murmurhash3 0.1.1
Sign up to get free protection for your applications and to get access to all the features.
- data/ext/murmurhash3/extconf.rb +8 -0
- data/ext/murmurhash3/murmur3.c +441 -0
- data/lib/murmurhash3.rb +16 -0
- data/lib/murmurhash3/pure_ruby.rb +155 -0
- data/lib/murmurhash3/version.rb +3 -0
- data/test/test_murmur.rb +76 -0
- metadata +54 -0
@@ -0,0 +1,441 @@
|
|
1
|
+
#include <ruby.h>
|
2
|
+
/*-----------------------------------------------------------------------------
|
3
|
+
* MurmurHash3 was written by Austin Appleby, and is placed in the public
|
4
|
+
* domain. The author hereby disclaims copyright to this source code.
|
5
|
+
|
6
|
+
* Note - The x86 and x64 versions do _not_ produce the same results, as the
|
7
|
+
* algorithms are optimized for their respective platforms. You can still
|
8
|
+
* compile and run any of them on any platform, but your performance with the
|
9
|
+
* non-native version will be less than optimal.
|
10
|
+
*/
|
11
|
+
|
12
|
+
typedef unsigned char uint8_t;
|
13
|
+
typedef unsigned int uint32_t;
|
14
|
+
#if SIZEOF_LONG == 8
|
15
|
+
typedef unsigned long uint64_t;
|
16
|
+
#else
|
17
|
+
typedef unsigned long long uint64_t;
|
18
|
+
#endif
|
19
|
+
|
20
|
+
/*-----------------------------------------------------------------------------
|
21
|
+
* Platform-specific functions and macros
|
22
|
+
*/
|
23
|
+
|
24
|
+
#ifdef __GNUC__
|
25
|
+
#define FORCE_INLINE __attribute__((always_inline))
|
26
|
+
#elif defined(_MSC_VER)
|
27
|
+
#define FORCE_INLINE __forceinline
|
28
|
+
#else
|
29
|
+
#define FORCE_INLINE
|
30
|
+
#endif
|
31
|
+
|
32
|
+
#if defined(_MSC_VER)
|
33
|
+
|
34
|
+
#define ROTL32(x,y) _rotl(x,y)
|
35
|
+
#define ROTL64(x,y) _rotl64(x,y)
|
36
|
+
|
37
|
+
#define BIG_CONSTANT(x) (x)
|
38
|
+
|
39
|
+
#else
|
40
|
+
|
41
|
+
static inline FORCE_INLINE uint32_t
|
42
|
+
rotl32 ( uint32_t x, int8_t r )
|
43
|
+
{
|
44
|
+
return (x << r) | (x >> (32 - r));
|
45
|
+
}
|
46
|
+
|
47
|
+
static inline FORCE_INLINE uint64_t
|
48
|
+
rotl64 ( uint64_t x, int8_t r )
|
49
|
+
{
|
50
|
+
return (x << r) | (x >> (64 - r));
|
51
|
+
}
|
52
|
+
|
53
|
+
#define ROTL32(x,y) rotl32(x,y)
|
54
|
+
#define ROTL64(x,y) rotl64(x,y)
|
55
|
+
|
56
|
+
#define BIG_CONSTANT(x) (x##LLU)
|
57
|
+
#endif
|
58
|
+
|
59
|
+
/* end platform specific */
|
60
|
+
|
61
|
+
/* Block read - if your platform needs to do endian-swapping or can only
|
62
|
+
* handle aligned reads, do the conversion here */
|
63
|
+
#ifdef WORDS_BIGENDIAN
|
64
|
+
#if GCC_VERSION_SINCE(4,3,0)
|
65
|
+
# define swap32(x) __builtin_bswap32(x)
|
66
|
+
# define swap64(x) __builtin_bswap64(x)
|
67
|
+
#endif
|
68
|
+
|
69
|
+
#ifndef swap32
|
70
|
+
# define swap32(x) ((((x)&0xFF)<<24) \
|
71
|
+
|(((x)>>24)&0xFF) \
|
72
|
+
|(((x)&0x0000FF00)<<8) \
|
73
|
+
|(((x)&0x00FF0000)>>8) )
|
74
|
+
#endif
|
75
|
+
|
76
|
+
#ifndef swap64
|
77
|
+
# ifdef HAVE_INT64_T
|
78
|
+
static inline FORCE_INLINE uint64_t
|
79
|
+
swap64(uint64_t x) {
|
80
|
+
x = (x>>32) | (x << 32);
|
81
|
+
x = ((x & BIG_CONSTANT(0xFFFF0000FFFF0000)) >> 16) |
|
82
|
+
((x & BIG_CONSTANT(0x0000FFFF0000FFFF)) << 16);
|
83
|
+
