Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 1 | // Copyright 2010 the V8 project authors. All rights reserved. |
| 2 | // Redistribution and use in source and binary forms, with or without |
| 3 | // modification, are permitted provided that the following conditions are |
| 4 | // met: |
| 5 | // |
| 6 | // * Redistributions of source code must retain the above copyright |
| 7 | // notice, this list of conditions and the following disclaimer. |
| 8 | // * Redistributions in binary form must reproduce the above |
| 9 | // copyright notice, this list of conditions and the following |
| 10 | // disclaimer in the documentation and/or other materials provided |
| 11 | // with the distribution. |
| 12 | // * Neither the name of Google Inc. nor the names of its |
| 13 | // contributors may be used to endorse or promote products derived |
| 14 | // from this software without specific prior written permission. |
| 15 | // |
| 16 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 17 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 18 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 19 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 20 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 21 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 22 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 23 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 24 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 25 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 26 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 27 | |
| 28 | #ifndef V8_STRING_SEARCH_H_ |
| 29 | #define V8_STRING_SEARCH_H_ |
| 30 | |
| 31 | namespace v8 { |
| 32 | namespace internal { |
| 33 | |
| 34 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 35 | //--------------------------------------------------------------------- |
| 36 | // String Search object. |
| 37 | //--------------------------------------------------------------------- |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 38 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 39 | // Class holding constants and methods that apply to all string search variants, |
| 40 | // independently of subject and pattern char size. |
| 41 | class StringSearchBase { |
| 42 | protected: |
| 43 | // Cap on the maximal shift in the Boyer-Moore implementation. By setting a |
| 44 | // limit, we can fix the size of tables. For a needle longer than this limit, |
| 45 | // search will not be optimal, since we only build tables for a suffix |
| 46 | // of the string, but it is a safe approximation. |
| 47 | static const int kBMMaxShift = 250; |
| 48 | |
| 49 | // Reduce alphabet to this size. |
| 50 | // One of the tables used by Boyer-Moore and Boyer-Moore-Horspool has size |
| 51 | // proportional to the input alphabet. We reduce the alphabet size by |
| 52 | // equating input characters modulo a smaller alphabet size. This gives |
| 53 | // a potentially less efficient searching, but is a safe approximation. |
| 54 | // For needles using only characters in the same Unicode 256-code point page, |
| 55 | // there is no search speed degradation. |
| 56 | static const int kAsciiAlphabetSize = 128; |
| 57 | static const int kUC16AlphabetSize = 256; |
| 58 | |
| 59 | // Bad-char shift table stored in the state. It's length is the alphabet size. |
| 60 | // For patterns below this length, the skip length of Boyer-Moore is too short |
| 61 | // to compensate for the algorithmic overhead compared to simple brute force. |
| 62 | static const int kBMMinPatternLength = 7; |
| 63 | |
| 64 | static inline bool IsAsciiString(Vector<const char>) { |
| 65 | return true; |
| 66 | } |
| 67 | |
| 68 | static inline bool IsAsciiString(Vector<const uc16> string) { |
Steve Block | 9fac840 | 2011-05-12 15:51:54 +0100 | [diff] [blame] | 69 | return String::IsAscii(string.start(), string.length()); |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 70 | } |
| 71 | |
| 72 | // The following tables are shared by all searches. |
| 73 | // TODO(lrn): Introduce a way for a pattern to keep its tables |
| 74 | // between searches (e.g., for an Atom RegExp). |
| 75 | |
| 76 | // Store for the BoyerMoore(Horspool) bad char shift table. |
| 77 | static int kBadCharShiftTable[kUC16AlphabetSize]; |
| 78 | // Store for the BoyerMoore good suffix shift table. |
| 79 | static int kGoodSuffixShiftTable[kBMMaxShift + 1]; |
