Wed, 31 Dec 2014 06:09:35 +0100
Cloned upstream origin tor-browser at tor-browser-31.3.0esr-4.5-1-build1
revision ID fc1c9ff7c1b2defdbc039f12214767608f46423f for hacking purpose.
michael@0 | 1 | /* |
michael@0 | 2 | * Copyright © 2012 Google, Inc. |
michael@0 | 3 | * |
michael@0 | 4 | * This is part of HarfBuzz, a text shaping library. |
michael@0 | 5 | * |
michael@0 | 6 | * Permission is hereby granted, without written agreement and without |
michael@0 | 7 | * license or royalty fees, to use, copy, modify, and distribute this |
michael@0 | 8 | * software and its documentation for any purpose, provided that the |
michael@0 | 9 | * above copyright notice and the following two paragraphs appear in |
michael@0 | 10 | * all copies of this software. |
michael@0 | 11 | * |
michael@0 | 12 | * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR |
michael@0 | 13 | * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES |
michael@0 | 14 | * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN |
michael@0 | 15 | * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH |
michael@0 | 16 | * DAMAGE. |
michael@0 | 17 | * |
michael@0 | 18 | * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING, |
michael@0 | 19 | * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND |
michael@0 | 20 | * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS |
michael@0 | 21 | * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO |
michael@0 | 22 | * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS. |
michael@0 | 23 | * |
michael@0 | 24 | * Google Author(s): Behdad Esfahbod |
michael@0 | 25 | */ |
michael@0 | 26 | |
michael@0 | 27 | #ifndef HB_SET_PRIVATE_HH |
michael@0 | 28 | #define HB_SET_PRIVATE_HH |
michael@0 | 29 | |
michael@0 | 30 | #include "hb-private.hh" |
michael@0 | 31 | #include "hb-object-private.hh" |
michael@0 | 32 | |
michael@0 | 33 | |
michael@0 | 34 | /* |
michael@0 | 35 | * The set digests here implement various "filters" that support |
michael@0 | 36 | * "approximate member query". Conceptually these are like Bloom |
michael@0 | 37 | * Filter and Quotient Filter, however, much smaller, faster, and |
michael@0 | 38 | * designed to fit the requirements of our uses for glyph coverage |
michael@0 | 39 | * queries. As a result, our filters have much higher. |
michael@0 | 40 | */ |
michael@0 | 41 | |
michael@0 | 42 | template <typename mask_t, unsigned int shift> |
michael@0 | 43 | struct hb_set_digest_lowest_bits_t |
michael@0 | 44 | { |
michael@0 | 45 | ASSERT_POD (); |
michael@0 | 46 | |
michael@0 | 47 | static const unsigned int mask_bytes = sizeof (mask_t); |
michael@0 | 48 | static const unsigned int mask_bits = sizeof (mask_t) * 8; |
michael@0 | 49 | static const unsigned int num_bits = 0 |
michael@0 | 50 | + (mask_bytes >= 1 ? 3 : 0) |
michael@0 | 51 | + (mask_bytes >= 2 ? 1 : 0) |
michael@0 | 52 | + (mask_bytes >= 4 ? 1 : 0) |
michael@0 | 53 | + (mask_bytes >= 8 ? 1 : 0) |
michael@0 | 54 | + (mask_bytes >= 16? 1 : 0) |
michael@0 | 55 | + 0; |
michael@0 | 56 | |
michael@0 | 57 | ASSERT_STATIC (shift < sizeof (hb_codepoint_t) * 8); |
michael@0 | 58 | ASSERT_STATIC (shift + num_bits <= sizeof (hb_codepoint_t) * 8); |
michael@0 | 59 | |
michael@0 | 60 | inline void init (void) { |
michael@0 | 61 | mask = 0; |
michael@0 | 62 | } |
michael@0 | 63 | |
michael@0 | 64 | inline void add (hb_codepoint_t g) { |
michael@0 | 65 | mask |= mask_for (g); |
michael@0 | 66 | } |
michael@0 | 67 | |
michael@0 | 68 | inline void add_range (hb_codepoint_t a, hb_codepoint_t b) { |
michael@0 | 69 | if ((b >> shift) - (a >> shift) >= mask_bits - 1) |
michael@0 | 70 | mask = (mask_t) -1; |