return ((x & BIG_CONSTANT(0xFF00FF00FF00FF00)) >> 8) |
|
84
|
+
((x & BIG_CONSTANT(0x00FF00FF00FF00FF)) << 8);
|
85
|
+
}
|
86
|
+
# endif
|
87
|
+
|
88
|
+
#endif
|
89
|
+
static inline FORCE_INLINE uint32_t
|
90
|
+
getblock32(const uint32_t * p, int i)
|
91
|
+
{
|
92
|
+
return swap32(p[i]);
|
93
|
+
}
|
94
|
+
|
95
|
+
static inline FORCE_INLINE uint64_t
|
96
|
+
getblock64(const uint64_t * p, int i)
|
97
|
+
{
|
98
|
+
return swap64(p[i]);
|
99
|
+
}
|
100
|
+
#else
|
101
|
+
#define getblock32(p, i) (p[i])
|
102
|
+
#define getblock64(p, i) (p[i])
|
103
|
+
#endif
|
104
|
+
|
105
|
+
/* Finalization mix - force all bits of a hash block to avalanche */
|
106
|
+
|
107
|
+
static inline FORCE_INLINE uint32_t
|
108
|
+
fmix32 ( uint32_t h )
|
109
|
+
{
|
110
|
+
h ^= h >> 16;
|
111
|
+
h *= 0x85ebca6b;
|
112
|
+
h ^= h >> 13;
|
113
|
+
h *= 0xc2b2ae35;
|
114
|
+
h ^= h >> 16;
|
115
|
+
|
116
|
+
return h;
|
117
|
+
}
|
118
|
+
|
119
|
+
static inline FORCE_INLINE uint64_t
|
120
|
+
fmix64 ( uint64_t k )
|
121
|
+
{
|
122
|
+
k ^= k >> 33;
|
123
|
+
k *= BIG_CONSTANT(0xff51afd7ed558ccd);
|
124
|
+
k ^= k >> 33;
|
125
|
+
k *= BIG_CONSTANT(0xc4ceb9fe1a85ec53);
|
126
|
+
k ^= k >> 33;
|
127
|
+
|
128
|
+
return k;
|
129
|
+
}
|
130
|
+
|
131
|
+
static inline FORCE_INLINE uint32_t
|
132
|
+
mmix32(uint32_t k1)
|
133
|
+
{
|
134
|
+
k1 *= 0xcc9e2d51;
|
135
|
+
k1 = ROTL32(k1, 15);
|
136
|
+
return k1 * 0x1b873593;
|
137
|
+
}
|
138
|
+
|
139
|
+
static uint32_t
|
140
|
+
MurmurHash3_x86_32 ( const void * key, int len, uint32_t seed)
|
141
|
+
{
|
142
|
+
const uint8_t * data = (const uint8_t*)key;
|
143
|
+
const int nblocks = len / 4;
|
144
|
+
int i;
|
145
|
+
|
146
|
+
uint32_t h1 = seed;
|
147
|
+
uint32_t k1 = 0;
|
148
|
+
|
149
|
+
|
150
|
+
/* body */
|
151
|
+
|
152
|
+
const uint32_t * blocks = (const uint32_t *)(data + nblocks*4);
|
153
|
+
|
154
|
+
for(i = -nblocks; i; i++)
|
155
|
+
{
|
156
|
+
h1 ^= mmix32(getblock32(blocks, i));
|
157
|
+
h1 = ROTL32(h1,13);
|
158
|
+
h1 = h1*5+0xe6546b64;
|
159
|
+
}
|
160
|
+
|
161
|
+
/* tail */
|
162
|
+
|
163
|
+
data += nblocks*4;
|
164
|
+
|
165
|
+
switch(len & 3)
|
166
|
+
{
|
167
|
+
case 3: k1 ^= data[2] << 16;
|
168
|
+
case 2: k1 ^= data[1] << 8;
|
169
|
+
case 1: k1 ^= data[0];
|
170
|
+
h1 ^= mmix32(k1);
|
171
|
+
};
|
172
|
+
|
173
|
+
/* finalization */
|
174
|
+
|
175
|
+
h1 ^= len;
|
176
|
+
|
177
|
+
h1 = fmix32(h1);
|
178
|
+
|
179
|
+
return h1;
|
180
|
+
}
|
181
|
+
|
182
|
+
#define C1_128 BIG_CONSTANT(0x87c37b91114253d5)
|
183
|
+
#define C2_128 BIG_CONSTANT(0x4cf5ad432745937f)
|
184
|
+
|
185
|
+
static inline FORCE_INLINE uint64_t
|
186
|
+
mmix128_1(uint64_t k1)
|
187
|
+
{
|
188
|
+
k1 *= C1_128;
|
189
|
+
k1 = ROTL64(k1, 31);
|
190
|
+
return k1 * C2_128;
|
191
|
+
}
|
192
|
+
|
193
|
+
static inline FORCE_INLINE uint64_t
|
194
|
+
mmix128_2(uint64_t k2)
|
195
|
+
{
|
196
|
+
k2 *= C2_128;
|
197
|
+
k2 = ROTL64(k2, 33);
|
198
|
+
return k2 * C1_128;
|
199
|
+
}
|
200
|
+
|
201
|
+
static void MurmurHash3_x64_128 ( const void * key, const int len,
|
202
|
+
const uint32_t seed, void * out )
|
203
|
+
{
|
204
|
+
const uint8_t * data = (const uint8_t*)key;