| 80 | // Table used temporarily while building the BoyerMoore good suffix |
| 81 | // shift table. |
| 82 | static int kSuffixTable[kBMMaxShift + 1]; |
| 83 | }; |
| 84 | |
| 85 | |
| 86 | template <typename PatternChar, typename SubjectChar> |
| 87 | class StringSearch : private StringSearchBase { |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 88 | public: |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 89 | explicit StringSearch(Vector<const PatternChar> pattern) |
| 90 | : pattern_(pattern), |
| 91 | start_(Max(0, pattern.length() - kBMMaxShift)) { |
| 92 | if (sizeof(PatternChar) > sizeof(SubjectChar)) { |
| 93 | if (!IsAsciiString(pattern_)) { |
| 94 | strategy_ = &FailSearch; |
| 95 | return; |
| 96 | } |
| 97 | } |
| 98 | int pattern_length = pattern_.length(); |
| 99 | if (pattern_length < kBMMinPatternLength) { |
| 100 | if (pattern_length == 1) { |
| 101 | strategy_ = &SingleCharSearch; |
| 102 | return; |
| 103 | } |
| 104 | strategy_ = &LinearSearch; |
| 105 | return; |
| 106 | } |
| 107 | strategy_ = &InitialSearch; |
| 108 | } |
| 109 | |
| 110 | int Search(Vector<const SubjectChar> subject, int index) { |
| 111 | return strategy_(this, subject, index); |
| 112 | } |
| 113 | |
| 114 | static inline int AlphabetSize() { |
| 115 | if (sizeof(PatternChar) == 1) { |
| 116 | // ASCII needle. |
| 117 | return kAsciiAlphabetSize; |
| 118 | } else { |
| 119 | ASSERT(sizeof(PatternChar) == 2); |
| 120 | // UC16 needle. |
| 121 | return kUC16AlphabetSize; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 122 | } |
| 123 | } |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 124 | |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 125 | private: |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 126 | typedef int (*SearchFunction)( // NOLINT - it's not a cast! |
| 127 | StringSearch<PatternChar, SubjectChar>*, |
| 128 | Vector<const SubjectChar>, |
| 129 | int); |
| 130 | |
| 131 | static int FailSearch(StringSearch<PatternChar, SubjectChar>*, |
| 132 | Vector<const SubjectChar>, |
| 133 | int) { |
| 134 | return -1; |
| 135 | } |
| 136 | |
| 137 | static int SingleCharSearch(StringSearch<PatternChar, SubjectChar>* search, |
| 138 | Vector<const SubjectChar> subject, |
| 139 | int start_index); |
| 140 | |
| 141 | static int LinearSearch(StringSearch<PatternChar, SubjectChar>* search, |
| 142 | Vector<const SubjectChar> subject, |
| 143 | int start_index); |
| 144 | |
| 145 | static int InitialSearch(StringSearch<PatternChar, SubjectChar>* search, |
| 146 | Vector<const SubjectChar> subject, |
| 147 | int start_index); |
| 148 | |
| 149 | static int BoyerMooreHorspoolSearch( |
| 150 | StringSearch<PatternChar, SubjectChar>* search, |
| 151 | Vector<const SubjectChar> subject, |
| 152 | int start_index); |
| 153 | |
| 154 | static int BoyerMooreSearch(StringSearch<PatternChar, SubjectChar>* search, |
| 155 | Vector<const SubjectChar> subject, |
| 156 | int start_index); |
| 157 | |
| 158 | void PopulateBoyerMooreHorspoolTable(); |
| 159 | |
| 160 | void PopulateBoyerMooreTable(); |
| 161 | |
| 162 | static inline int CharOccurrence(int* bad_char_occurrence, |
| 163 | SubjectChar char_code) { |
| 164 | if (sizeof(SubjectChar) == 1) { |
| 165 | return bad_char_occurrence[static_cast<int>(char_code)]; |
| 166 | } |
| 167 | if (sizeof(PatternChar) == 1) { |
| 168 | if (static_cast<unsigned int>(char_code) > String::kMaxAsciiCharCodeU) { |
| 169 | return -1; |
| 170 | } |
| 171 | return bad_char_occurrence[static_cast<unsigned int>(char_code)]; |
| 172 | } |
| 173 | // Both pattern and subject are UC16. Reduce character to equivalence class. |
| 174 | int equiv_class = char_code % kUC16AlphabetSize; |
| 175 | return bad_char_occurrence[equiv_class]; |
| 176 | } |
| 177 | |
| 178 | // Return a table covering the last kBMMaxShift+1 positions of |
| 179 | // pattern. |
| 180 | int* bad_char_table() { |
| 181 | return kBadCharShiftTable; |
| 182 | } |
| 183 | |
| 184 | int* good_suffix_shift_table() { |
| 185 | // Return biased pointer that maps the range [start_..pattern_.length() |
| 186 | // to the kGoodSuffixShiftTable array. |
| 187 | return kGoodSuffixShiftTable - start_; |
| 188 | } |
| 189 | |
| 190 | int* suffix_table() { |
| 191 | // Return biased pointer that maps the range [start_..pattern_.length() |
| 192 | // to the kSuffixTable array. |
