1 2 // 3 // file: regexcmp.cpp 4 // 5 // Copyright (C) 2002-2008 International Business Machines Corporation and others. 6 // All Rights Reserved. 7 // 8 // This file contains the ICU regular expression compiler, which is responsible 9 // for processing a regular expression pattern into the compiled form that 10 // is used by the match finding engine. 11 // 12 13 #include "unicode/utypes.h" 14 15 #if !UCONFIG_NO_REGULAR_EXPRESSIONS 16 17 #include "unicode/unistr.h" 18 #include "unicode/uniset.h" 19 #include "unicode/uchar.h" 20 #include "unicode/uchriter.h" 21 #include "unicode/parsepos.h" 22 #include "unicode/parseerr.h" 23 #include "unicode/regex.h" 24 #include "../common/util.h" 25 #include "cmemory.h" 26 #include "cstring.h" 27 #include "uvectr32.h" 28 #include "uassert.h" 29 #include "ucln_in.h" 30 #include "uinvchar.h" 31 32 #include "regeximp.h" 33 #include "regexcst.h" // Contains state table for the regex pattern parser. 34 // generated by a Perl script. 35 #include "regexcmp.h" 36 #include "regexst.h" 37 38 39 40 U_NAMESPACE_BEGIN 41 42 43 //------------------------------------------------------------------------------ 44 // 45 // Constructor. 46 // 47 //------------------------------------------------------------------------------ 48 RegexCompile::RegexCompile(RegexPattern *rxp, UErrorCode &status) : 49 fParenStack(status), fSetStack(status), fSetOpStack(status) 50 { 51 fStatus = &status; 52 53 fRXPat = rxp; 54 fScanIndex = 0; 55 fNextIndex = 0; 56 fPeekChar = -1; 57 fLineNum = 1; 58 fCharNum = 0; 59 fQuoteMode = FALSE; 60 fInBackslashQuote = FALSE; 61 fModeFlags = fRXPat->fFlags | 0x80000000; 62 fEOLComments = TRUE; 63 64 fMatchOpenParen = -1; 65 fMatchCloseParen = -1; 66 fStringOpStart = -1; 67 68 if (U_SUCCESS(status) && U_FAILURE(rxp->fDeferredStatus)) { 69 status = rxp->fDeferredStatus; 70 } 71 } 72 73 static const UChar chAmp = 0x26; // '&' 74 static const UChar chDash = 0x2d; // '-' 75 76 77 //------------------------------------------------------------------------------ 78 // 79 // Destructor 80 // 81 //------------------------------------------------------------------------------ 82 RegexCompile::~RegexCompile() { 83 } 84 85 static inline void addCategory(UnicodeSet *set, int32_t value, UErrorCode& ec) { 86 set->addAll(UnicodeSet().applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, value, ec)); 87 } 88 89 //------------------------------------------------------------------------------ 90 // 91 // Compile regex pattern. The state machine for rexexp pattern parsing is here. 92 // The state tables are hand-written in the file regexcst.txt, 93 // and converted to the form used here by a perl 94 // script regexcst.pl 95 // 96 //------------------------------------------------------------------------------ 97 void RegexCompile::compile( 98 const UnicodeString &pat, // Source pat to be compiled. 99 UParseError &pp, // Error position info 100 UErrorCode &e) // Error Code 101 { 102 fStatus = &e; 103 fParseErr = &pp; 104 fStackPtr = 0; 105 fStack[fStackPtr] = 0; 106 107 if (U_FAILURE(*fStatus)) { 108 return; 109 } 110 111 // There should be no pattern stuff in the RegexPattern object. They can not be reused. 112 U_ASSERT(fRXPat->fPattern.length() == 0); 113 114 // Prepare the RegexPattern object to receive the compiled pattern. 115 fRXPat->fPattern = pat; 116 fRXPat->fStaticSets = RegexStaticSets::gStaticSets->fPropSets; 117 fRXPat->fStaticSets8 = RegexStaticSets::gStaticSets->fPropSets8; 118 119 120 // Initialize the pattern scanning state machine 121 fPatternLength = pat.length(); 122 uint16_t state = 1; 123 const RegexTableEl *tableEl; 124 nextChar(fC); // Fetch the first char from the pattern string. 125 126 // 127 // Main loop for the regex pattern parsing state machine. 128 // Runs once per state transition. 129 // Each time through optionally performs, depending on the state table, 130 // - an advance to the the next pattern char 131 // - an action to be performed. 132 // - pushing or popping a state to/from the local state return stack. 133 // file regexcst.txt is the source for the state table. The logic behind 134 // recongizing the pattern syntax is there, not here. 135 // 136 for (;;) { 137 // Bail out if anything has gone wrong. 138 // Regex pattern parsing stops on the first error encountered. 139 if (U_FAILURE(*fStatus)) { 140 break; 141 } 142 143 U_ASSERT(state != 0); 144 145 // Find the state table element that matches the input char from the pattern, or the 146 // class of the input character. Start with the first table row for this 147 // state, then linearly scan forward until we find a row that matches the 148 // character. The last row for each state always matches all characters, so 149 // the search will stop there, if not before. 