry_w_hash(struct lh_table *t, const void *k,
635 : const unsigned long h)
636 : {
637 2489120 : unsigned long n = h % t->size;
638 2489120 : int count = 0;
639 :
640 2851070 : while (count < t->size)
641 : {
642 2851070 : if (t->table[n].k == LH_EMPTY)
643 1702240 : return NULL;
644 1148820 : if (t->table[n].k != LH_FREED && t->equal_fn(t->table[n].k, k))
645 786873 : return &t->table[n];
646 361952 : if ((int)++n == t->size)
647 38672 : n = 0;
648 361952 : count++;
649 : }
650 0 : return NULL;
651 : }
652 :
653 999218 : struct lh_entry *lh_table_lookup_entry(struct lh_table *t, const void *k)
654 : {
655 999218 : return lh_table_lookup_entry_w_hash(t, k, lh_get_hash(t, k));
656 : }
657 :
658 988899 : json_bool lh_table_lookup_ex(struct lh_table *t, const void *k, void **v)
659 : {
660 988899 : struct lh_entry *e = lh_table_lookup_entry(t, k);
661 988899 : if (e != NULL)
662 : {
663 785456 : if (v != NULL)
664 785456 : *v = lh_entry_v(e);
665 785456 : return 1; /* key found */
666 : }
667 203443 : if (v != NULL)
668 203443 : *v = NULL;
669 203443 : return 0; /* key not found */
670 : }
671 :
672 1273 : int lh_table_delete_entry(struct lh_table *t, struct lh_entry *e)
673 : {
674 : /* CAW: fixed to be 64bit nice, still need the crazy negative case... */
675 1273 : ptrdiff_t n = (ptrdiff_t)(e - t->table);
676 :
677 : /* CAW: this is bad, really bad, maybe stack goes other direction on this machine... */
678 1273 : if (n < 0)
679 : {
680 0 : return -2;
681 : }
682 :
683 1273 : if (t->table[n].k == LH_EMPTY || t->table[n].k == LH_FREED)
684 0 : return -1;
685 1273 : t->count--;
686 1273 : if (t->free_fn)
687 1273 : t->free_fn(e);
688 1273 : t->table[n].v = NULL;
689 1273 : t->table[n].k = LH_FREED;
690 1273 : if (t->tail == &t->table[n] && t->head == &t->table[n])
691 : {
692 311 : t->head = t->tail = NULL;
693 : }
694 962 : else if (t->head == &t->table[n])
695 : {
696 376 : t->head->next->prev = NULL;
697 376 : t->head = t->head->next;
698 : }
699 586 : else if (t->tail == &t->table[n])
700 : {
701 100 : t->tail->prev->next = NULL;
702 100 : t->tail = t->tail->prev;
703 : }
704 : else
705 : {
706 486 : t->table[n].prev->next = t->table[n].next;
707 486 : t->table[n].next->prev = t->table[n].prev;
708 : }
709 1273 : t->table[n].next = t->table[n].prev = NULL;
710 1273 : return 0;
711 : }
712 :
713 10319 : int lh_table_delete(struct lh_table *t, const void *k)
714 : {
715 10319 : struct lh_entry *e = lh_table_lookup_entry(t, k);
716 10319 : if (!e)
717 9046 : return -1;
718 1273 : return lh_table_delete_entry(t, e);
719 : }
720 :
721 3 : int lh_table_length(struct lh_table *t)
722 : {
723 3 : return t->count;
724 : }
|