| 193 | return kSuffixTable - start_; |
| 194 | } |
| 195 | |
| 196 | // The pattern to search for. |
| 197 | Vector<const PatternChar> pattern_; |
| 198 | // Pointer to implementation of the search. |
| 199 | SearchFunction strategy_; |
| 200 | // Cache value of Max(0, pattern_length() - kBMMaxShift) |
| 201 | int start_; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 202 | }; |
| 203 | |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 204 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 205 | //--------------------------------------------------------------------- |
| 206 | // Single Character Pattern Search Strategy |
| 207 | //--------------------------------------------------------------------- |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 208 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 209 | template <typename PatternChar, typename SubjectChar> |
| 210 | int StringSearch<PatternChar, SubjectChar>::SingleCharSearch( |
| 211 | StringSearch<PatternChar, SubjectChar>* search, |
| 212 | Vector<const SubjectChar> subject, |
| 213 | int index) { |
| 214 | ASSERT_EQ(1, search->pattern_.length()); |
| 215 | PatternChar pattern_first_char = search->pattern_[0]; |
| 216 | int i = index; |
| 217 | if (sizeof(SubjectChar) == 1 && sizeof(PatternChar) == 1) { |
| 218 | const SubjectChar* pos = reinterpret_cast<const SubjectChar*>( |
| 219 | memchr(subject.start() + i, |
| 220 | pattern_first_char, |
| 221 | subject.length() - i)); |
| 222 | if (pos == NULL) return -1; |
| 223 | return static_cast<int>(pos - subject.start()); |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 224 | } else { |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 225 | if (sizeof(PatternChar) > sizeof(SubjectChar)) { |
| 226 | if (static_cast<uc16>(pattern_first_char) > String::kMaxAsciiCharCodeU) { |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 227 | return -1; |
| 228 | } |
| 229 | } |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 230 | SubjectChar search_char = static_cast<SubjectChar>(pattern_first_char); |
| 231 | int n = subject.length(); |
| 232 | while (i < n) { |
| 233 | if (subject[i++] == search_char) return i - 1; |
| 234 | } |
| 235 | return -1; |
| 236 | } |
| 237 | } |
| 238 | |
| 239 | //--------------------------------------------------------------------- |
| 240 | // Linear Search Strategy |
| 241 | //--------------------------------------------------------------------- |
| 242 | |
| 243 | |
| 244 | template <typename PatternChar, typename SubjectChar> |
| 245 | static inline bool CharCompare(const PatternChar* pattern, |
| 246 | const SubjectChar* subject, |
| 247 | int length) { |
| 248 | ASSERT(length > 0); |
| 249 | int pos = 0; |
| 250 | do { |
| 251 | if (pattern[pos] != subject[pos]) { |
| 252 | return false; |
| 253 | } |
| 254 | pos++; |
| 255 | } while (pos < length); |
| 256 | return true; |
| 257 | } |
| 258 | |
| 259 | |
| 260 | // Simple linear search for short patterns. Never bails out. |
| 261 | template <typename PatternChar, typename SubjectChar> |
| 262 | int StringSearch<PatternChar, SubjectChar>::LinearSearch( |
| 263 | StringSearch<PatternChar, SubjectChar>* search, |
| 264 | Vector<const SubjectChar> subject, |
| 265 | int index) { |
| 266 | Vector<const PatternChar> pattern = search->pattern_; |
| 267 | ASSERT(pattern.length() > 1); |
| 268 | int pattern_length = pattern.length(); |
| 269 | PatternChar pattern_first_char = pattern[0]; |
| 270 | int i = index; |
| 271 | int n = subject.length() - pattern_length; |
| 272 | while (i <= n) { |
| 273 | if (sizeof(SubjectChar) == 1 && sizeof(PatternChar) == 1) { |
| 274 | const SubjectChar* pos = reinterpret_cast<const SubjectChar*>( |
| 275 | memchr(subject.start() + i, |
| 276 | pattern_first_char, |
| 277 | n - i + 1)); |
| 278 | if (pos == NULL) return -1; |
| 279 | i = static_cast<int>(pos - subject.start()) + 1; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 280 | } else { |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 281 | if (subject[i++] != pattern_first_char) continue; |
| 282 | } |
| 283 | // Loop extracted to separate function to allow using return to do |
| 284 | // a deeper break. |
| 285 | if (CharCompare(pattern.start() + 1, |
| 286 | subject.start() + i, |
| 287 | pattern_length - 1)) { |
| 288 | return i - 1; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 289 | } |