|
205
|
+
const int nblocks = len / 16;
|
206
|
+
int i;
|
207
|
+
|
208
|
+
uint64_t h1 = seed;
|
209
|
+
uint64_t h2 = seed;
|
210
|
+
uint64_t k1 = 0, k2 = 0;
|
211
|
+
|
212
|
+
/* body */
|
213
|
+
|
214
|
+
const uint64_t * blocks = (const uint64_t *)(data);
|
215
|
+
|
216
|
+
for(i = 0; i < nblocks; i++)
|
217
|
+
{
|
218
|
+
k1 = getblock64(blocks, i*2+0);
|
219
|
+
k2 = getblock64(blocks, i*2+1);
|
220
|
+
|
221
|
+
h1 ^= mmix128_1(k1);
|
222
|
+
h1 = ROTL64(h1,27); h1 += h2; h1 = h1*5+0x52dce729;
|
223
|
+
|
224
|
+
h2 ^= mmix128_2(k2);
|
225
|
+
h2 = ROTL64(h2,31); h2 += h1; h2 = h2*5+0x38495ab5;
|
226
|
+
}
|
227
|
+
|
228
|
+
/* tail */
|
229
|
+
|
230
|
+
data += nblocks*16;
|
231
|
+
k1 = k2 = 0;
|
232
|
+
|
233
|
+
switch(len & 15)
|
234
|
+
{
|
235
|
+
case 15: k2 ^= (uint64_t)(data[14]) << 48;
|
236
|
+
case 14: k2 ^= (uint64_t)(data[13]) << 40;
|
237
|
+
case 13: k2 ^= (uint64_t)(data[12]) << 32;
|
238
|
+
case 12: k2 ^= (uint64_t)(data[11]) << 24;
|
239
|
+
case 11: k2 ^= (uint64_t)(data[10]) << 16;
|
240
|
+
case 10: k2 ^= (uint64_t)(data[ 9]) << 8;
|
241
|
+
case 9: k2 ^= (uint64_t)(data[ 8]) << 0;
|
242
|
+
h2 ^= mmix128_2(k2);
|
243
|
+
|
244
|
+
case 8: k1 ^= (uint64_t)(data[ 7]) << 56;
|
245
|
+
case 7: k1 ^= (uint64_t)(data[ 6]) << 48;
|
246
|
+
case 6: k1 ^= (uint64_t)(data[ 5]) << 40;
|
247
|
+
case 5: k1 ^= (uint64_t)(data[ 4]) << 32;
|
248
|
+
case 4: k1 ^= (uint64_t)(data[ 3]) << 24;
|
249
|
+
case 3: k1 ^= (uint64_t)(data[ 2]) << 16;
|
250
|
+
case 2: k1 ^= (uint64_t)(data[ 1]) << 8;
|
251
|
+
case 1: k1 ^= (uint64_t)(data[ 0]) << 0;
|
252
|
+
h1 ^= mmix128_1(k1);
|
253
|
+
};
|
254
|
+
|
255
|
+
/* finalization */
|
256
|
+
|
257
|
+
h1 ^= len; h2 ^= len;
|
258
|
+
|
259
|
+
h1 += h2;
|
260
|
+
h2 += h1;
|
261
|
+
|
262
|
+
h1 = fmix64(h1);
|
263
|
+
h2 = fmix64(h2);
|
264
|
+
|
265
|
+
h1 += h2;
|
266
|
+
h2 += h1;
|
267
|
+
|
268
|
+
((uint64_t*)out)[0] = h1;
|
269
|
+
((uint64_t*)out)[1] = h2;
|
270
|
+
}
|
271
|
+
|
272
|
+
/* end of MurmurHash3 algorithm */
|
273
|
+
|
274
|
+
static VALUE
|
275
|
+
rb_fmix32(VALUE self, VALUE integer)
|
276
|
+
{
|
277
|
+
uint32_t _int = NUM2UINT(integer);
|
278
|
+
return UINT2NUM(fmix32(_int));
|
279
|
+
}
|
280
|
+
|
281
|
+
static VALUE
|
282
|
+
rb_fmix64(VALUE self, VALUE integer)
|
283
|
+
{
|
284
|
+
#if SIZEOF_LONG == 8
|
285
|
+
uint64_t _int = NUM2ULONG(integer);
|
286
|
+
return ULONG2NUM(fmix64(_int));
|
287
|
+
#else
|
288
|
+
uint64_t _int = NUM2ULL(integer);
|
289
|
+
return ULL2NUM(fmix64(_int));
|
290
|
+
#endif
|
291
|
+
}
|
292
|
+
|
293
|
+
static VALUE
|
294
|
+
rb_murmur3_32_str_hash(int argc, VALUE* argv, VALUE self)
|
295
|
+
{
|
296
|
+
VALUE rstr, rseed;
|
297
|
+
uint32_t result;
|
298
|
+
|
299
|
+
rb_scan_args(argc, argv, "11", &rstr, &rseed);
|
300
|
+
|
301
|
+
result = MurmurHash3_x86_32(RSTRING_PTR(rstr), RSTRING_LEN(rstr), argc == 1 ? 0 : NUM2UINT(rseed));
|
302
|
+
|
303
|
+
return UINT2NUM(result);
|
304
|
+
}
|
305
|
+
|
306
|
+
static VALUE
|
307
|
+
rb_murmur3_32_int32_hash(int argc, VALUE* argv, VALUE self)
|
308
|
+
{
|
309
|
+
VALUE rint, rseed;
|
310
|
+
uint32_t _int;
|
311
|
+
uint32_t result;