michael@0 | 71 | else { |
michael@0 | 72 | mask_t ma = mask_for (a); |
michael@0 | 73 | mask_t mb = mask_for (b); |
michael@0 | 74 | mask |= mb + (mb - ma) - (mb < ma); |
michael@0 | 75 | } |
michael@0 | 76 | } |
michael@0 | 77 | |
michael@0 | 78 | inline bool may_have (hb_codepoint_t g) const { |
michael@0 | 79 | return !!(mask & mask_for (g)); |
michael@0 | 80 | } |
michael@0 | 81 | |
michael@0 | 82 | private: |
michael@0 | 83 | |
michael@0 | 84 | static inline mask_t mask_for (hb_codepoint_t g) { |
michael@0 | 85 | return ((mask_t) 1) << ((g >> shift) & (mask_bits - 1)); |
michael@0 | 86 | } |
michael@0 | 87 | mask_t mask; |
michael@0 | 88 | }; |
michael@0 | 89 | |
michael@0 | 90 | template <typename head_t, typename tail_t> |
michael@0 | 91 | struct hb_set_digest_combiner_t |
michael@0 | 92 | { |
michael@0 | 93 | ASSERT_POD (); |
michael@0 | 94 | |
michael@0 | 95 | inline void init (void) { |
michael@0 | 96 | head.init (); |
michael@0 | 97 | tail.init (); |
michael@0 | 98 | } |
michael@0 | 99 | |
michael@0 | 100 | inline void add (hb_codepoint_t g) { |
michael@0 | 101 | head.add (g); |
michael@0 | 102 | tail.add (g); |
michael@0 | 103 | } |
michael@0 | 104 | |
michael@0 | 105 | inline void add_range (hb_codepoint_t a, hb_codepoint_t b) { |
michael@0 | 106 | head.add_range (a, b); |
michael@0 | 107 | tail.add_range (a, b); |
michael@0 | 108 | } |
michael@0 | 109 | |
michael@0 | 110 | inline bool may_have (hb_codepoint_t g) const { |
michael@0 | 111 | return head.may_have (g) && tail.may_have (g); |
michael@0 | 112 | } |
michael@0 | 113 | |
michael@0 | 114 | private: |
michael@0 | 115 | head_t head; |
michael@0 | 116 | tail_t tail; |
michael@0 | 117 | }; |
michael@0 | 118 | |
michael@0 | 119 | |
michael@0 | 120 | /* |
michael@0 | 121 | * hb_set_digest_t |
michael@0 | 122 | * |
michael@0 | 123 | * This is a combination of digests that performs "best". |
michael@0 | 124 | * There is not much science to this: it's a result of intuition |
michael@0 | 125 | * and testing. |
michael@0 | 126 | */ |
michael@0 | 127 | typedef hb_set_digest_combiner_t |
michael@0 | 128 | < |
michael@0 | 129 | hb_set_digest_lowest_bits_t<unsigned long, 4>, |
michael@0 | 130 | hb_set_digest_combiner_t |
michael@0 | 131 | < |
michael@0 | 132 | hb_set_digest_lowest_bits_t<unsigned long, 0>, |
michael@0 | 133 | hb_set_digest_lowest_bits_t<unsigned long, 9> |
michael@0 | 134 | > |
michael@0 | 135 | > hb_set_digest_t; |
michael@0 | 136 | |
michael@0 | 137 | |
michael@0 | 138 | |
michael@0 | 139 | /* |
michael@0 | 140 | * hb_set_t |
michael@0 | 141 | */ |
michael@0 | 142 | |
michael@0 | 143 | |
michael@0 | 144 | /* TODO Make this faster and memmory efficient. */ |
michael@0 | 145 | |
michael@0 | 146 | struct hb_set_t |
michael@0 | 147 | { |
michael@0 | 148 | hb_object_header_t header; |
michael@0 | 149 | ASSERT_POD (); |
michael@0 | 150 | bool in_error; |
michael@0 | 151 | |
michael@0 | 152 | inline void init (void) { |
michael@0 | 153 | header.init (); |
michael@0 | 154 | clear (); |
michael@0 | 155 | } |
michael@0 | 156 | inline void fini (void) { |
michael@0 | 157 | } |
michael@0 | 158 | inline void clear (void) { |
michael@0 | 159 | if (unlikely (hb_object_is_inert (this))) |
michael@0 | 160 | return; |
michael@0 | 161 | in_error = false; |
michael@0 | 162 | memset (elts, 0, sizeof elts); |
michael@0 | 163 | } |
michael@0 | 164 | inline bool is_empty (void) const { |
michael@0 | 165 | for (unsigned int i = 0; i < ARRAY_LENGTH (elts); i++) |
michael@0 | 166 | if (elts[i]) |
michael@0 | 167 | return false; |
michael@0 | 168 | return true; |
michael@0 | 169 | } |
michael@0 | 170 | inline void add (hb_codepoint_t g) |