| 290 | } |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 291 | return -1; |
| 292 | } |
| 293 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 294 | //--------------------------------------------------------------------- |
| 295 | // Boyer-Moore string search |
| 296 | //--------------------------------------------------------------------- |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 297 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 298 | template <typename PatternChar, typename SubjectChar> |
| 299 | int StringSearch<PatternChar, SubjectChar>::BoyerMooreSearch( |
| 300 | StringSearch<PatternChar, SubjectChar>* search, |
| 301 | Vector<const SubjectChar> subject, |
| 302 | int start_index) { |
| 303 | Vector<const PatternChar> pattern = search->pattern_; |
| 304 | int subject_length = subject.length(); |
| 305 | int pattern_length = pattern.length(); |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 306 | // Only preprocess at most kBMMaxShift last characters of pattern. |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 307 | int start = search->start_; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 308 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 309 | int* bad_char_occurence = search->bad_char_table(); |
| 310 | int* good_suffix_shift = search->good_suffix_shift_table(); |
| 311 | |
| 312 | PatternChar last_char = pattern[pattern_length - 1]; |
| 313 | int index = start_index; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 314 | // Continue search from i. |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 315 | while (index <= subject_length - pattern_length) { |
| 316 | int j = pattern_length - 1; |
| 317 | int c; |
| 318 | while (last_char != (c = subject[index + j])) { |
| 319 | int shift = |
| 320 | j - CharOccurrence(bad_char_occurence, c); |
| 321 | index += shift; |
| 322 | if (index > subject_length - pattern_length) { |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 323 | return -1; |
| 324 | } |
| 325 | } |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 326 | while (j >= 0 && pattern[j] == (c = subject[index + j])) j--; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 327 | if (j < 0) { |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 328 | return index; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 329 | } else if (j < start) { |
| 330 | // we have matched more than our tables allow us to be smart about. |
| 331 | // Fall back on BMH shift. |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 332 | index += pattern_length - 1 |
| 333 | - CharOccurrence(bad_char_occurence, |
| 334 | static_cast<SubjectChar>(last_char)); |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 335 | } else { |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 336 | int gs_shift = good_suffix_shift[j + 1]; |
| 337 | int bc_occ = |
| 338 | CharOccurrence(bad_char_occurence, c); |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 339 | int shift = j - bc_occ; |
| 340 | if (gs_shift > shift) { |
| 341 | shift = gs_shift; |
| 342 | } |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 343 | index += shift; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 344 | } |
| 345 | } |
| 346 | |
| 347 | return -1; |
| 348 | } |
| 349 | |
| 350 | |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 351 | template <typename PatternChar, typename SubjectChar> |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 352 | void StringSearch<PatternChar, SubjectChar>::PopulateBoyerMooreTable() { |
| 353 | int pattern_length = pattern_.length(); |
| 354 | const PatternChar* pattern = pattern_.start(); |
| 355 | // Only look at the last kBMMaxShift characters of pattern (from start_ |
| 356 | // to pattern_length). |
| 357 | int start = start_; |
| 358 | int length = pattern_length - start; |
| 359 | |
| 360 | // Biased tables so that we can use pattern indices as table indices, |
| 361 | // even if we only cover the part of the pattern from offset start. |
| 362 | int* shift_table = good_suffix_shift_table(); |
| 363 | int* suffix_table = this->suffix_table(); |
| 364 | |
| 365 | // Initialize table. |
| 366 | for (int i = start; i < pattern_length; i++) { |
| 367 | shift_table[i] = length; |
| 368 | } |
| 369 | shift_table[pattern_length] = 1; |
| 370 | suffix_table[pattern_length] = pattern_length + 1; |
| 371 | |
| 372 | // Find suffixes. |