|
312
|
+
|
313
|
+
rb_scan_args(argc, argv, "11", &rint, &rseed);
|
314
|
+
_int = NUM2UINT(rint);
|
315
|
+
|
316
|
+
result = MurmurHash3_x86_32(&_int, 4, argc == 1 ? 0 : NUM2UINT(rseed));
|
317
|
+
|
318
|
+
return UINT2NUM(result);
|
319
|
+
}
|
320
|
+
|
321
|
+
static VALUE
|
322
|
+
rb_murmur3_32_int64_hash(int argc, VALUE* argv, VALUE self)
|
323
|
+
{
|
324
|
+
VALUE rint, rseed;
|
325
|
+
uint64_t _int;
|
326
|
+
uint32_t result;
|
327
|
+
|
328
|
+
rb_scan_args(argc, argv, "11", &rint, &rseed);
|
329
|
+
#if SIZEOF_LONG == 8
|
330
|
+
_int = NUM2ULONG(rint);
|
331
|
+
#else
|
332
|
+
_int = NUM2ULL(rint);
|
333
|
+
#endif
|
334
|
+
|
335
|
+
result = MurmurHash3_x86_32(&_int, 8, argc == 1 ? 0 : NUM2UINT(rseed));
|
336
|
+
|
337
|
+
return UINT2NUM(result);
|
338
|
+
}
|
339
|
+
|
340
|
+
#define PREPARE_128_BIT() \
|
341
|
+
VALUE rstr, rseed, ar_result; \
|
342
|
+
uint32_t result[4]; \
|
343
|
+
rb_scan_args(argc, argv, "11", &rstr, &rseed)
|
344
|
+
|
345
|
+
#define SWAP_128_BIT() do { \
|
346
|
+
uint32_t tmp; \
|
347
|
+
tmp = result[0]; \
|
348
|
+
result[0] = result[1]; \
|
349
|
+
result[1] = tmp; \
|
350
|
+
tmp = result[2]; \
|
351
|
+
result[2] = result[3]; \
|
352
|
+
result[3] = tmp; \
|
353
|
+
} while (0)
|
354
|
+
|
355
|
+
#define RETURN_128_BIT() \
|
356
|
+
ar_result = rb_ary_new2(4); \
|
357
|
+
rb_ary_push(ar_result, UINT2NUM(result[0])); \
|
358
|
+
rb_ary_push(ar_result, UINT2NUM(result[1])); \
|
359
|
+
rb_ary_push(ar_result, UINT2NUM(result[2])); \
|
360
|
+
rb_ary_push(ar_result, UINT2NUM(result[3])); \
|
361
|
+
return ar_result
|
362
|
+
|
363
|
+
static VALUE
|
364
|
+
rb_murmur3_128_str_hash(int argc, VALUE* argv, VALUE self)
|
365
|
+
{
|
366
|
+
PREPARE_128_BIT();
|
367
|
+
|
368
|
+
MurmurHash3_x64_128(RSTRING_PTR(rstr), RSTRING_LEN(rstr), argc == 1 ? 0 : NUM2UINT(rseed), result);
|
369
|
+
#if WORDS_BIGENDIAN
|
370
|
+
SWAP_128_BIT();
|
371
|
+
#endif
|
372
|
+
RETURN_128_BIT();
|
373
|
+
}
|
374
|
+
|
375
|
+
static VALUE
|
376
|
+
rb_murmur3_128_int32_hash(int argc, VALUE* argv, VALUE self)
|
377
|
+
{
|
378
|
+
PREPARE_128_BIT();
|
379
|
+
|
380
|
+
{
|
381
|
+
uint32_t _int = NUM2UINT(rstr);
|
382
|
+
MurmurHash3_x64_128(&_int, 4, argc == 1 ? 0 : NUM2UINT(rseed), result);
|
383
|
+
}
|
384
|
+
#if WORDS_BIGENDIAN
|
385
|
+
SWAP_128_BIT();
|
386
|
+
#endif
|
387
|
+
RETURN_128_BIT();
|
388
|
+
}
|
389
|
+
|
390
|
+
static VALUE
|
391
|
+
rb_murmur3_128_int64_hash(int argc, VALUE* argv, VALUE self)
|
392
|
+
{
|
393
|
+
PREPARE_128_BIT();
|
394
|
+
|
395
|
+
{
|
396
|
+
#if SIZEOF_LONG == 8
|
397
|
+
uint64_t _int = NUM2ULONG(rstr);
|
398
|
+
#else
|
399
|
+
uint64_t _int = NUM2ULL(rstr);
|
400
|
+
#endif
|
401
|
+
MurmurHash3_x64_128(&_int, 8, argc == 1 ? 0 : NUM2UINT(rseed), result);
|
402
|
+
}
|
403
|
+
#if WORDS_BIGENDIAN
|
404
|
+
SWAP_128_BIT();
|
405
|
+
#endif
|
406
|
+
RETURN_128_BIT();
|
407
|
+
}
|
408
|
+
|
409
|
+
void
|
410
|
+
Init_native_murmur() {
|
411
|
+
VALUE singleton;
|
412
|
+
VALUE mod_murmur = rb_define_module("MurmurHash3");
|
413
|
+
VALUE mod_murmur32 = rb_define_module_under(mod_murmur, "Native32");
|
414
|
+