michael@0 | 171 | { |
michael@0 | 172 | if (unlikely (in_error)) return; |
michael@0 | 173 | if (unlikely (g == INVALID)) return; |
michael@0 | 174 | if (unlikely (g > MAX_G)) return; |
michael@0 | 175 | elt (g) |= mask (g); |
michael@0 | 176 | } |
michael@0 | 177 | inline void add_range (hb_codepoint_t a, hb_codepoint_t b) |
michael@0 | 178 | { |
michael@0 | 179 | if (unlikely (in_error)) return; |
michael@0 | 180 | /* TODO Speedup */ |
michael@0 | 181 | for (unsigned int i = a; i < b + 1; i++) |
michael@0 | 182 | add (i); |
michael@0 | 183 | } |
michael@0 | 184 | inline void del (hb_codepoint_t g) |
michael@0 | 185 | { |
michael@0 | 186 | if (unlikely (in_error)) return; |
michael@0 | 187 | if (unlikely (g > MAX_G)) return; |
michael@0 | 188 | elt (g) &= ~mask (g); |
michael@0 | 189 | } |
michael@0 | 190 | inline void del_range (hb_codepoint_t a, hb_codepoint_t b) |
michael@0 | 191 | { |
michael@0 | 192 | if (unlikely (in_error)) return; |
michael@0 | 193 | /* TODO Speedup */ |
michael@0 | 194 | for (unsigned int i = a; i < b + 1; i++) |
michael@0 | 195 | del (i); |
michael@0 | 196 | } |
michael@0 | 197 | inline bool has (hb_codepoint_t g) const |
michael@0 | 198 | { |
michael@0 | 199 | if (unlikely (g > MAX_G)) return false; |
michael@0 | 200 | return !!(elt (g) & mask (g)); |
michael@0 | 201 | } |
michael@0 | 202 | inline bool intersects (hb_codepoint_t first, |
michael@0 | 203 | hb_codepoint_t last) const |
michael@0 | 204 | { |
michael@0 | 205 | if (unlikely (first > MAX_G)) return false; |
michael@0 | 206 | if (unlikely (last > MAX_G)) last = MAX_G; |
michael@0 | 207 | unsigned int end = last + 1; |
michael@0 | 208 | for (hb_codepoint_t i = first; i < end; i++) |
michael@0 | 209 | if (has (i)) |
michael@0 | 210 | return true; |
michael@0 | 211 | return false; |
michael@0 | 212 | } |
michael@0 | 213 | inline bool is_equal (const hb_set_t *other) const |
michael@0 | 214 | { |
michael@0 | 215 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 216 | if (elts[i] != other->elts[i]) |
michael@0 | 217 | return false; |
michael@0 | 218 | return true; |
michael@0 | 219 | } |
michael@0 | 220 | inline void set (const hb_set_t *other) |
michael@0 | 221 | { |
michael@0 | 222 | if (unlikely (in_error)) return; |
michael@0 | 223 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 224 | elts[i] = other->elts[i]; |
michael@0 | 225 | } |
michael@0 | 226 | inline void union_ (const hb_set_t *other) |
michael@0 | 227 | { |
michael@0 | 228 | if (unlikely (in_error)) return; |
michael@0 | 229 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 230 | elts[i] |= other->elts[i]; |
michael@0 | 231 | } |
michael@0 | 232 | inline void intersect (const hb_set_t *other) |
michael@0 | 233 | { |
michael@0 | 234 | if (unlikely (in_error)) return; |
michael@0 | 235 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 236 | elts[i] &= other->elts[i]; |
michael@0 | 237 | } |
michael@0 | 238 | inline void subtract (const hb_set_t *other) |
michael@0 | 239 | { |
michael@0 | 240 | if (unlikely (in_error)) return; |
michael@0 | 241 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 242 | elts[i] &= ~other->elts[i]; |
michael@0 | 243 | } |
michael@0 | 244 | inline void symmetric_difference (const hb_set_t *other) |
michael@0 | 245 | { |
michael@0 | 246 | if (unlikely (in_error)) return; |
michael@0 | 247 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 248 | elts[i] ^= other->elts[i]; |
michael@0 | 249 | } |
michael@0 | 250 | inline void invert (void) |
michael@0 | 251 | { |
michael@0 | 252 | if (unlikely (in_error)) return; |
michael@0 | 253 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 254 | elts[i] = ~elts[i]; |