| 373 | PatternChar last_char = pattern[pattern_length - 1]; |
| 374 | int suffix = pattern_length + 1; |
| 375 | { |
| 376 | int i = pattern_length; |
| 377 | while (i > start) { |
| 378 | PatternChar c = pattern[i - 1]; |
| 379 | while (suffix <= pattern_length && c != pattern[suffix - 1]) { |
| 380 | if (shift_table[suffix] == length) { |
| 381 | shift_table[suffix] = suffix - i; |
| 382 | } |
| 383 | suffix = suffix_table[suffix]; |
| 384 | } |
| 385 | suffix_table[--i] = --suffix; |
| 386 | if (suffix == pattern_length) { |
| 387 | // No suffix to extend, so we check against last_char only. |
| 388 | while ((i > start) && (pattern[i - 1] != last_char)) { |
| 389 | if (shift_table[pattern_length] == length) { |
| 390 | shift_table[pattern_length] = pattern_length - i; |
| 391 | } |
| 392 | suffix_table[--i] = pattern_length; |
| 393 | } |
| 394 | if (i > start) { |
| 395 | suffix_table[--i] = --suffix; |
| 396 | } |
| 397 | } |
| 398 | } |
| 399 | } |
| 400 | // Build shift table using suffixes. |
| 401 | if (suffix < pattern_length) { |
| 402 | for (int i = start; i <= pattern_length; i++) { |
| 403 | if (shift_table[i] == length) { |
| 404 | shift_table[i] = suffix - start; |
| 405 | } |
| 406 | if (i == suffix) { |
| 407 | suffix = suffix_table[suffix]; |
| 408 | } |
| 409 | } |
| 410 | } |
| 411 | } |
| 412 | |
| 413 | //--------------------------------------------------------------------- |
| 414 | // Boyer-Moore-Horspool string search. |
| 415 | //--------------------------------------------------------------------- |
| 416 | |
| 417 | template <typename PatternChar, typename SubjectChar> |
| 418 | int StringSearch<PatternChar, SubjectChar>::BoyerMooreHorspoolSearch( |
| 419 | StringSearch<PatternChar, SubjectChar>* search, |
| 420 | Vector<const SubjectChar> subject, |
| 421 | int start_index) { |
| 422 | Vector<const PatternChar> pattern = search->pattern_; |
| 423 | int subject_length = subject.length(); |
| 424 | int pattern_length = pattern.length(); |
| 425 | int* char_occurrences = search->bad_char_table(); |
| 426 | int badness = -pattern_length; |
| 427 | |
| 428 | // How bad we are doing without a good-suffix table. |
| 429 | PatternChar last_char = pattern[pattern_length - 1]; |
| 430 | int last_char_shift = pattern_length - 1 - |
| 431 | CharOccurrence(char_occurrences, static_cast<SubjectChar>(last_char)); |
| 432 | // Perform search |
| 433 | int index = start_index; // No matches found prior to this index. |
| 434 | while (index <= subject_length - pattern_length) { |
| 435 | int j = pattern_length - 1; |
| 436 | int subject_char; |
| 437 | while (last_char != (subject_char = subject[index + j])) { |
| 438 | int bc_occ = CharOccurrence(char_occurrences, subject_char); |
| 439 | int shift = j - bc_occ; |
| 440 | index += shift; |
| 441 | badness += 1 - shift; // at most zero, so badness cannot increase. |
| 442 | if (index > subject_length - pattern_length) { |
| 443 | return -1; |
| 444 | } |
| 445 | } |
| 446 | j--; |
| 447 | while (j >= 0 && pattern[j] == (subject[index + j])) j--; |
| 448 | if (j < 0) { |
| 449 | return index; |
| 450 | } else { |
| 451 | index += last_char_shift; |
| 452 | // Badness increases by the number of characters we have |
| 453 | // checked, and decreases by the number of characters we |
| 454 | // can skip by shifting. It's a measure of how we are doing |
| 455 | // compared to reading each character exactly once. |
| 456 | badness += (pattern_length - j) - last_char_shift; |
| 457 | if (badness > 0) { |
| 458 | search->PopulateBoyerMooreTable(); |
| 459 | search->strategy_ = &BoyerMooreSearch; |
| 460 | return BoyerMooreSearch(search, subject, index); |
| 461 | } |
| 462 | } |
| 463 | } |
| 464 | return -1; |
| 465 | } |
| 466 | |
| 467 | |
| 468 | template <typename PatternChar, typename SubjectChar> |
| 469 | void StringSearch<PatternChar, SubjectChar>::PopulateBoyerMooreHorspoolTable() { |
| 470 | int pattern_length = pattern_.length(); |
| 471 | |
| 472 | int* bad_char_occurrence = bad_char_table(); |
| 473 | |
| 474 | // Only preprocess at most kBMMaxShift last characters of pattern. |
| 475 | int start = start_; |
| 476 | // Run forwards to populate bad_char_table, so that *last* instance |
| 477 | // of character equivalence class is the one registered. |
| 478 | // Notice: Doesn't include the last character. |
| 479 | int table_size = AlphabetSize(); |