VALUE mod_murmur128 = rb_define_module_under(mod_murmur, "Native128");
|
415
|
+
|
416
|
+
rb_define_method(mod_murmur32, "murmur3_32_fmix", rb_fmix32, 1);
|
417
|
+
rb_define_method(mod_murmur32, "murmur3_32_str_hash", rb_murmur3_32_str_hash, -1);
|
418
|
+
rb_define_method(mod_murmur32, "murmur3_32_int32_hash", rb_murmur3_32_int32_hash, -1);
|
419
|
+
rb_define_method(mod_murmur32, "murmur3_32_int64_hash", rb_murmur3_32_int64_hash, -1);
|
420
|
+
|
421
|
+
rb_extend_object(mod_murmur32, mod_murmur32);
|
422
|
+
singleton = rb_singleton_class(mod_murmur32);
|
423
|
+
rb_define_alias(singleton, "fmix", "murmur3_32_fmix");
|
424
|
+
rb_define_alias(singleton, "str_hash", "murmur3_32_str_hash");
|
425
|
+
rb_define_alias(singleton, "int32_hash", "murmur3_32_int32_hash");
|
426
|
+
rb_define_alias(singleton, "int64_hash", "murmur3_32_int64_hash");
|
427
|
+
|
428
|
+
|
429
|
+
rb_define_method(mod_murmur128, "murmur3_128_fmix", rb_fmix64, 1);
|
430
|
+
rb_define_method(mod_murmur128, "murmur3_128_str_hash", rb_murmur3_128_str_hash, -1);
|
431
|
+
rb_define_method(mod_murmur128, "murmur3_128_int32_hash", rb_murmur3_128_int32_hash, -1);
|
432
|
+
rb_define_method(mod_murmur128, "murmur3_128_int64_hash", rb_murmur3_128_int64_hash, -1);
|
433
|
+
|
434
|
+
rb_extend_object(mod_murmur128, mod_murmur128);
|
435
|
+
singleton = rb_singleton_class(mod_murmur128);
|
436
|
+
rb_define_alias(singleton, "fmix", "murmur3_128_fmix");
|
437
|
+
rb_define_alias(singleton, "str_hash", "murmur3_128_str_hash");
|
438
|
+
rb_define_alias(singleton, "int32_hash", "murmur3_128_int32_hash");
|
439
|
+
rb_define_alias(singleton, "int64_hash", "murmur3_128_int64_hash");
|
440
|
+
|
441
|
+
}
|
data/lib/murmurhash3.rb
ADDED
@@ -0,0 +1,16 @@
|
|
1
|
+
require "murmurhash3/version"
|
2
|
+
|
3
|
+
module MurmurHash3
|
4
|
+
begin
|
5
|
+
require 'murmurhash3/native_murmur'
|
6
|
+
V32 = Native32
|
7
|
+
V128 = Native128
|
8
|
+
rescue LoadError
|
9
|
+
require 'murmurhash3/pure_ruby'
|
10
|
+
if RUBY_ENGINE == 'ruby'
|
11
|
+
$stderr.puts "Attention: used pure ruby version of MurmurHash3"
|
12
|
+
end
|
13
|
+
V32 = PureRuby32
|
14
|
+
V128 = PureRuby128
|
15
|
+
end
|
16
|
+
end
|
@@ -0,0 +1,155 @@
|
|
1
|
+
require 'digest'
|
2
|
+
module MurmurHash3
|
3
|
+
module PureRuby32
|
4
|
+
MASK32 = 0xffffffff
|
5
|
+
def murmur3_32_rotl(x, r)
|
6
|
+
((x << r) | (x >> (32 - r))) & MASK32
|
7
|
+
end
|
8
|
+
|
9
|
+
|
10
|
+
def murmur3_32_fmix(h)
|
11
|
+
h &= MASK32
|
12
|
+
h ^= h >> 16
|
13
|
+
h = (h * 0x85ebca6b) & MASK32
|
14
|
+
h ^= h >> 13
|
15
|
+
h = (h * 0xc2b2ae35) & MASK32
|
16
|
+
h ^ (h >> 16)
|
17
|
+
end
|
18
|
+
|
19
|
+
def murmur3_32__mmix(k1)
|
20
|
+
k1 = (k1 * 0xcc9e2d51) & MASK32
|
21
|
+
k1 = murmur3_32_rotl(k1, 15)
|
22
|
+
(k1 * 0x1b873593) & MASK32
|
23
|
+
end
|
24
|
+
|
25
|
+
def murmur3_32_str_hash(str, seed=0)
|
26
|
+
h1 = seed
|
27
|
+
numbers = str.unpack('V*C*')
|
28
|
+
tailn = str.bytesize % 4
|
29
|
+
tail = numbers.slice!(numbers.size - tailn, tailn)
|
30
|
+
for k1 in numbers
|
31
|
+
h1 ^= murmur3_32__mmix(k1)
|
32
|
+
h1 = murmur3_32_rotl(h1, 13)
|
33
|
+
h1 = (h1*5 + 0xe6546b64) & MASK32
|
34
|
+
end
|
35
|
+
|
36
|
+
unless tail.empty?