michael@0 | 255 | } |
michael@0 | 256 | inline bool next (hb_codepoint_t *codepoint) const |
michael@0 | 257 | { |
michael@0 | 258 | if (unlikely (*codepoint == INVALID)) { |
michael@0 | 259 | hb_codepoint_t i = get_min (); |
michael@0 | 260 | if (i != INVALID) { |
michael@0 | 261 | *codepoint = i; |
michael@0 | 262 | return true; |
michael@0 | 263 | } else { |
michael@0 | 264 | *codepoint = INVALID; |
michael@0 | 265 | return false; |
michael@0 | 266 | } |
michael@0 | 267 | } |
michael@0 | 268 | for (hb_codepoint_t i = *codepoint + 1; i < MAX_G + 1; i++) |
michael@0 | 269 | if (has (i)) { |
michael@0 | 270 | *codepoint = i; |
michael@0 | 271 | return true; |
michael@0 | 272 | } |
michael@0 | 273 | *codepoint = INVALID; |
michael@0 | 274 | return false; |
michael@0 | 275 | } |
michael@0 | 276 | inline bool next_range (hb_codepoint_t *first, hb_codepoint_t *last) const |
michael@0 | 277 | { |
michael@0 | 278 | hb_codepoint_t i; |
michael@0 | 279 | |
michael@0 | 280 | i = *last; |
michael@0 | 281 | if (!next (&i)) |
michael@0 | 282 | { |
michael@0 | 283 | *last = *first = INVALID; |
michael@0 | 284 | return false; |
michael@0 | 285 | } |
michael@0 | 286 | |
michael@0 | 287 | *last = *first = i; |
michael@0 | 288 | while (next (&i) && i == *last + 1) |
michael@0 | 289 | (*last)++; |
michael@0 | 290 | |
michael@0 | 291 | return true; |
michael@0 | 292 | } |
michael@0 | 293 | |
michael@0 | 294 | inline unsigned int get_population (void) const |
michael@0 | 295 | { |
michael@0 | 296 | unsigned int count = 0; |
michael@0 | 297 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 298 | count += _hb_popcount32 (elts[i]); |
michael@0 | 299 | return count; |
michael@0 | 300 | } |
michael@0 | 301 | inline hb_codepoint_t get_min (void) const |
michael@0 | 302 | { |
michael@0 | 303 | for (unsigned int i = 0; i < ELTS; i++) |
michael@0 | 304 | if (elts[i]) |
michael@0 | 305 | for (unsigned int j = 0; j < BITS; j++) |
michael@0 | 306 | if (elts[i] & (1 << j)) |
michael@0 | 307 | return i * BITS + j; |
michael@0 | 308 | return INVALID; |
michael@0 | 309 | } |
michael@0 | 310 | inline hb_codepoint_t get_max (void) const |
michael@0 | 311 | { |
michael@0 | 312 | for (unsigned int i = ELTS; i; i--) |
michael@0 | 313 | if (elts[i - 1]) |
michael@0 | 314 | for (unsigned int j = BITS; j; j--) |
michael@0 | 315 | if (elts[i - 1] & (1 << (j - 1))) |
michael@0 | 316 | return (i - 1) * BITS + (j - 1); |
michael@0 | 317 | return INVALID; |
michael@0 | 318 | } |
michael@0 | 319 | |
michael@0 | 320 | typedef uint32_t elt_t; |
michael@0 | 321 | static const unsigned int MAX_G = 65536 - 1; /* XXX Fix this... */ |
michael@0 | 322 | static const unsigned int SHIFT = 5; |
michael@0 | 323 | static const unsigned int BITS = (1 << SHIFT); |
michael@0 | 324 | static const unsigned int MASK = BITS - 1; |
michael@0 | 325 | static const unsigned int ELTS = (MAX_G + 1 + (BITS - 1)) / BITS; |
michael@0 | 326 | static const hb_codepoint_t INVALID = HB_SET_VALUE_INVALID; |
michael@0 | 327 | |
michael@0 | 328 | elt_t &elt (hb_codepoint_t g) { return elts[g >> SHIFT]; } |
michael@0 | 329 | elt_t elt (hb_codepoint_t g) const { return elts[g >> SHIFT]; } |
michael@0 | 330 | elt_t mask (hb_codepoint_t g) const { return elt_t (1) << (g & MASK); } |
michael@0 | 331 | |
michael@0 | 332 | elt_t elts[ELTS]; /* XXX 8kb */ |
michael@0 | 333 | |
michael@0 | 334 | ASSERT_STATIC (sizeof (elt_t) * 8 == BITS); |
michael@0 | 335 | ASSERT_STATIC (sizeof (elt_t) * 8 * ELTS > MAX_G); |
michael@0 | 336 | }; |
michael@0 | 337 | |
michael@0 | 338 | |
michael@0 | 339 | |
michael@0 | 340 | #endif /* HB_SET_PRIVATE_HH */ |