| 480 | if (start == 0) { // All patterns less than kBMMaxShift in length. |
| 481 | memset(bad_char_occurrence, |
| 482 | -1, |
| 483 | table_size * sizeof(*bad_char_occurrence)); |
| 484 | } else { |
| 485 | for (int i = 0; i < table_size; i++) { |
| 486 | bad_char_occurrence[i] = start - 1; |
| 487 | } |
| 488 | } |
| 489 | for (int i = start; i < pattern_length - 1; i++) { |
| 490 | PatternChar c = pattern_[i]; |
| 491 | int bucket = (sizeof(PatternChar) == 1) ? c : c % AlphabetSize(); |
| 492 | bad_char_occurrence[bucket] = i; |
| 493 | } |
| 494 | } |
| 495 | |
| 496 | //--------------------------------------------------------------------- |
| 497 | // Linear string search with bailout to BMH. |
| 498 | //--------------------------------------------------------------------- |
| 499 | |
| 500 | // Simple linear search for short patterns, which bails out if the string |
| 501 | // isn't found very early in the subject. Upgrades to BoyerMooreHorspool. |
| 502 | template <typename PatternChar, typename SubjectChar> |
| 503 | int StringSearch<PatternChar, SubjectChar>::InitialSearch( |
| 504 | StringSearch<PatternChar, SubjectChar>* search, |
| 505 | Vector<const SubjectChar> subject, |
| 506 | int index) { |
| 507 | Vector<const PatternChar> pattern = search->pattern_; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 508 | int pattern_length = pattern.length(); |
| 509 | // Badness is a count of how much work we have done. When we have |
| 510 | // done enough work we decide it's probably worth switching to a better |
| 511 | // algorithm. |
| 512 | int badness = -10 - (pattern_length << 2); |
| 513 | |
| 514 | // We know our pattern is at least 2 characters, we cache the first so |
| 515 | // the common case of the first character not matching is faster. |
| 516 | PatternChar pattern_first_char = pattern[0]; |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 517 | for (int i = index, n = subject.length() - pattern_length; i <= n; i++) { |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 518 | badness++; |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 519 | if (badness <= 0) { |
| 520 | if (sizeof(SubjectChar) == 1 && sizeof(PatternChar) == 1) { |
| 521 | const SubjectChar* pos = reinterpret_cast<const SubjectChar*>( |
| 522 | memchr(subject.start() + i, |
| 523 | pattern_first_char, |
| 524 | n - i + 1)); |
| 525 | if (pos == NULL) { |
| 526 | return -1; |
| 527 | } |
| 528 | i = static_cast<int>(pos - subject.start()); |
| 529 | } else { |
| 530 | if (subject[i] != pattern_first_char) continue; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 531 | } |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 532 | int j = 1; |
| 533 | do { |
| 534 | if (pattern[j] != subject[i + j]) { |
| 535 | break; |
| 536 | } |
| 537 | j++; |
| 538 | } while (j < pattern_length); |
| 539 | if (j == pattern_length) { |
| 540 | return i; |
| 541 | } |
| 542 | badness += j; |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 543 | } else { |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 544 | search->PopulateBoyerMooreHorspoolTable(); |
| 545 | search->strategy_ = &BoyerMooreHorspoolSearch; |
| 546 | return BoyerMooreHorspoolSearch(search, subject, i); |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 547 | } |
| 548 | } |
| 549 | return -1; |
| 550 | } |
| 551 | |
| 552 | |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 553 | // Perform a a single stand-alone search. |
| 554 | // If searching multiple times for the same pattern, a search |
| 555 | // object should be constructed once and the Search function then called |
| 556 | // for each search. |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 557 | template <typename SubjectChar, typename PatternChar> |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 558 | static int SearchString(Vector<const SubjectChar> subject, |
| 559 | Vector<const PatternChar> pattern, |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 560 | int start_index) { |
Ben Murdoch | f87a203 | 2010-10-22 12:50:53 +0100 | [diff] [blame] | 561 | StringSearch<PatternChar, SubjectChar> search(pattern); |
| 562 | return search.Search(subject, start_index); |
Steve Block | 5915150 | 2010-09-22 15:07:15 +0100 | [diff] [blame] | 563 | } |
| 564 | |
| 565 | }} // namespace v8::internal |
| 566 | |
| 567 | #endif // V8_STRING_SEARCH_H_ |