|
37
|
+
k1 = 0
|
38
|
+
tail.reverse_each do |c1|
|
39
|
+
k1 = (k1 << 8) | c1
|
40
|
+
end
|
41
|
+
h1 ^= murmur3_32__mmix(k1)
|
42
|
+
end
|
43
|
+
|
44
|
+
h1 ^= str.bytesize
|
45
|
+
murmur3_32_fmix(h1)
|
46
|
+
end
|
47
|
+
|
48
|
+
def murmur3_32_int32_hash(i, seed=0)
|
49
|
+
str_hash([i].pack("V"), seed)
|
50
|
+
end
|
51
|
+
|
52
|
+
def murmur3_32_int64_hash(i, seed=0)
|
53
|
+
str_hash([i].pack("Q<"), seed)
|
54
|
+
end
|
55
|
+
|
56
|
+
class << self
|
57
|
+
include MurmurHash3::PureRuby32
|
58
|
+
alias fmix murmur3_32_fmix
|
59
|
+
alias str_hash murmur3_32_str_hash
|
60
|
+
alias int32_hash murmur3_32_int32_hash
|
61
|
+
alias int64_hash murmur3_32_int64_hash
|
62
|
+
end
|
63
|
+
end
|
64
|
+
|
65
|
+
module PureRuby128
|
66
|
+
MASK64 = 0xffff_ffff_ffff_ffff
|
67
|
+
|
68
|
+
def murmur3_128_rotl(x, r)
|
69
|
+
((x << r) | (x >> (64 - r))) & MASK64
|
70
|
+
end
|
71
|
+
|
72
|
+
def murmur3_128_fmix(h)
|
73
|
+
h &= MASK64
|
74
|
+
h ^= h >> 33
|
75
|
+
h = (h * 0xff51afd7_ed558ccd) & MASK64
|
76
|
+
h ^= h >> 33
|
77
|
+
h = (h * 0xc4ceb9fe_1a85ec53) & MASK64
|
78
|
+
h ^ (h >> 33)
|
79
|
+
end
|
80
|
+
|
81
|
+
C1_128 = 0x87c37b91_114253d5
|
82
|
+
C2_128 = 0x4cf5ad43_2745937f
|
83
|
+
def murmur3_128__mmix1(k1)
|
84
|
+
k1 = (k1 * C1_128) & MASK64
|
85
|
+
k1 = murmur3_128_rotl(k1, 31)
|
86
|
+
(k1 * C2_128) & MASK64
|
87
|
+
end
|
88
|
+
|
89
|
+
def murmur3_128__mmix2(k2)
|
90
|
+
k2 = (k2 * C2_128) & MASK64
|
91
|
+
k2 = murmur3_128_rotl(k2, 33)
|
92
|
+
(k2 * C1_128) & MASK64
|
93
|
+
end
|
94
|
+
|
95
|
+
def murmur3_128_str_hash(str, seed=0)
|
96
|
+
h1 = h2 = seed
|
97
|
+
fast_part = ((str.bytesize / 16) * 16)
|
98
|
+
numbers = str.byteslice(0, fast_part).unpack('Q<*')
|
99
|
+
tail = str.byteslice(fast_part, str.bytesize - fast_part).unpack('C*')
|
100
|
+
|
101
|
+
numbers.each_slice(2) do |k1, k2|
|
102
|
+
h1 ^= murmur3_128__mmix1(k1)
|
103
|
+
h1 = murmur3_128_rotl(h1, 27)
|
104
|
+
h1 = (h1 + h2) & MASK64
|
105
|
+
h1 = (h1*5 + 0x52dce729) & MASK64
|
106
|
+
h2 ^= murmur3_128__mmix2(k2)
|
107
|
+
h2 = murmur3_128_rotl(h2, 31)
|
108
|
+
h2 = (h1 + h2) & MASK64
|
109
|
+
h2 = (h2*5 + 0x38495ab5) & MASK64
|
110
|
+
end
|
111
|
+
|
112
|
+
unless tail.empty?
|
113
|
+
if tail.size > 8
|
114
|
+
k2 = 0
|
115
|
+
tail[8,8].reverse_each do |c2|
|
116
|
+
k2 = (k2 << 8) | c2
|
117
|
+
end
|
118
|
+
h2 ^= murmur3_128__mmix2(k2)
|
119
|
+
end
|
120
|
+
k1 = 0
|
121
|
+
tail[0,8].reverse_each do |c1|
|
122
|
+
k1 = (k1 << 8) | c1
|
123
|
+
end
|
124
|
+
h1 ^= murmur3_128__mmix1(k1)
|
125
|
+
end
|
126
|
+
|
127
|
+
h1 ^= str.bytesize
|
128
|
+
h2 ^= str.bytesize
|
129
|
+
h1 = (h1 + h2) & MASK64
|
130
|
+
h2 = (h1 + h2) & MASK64
|
131
|
+
h1 = murmur3_128_fmix(h1)
|
132
|
+
h2 = murmur3_128_fmix(h2)
|
133
|
+
|
134
|
+
h1 = (h1 + h2) & MASK64
|
135
|
+
h2 = (h1 + h2) & MASK64
|
136
|
+
[h1 & 0xffffffff, h1 >> 32, h2 & 0xffffffff, h2 >> 32]
|
137
|
+
end
|
138
|
+
|
139
|
+
def murmur3_128_int32_hash(i, seed=0)
|
140
|
+
str_hash([i].pack("V"), seed)
|
141
|
+
end
|
142
|
+
|
143
|
+
def murmur3_128_int64_hash(i, seed=0)
|
144
|
+
str_hash([i].pack("Q<"), seed)
|
145
|
+
end
|
146
|
+
|
147
|
+
class << self
|
148
|
+
include MurmurHash3::PureRuby128
|
149
|
+
alias fmix murmur3_128_fmix
|
150
|
+
alias str_hash murmur3_128_str_hash
|
151
|
+
alias int32_hash murmur3_128_int32_hash
|
152
|
+
alias int64_hash murmur3_128_int64_hash
|
153
|
+
end
|
154
|
+
end
|
155
|
+
end
|
data/test/test_murmur.rb
ADDED
@@ -0,0 +1,76 @@
|
|
1
|
+
require 'minitest/spec'
|
2
|
+
require 'minitest/autorun'
|
3
|
+
|
4
|
+
shared_examples_128 = proc do
|
5
|
+
it 'should make correct hash for string' do
|
6
|
+
murmur.str_hash('asdfqwer', 0).must_equal [0xd6d7d367, 0xcb41f064, 0x8973cd72, 0xc345e72e]
|
7
|
+
murmur.str_hash('asdfqwerzxcvyui', 0).must_equal [0x007b2172f, 0x64ecae1b, 0x1813b5a5, 0x9c674ee6]
|
8
|
+
murmur.str_hash('asdfqwerzxcvyuio', 0).must_equal [0xf508df57, 0xbb38f3fd, 0xf48c9d98, 0xb65c36cd]
|
9
|
+
murmur.str_hash('asdfqwerzxcvyuio!', 0).must_equal [0x8a011755, 0xb13d463f, 0x8386d32a, 0x0df8884c]
|
10
|
+
end
|
11
|
+
|
12
|
+
it 'should make correct hash for 32bit integer' do
|
13
|
+
murmur.int32_hash(1717859169).must_equal [0x20b48108, 0x10369ceb, 0x3ad523cc, 0xdacb587f]
|
14
|
+
murmur.int32_hash(1717859169).must_equal murmur.str_hash('asdf')
|
15
|
+
end
|
16
|
+
|
17
|
+
it 'should make correct hash for 64bit integer' do
|
18
|
+
murmur.int64_hash(0x12345678).must_equal murmur.str_hash("\x78\x56\x34\x12\x00\x00\x00\x00")
|
19
|
+
murmur.int64_hash(0x1234567812345678).must_equal murmur.str_hash("\x78\x56\x34\x12\x78\x56\x34\x12")
|
20
|
+
end
|
21
|
+
|
22
|
+
it 'should make correct fmix for 64bit integer' do
|
23
|
+
murmur.fmix(1717859169).must_equal 0xbefb9076a3712207
|
24
|
+
murmur.fmix(12345678912345678).must_equal 0x197ef59146f5221c
|
25
|
+
end
|
26
|
+
end
|
27
|
+
|
28
|
+
shared_examples_32 = proc do
|
29
|
+
it 'should make correct hash for string' do
|
30
|
+
murmur.str_hash('asdfqwer', 0).must_equal 0xa46b5209
|
31
|
+
murmur.str_hash('asdfqwerty', 0).must_equal 0xa3cfe04b
|
32
|
+
murmur.str_hash('asd', 0).must_equal 0x14570c6f
|
33
|
+
end
|
34
|
+
|
35
|
+
it 'should make correct hash for 32bit integer' do
|
36
|
+
murmur.int32_hash(1717859169).must_equal 0x1b20e026
|
37
|
+
murmur.int32_hash(1717859169).must_equal murmur.str_hash('asdf')
|
38
|
+
end
|
39
|
+
|
40
|
+
it 'should make correct hash for 64bit integer' do
|
41
|
+
murmur.int64_hash(0x12345678).must_equal murmur.str_hash("\x78\x56\x34\x12\x00\x00\x00\x00")
|
42
|
+
murmur.int64_hash(0x1234567812345678).must_equal murmur.str_hash("\x78\x56\x34\x12\x78\x56\x34\x12")
|
43
|
+
end
|
44
|
+
|
45
|
+
it 'should make correct fmix for 32bit integer' do
|
46
|
+
murmur.fmix(1717859169).must_equal 0x17561734
|
47
|
+
end
|
48
|
+
end
|
49
|
+
|
50
|
+
require 'murmurhash3/pure_ruby'
|
51
|
+
describe "Pure ruby 32" do
|
52
|
+
let(:murmur) { MurmurHash3::PureRuby32 }
|
53
|
+
class_exec &shared_examples_32
|
54
|
+
end
|
55
|
+
|
56
|
+
describe "Pure ruby 128" do
|
57
|
+
let(:murmur) { MurmurHash3::PureRuby128 }
|
58
|
+
class_exec &shared_examples_128
|
59
|
+
end
|
60
|
+
|
61
|
+
begin
|
62
|
+
require 'murmurhash3/native_murmur'
|
63
|
+
|
64
|
+
describe "Native 32" do
|
65
|
+
let(:murmur) { MurmurHash3::Native32 }
|
66
|
+
class_exec &shared_examples_32
|
67
|
+
end
|
68
|
+
|
69
|
+
describe "Native 128" do
|
70
|
+
let(:murmur) { MurmurHash3::Native128 }
|
71
|
+
class_exec &shared_examples_128
|
72
|
+
end
|
73
|
+
|
74
|
+
rescue LoadError => e
|
75
|
+
puts "Could not load native extension: #{e}"
|
76
|
+
end
|
metadata
ADDED
@@ -0,0 +1,54 @@
|
|
1
|
+
--- !ruby/object:Gem::Specification
|
2
|
+
name: murmurhash3
|
3
|
+
version: !ruby/object:Gem::Version
|
4
|
+
version: 0.1.1
|
5
|
+
prerelease:
|
6
|
+
platform: ruby
|
7
|
+
authors:
|
8
|
+
- Sokolov Yura 'funny-falcon'
|
9
|
+
autorequire:
|
10
|
+
bindir: bin
|
11
|
+
cert_chain: []
|
12
|
+
date: 2012-08-02 00:00:00.000000000 Z
|
13
|
+
dependencies: []
|
14
|
+
description: implementation of murmur3 hashing function
|
15
|
+
email:
|
16
|
+
- funny.falcon@gmail.com
|
17
|
+
executables: []
|
18
|
+
extensions:
|
19
|
+
- ext/murmurhash3/extconf.rb
|
20
|
+
extra_rdoc_files: []
|
21
|
+
files:
|
22
|
+
- ext/murmurhash3/extconf.rb
|
23
|
+
- ext/murmurhash3/murmur3.c
|
24
|
+
- lib/murmurhash3/pure_ruby.rb
|
25
|
+
- lib/murmurhash3/version.rb
|
26
|
+
- lib/murmurhash3.rb
|
27
|
+
- test/test_murmur.rb
|
28
|
+
homepage: https://github.com/funny-falcon/murmurhash3
|
29
|
+
licenses: []
|
30
|
+
post_install_message:
|
31
|
+
rdoc_options: []
|
32
|
+
require_paths:
|
33
|
+
- lib
|
34
|
+
- ext
|
35
|
+
required_ruby_version: !ruby/object:Gem::Requirement
|
36
|
+
none: false
|
37
|
+
requirements:
|
38
|
+
- - ! '>='
|
39
|
+
- !ruby/object:Gem::Version
|
40
|
+
version: '0'
|
41
|
+
required_rubygems_version: !ruby/object:Gem::Requirement
|
42
|
+
none: false
|
43
|
+
requirements:
|
44
|
+
- - ! '>='
|
45
|
+
- !ruby/object:Gem::Version
|
46
|
+
version: '0'
|
47
|
+
requirements: []
|
48
|
+
rubyforge_project:
|
49
|
+
rubygems_version: 1.8.24
|
50
|
+
signing_key:
|
51
|
+
specification_version: 3
|
52
|
+
summary: implements mumur3 hashing function
|
53
|
+
test_files:
|
54
|
+
- test/test_murmur.rb
|