150 // 151 tableEl = &gRuleParseStateTable[state]; 152 REGEX_SCAN_DEBUG_PRINTF(("char, line, col = (\'%c\', %d, %d) state=%s ", 153 fC.fChar, fLineNum, fCharNum, RegexStateNames[state])); 154 155 for (;;) { // loop through table rows belonging to this state, looking for one 156 // that matches the current input char. 157 REGEX_SCAN_DEBUG_PRINTF((".")); 158 if (tableEl->fCharClass < 127 && fC.fQuoted == FALSE && tableEl->fCharClass == fC.fChar) { 159 // Table row specified an individual character, not a set, and 160 // the input character is not quoted, and 161 // the input character matched it. 162 break; 163 } 164 if (tableEl->fCharClass == 255) { 165 // Table row specified default, match anything character class. 166 break; 167 } 168 if (tableEl->fCharClass == 254 && fC.fQuoted) { 169 // Table row specified "quoted" and the char was quoted. 170 break; 171 } 172 if (tableEl->fCharClass == 253 && fC.fChar == (UChar32)-1) { 173 // Table row specified eof and we hit eof on the input. 174 break; 175 } 176 177 if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 && // Table specs a char class && 178 fC.fQuoted == FALSE && // char is not escaped && 179 fC.fChar != (UChar32)-1) { // char is not EOF 180 if (RegexStaticSets::gStaticSets->fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) { 181 // Table row specified a character class, or set of characters, 182 // and the current char matches it. 183 break; 184 } 185 } 186 187 // No match on this row, advance to the next row for this state, 188 tableEl++; 189 } 190 REGEX_SCAN_DEBUG_PRINTF(("\n")); 191 192 // 193 // We've found the row of the state table that matches the current input 194 // character from the rules string. 195 // Perform any action specified by this row in the state table. 196 if (doParseActions(tableEl->fAction) == FALSE) { 197 // Break out of the state machine loop if the 198 // the action signalled some kind of error, or 199 // the action was to exit, occurs on normal end-of-rules-input. 200 break; 201 } 202 203 if (tableEl->fPushState != 0) { 204 fStackPtr++; 205 if (fStackPtr >= kStackSize) { 206 error(U_REGEX_INTERNAL_ERROR); 207 REGEX_SCAN_DEBUG_PRINTF(("RegexCompile::parse() - state stack overflow.\n")); 208 fStackPtr--; 209 } 210 fStack[fStackPtr] = tableEl->fPushState; 211 } 212 213 // 214 // NextChar. This is where characters are actually fetched from the pattern. 215 // Happens under control of the 'n' tag in the state table. 216 // 217 if (tableEl->fNextChar) { 218 nextChar(fC); 219 } 220 221 // Get the next state from the table entry, or from the 222 // state stack if the next state was specified as "pop". 223 if (tableEl->fNextState != 255) { 224 state = tableEl->fNextState; 225 } else { 226 state = fStack[fStackPtr]; 227 fStackPtr--; 228 if (fStackPtr < 0) { 229 // state stack underflow 230 // This will occur if the user pattern has mis-matched parentheses, 231 // with extra close parens. 232 // 233 fStackPtr++; 234 error(U_REGEX_MISMATCHED_PAREN); 235 } 236 } 237 238 } 239 240 if (U_FAILURE(*fStatus)) { 241 // Bail out if the pattern had errors. 242 // Set stack cleanup: a successful compile would have left it empty, 243 // but errors can leave temporary sets hanging around. 244 while (!fSetStack.empty()) { 245 delete (UnicodeSet *)fSetStack.pop(); 246 } 247 return; 248 } 249 250 // 251 // The pattern has now been read and processed, and the compiled code generated. 252 // 253 254 // Back-reference fixup 255 // 256 int32_t loc; 257 for (loc=0; loc<fRXPat->fCompiledPat->size(); loc++) { 258 int32_t op = fRXPat->fCompiledPat->elementAti(loc); 259 int32_t opType = URX_TYPE(op); 260 if (opType == URX_BACKREF || opType == URX_BACKREF_I) { 261 int32_t where = URX_VAL(op); 262 if (where > fRXPat->fGroupMap->size()) { 263 error(U_REGEX_INVALID_BACK_REF); 264 break; 265 } 266 where = fRXPat->fGroupMap->elementAti(where-1); 267 op = URX_BUILD(opType, where); 268 fRXPat->fCompiledPat->setElementAt(op, loc); 269 } 270 } 271 272 273 // 274 // Compute the number of digits requried for the largest capture group number. 275 // 276 fRXPat->fMaxCaptureDigits = 1; 277 int32_t n = 10; 278 for (;;) { 279 if (n > fRXPat->fGroupMap->size()) { 280 break; 281 } 282 fRXPat->fMaxCaptureDigits++; 283 n *= 10; 284 } 285 286 // 287 // The pattern's fFrameSize so far has accumulated the requirements for 288 // storage for capture parentheses, counters, etc. that are encountered 289 // in the pattern. Add space for the two variables that are always 290 // present in the saved state: the input string position and the 291 // position in the compiled pattern. 292 // 293 fRXPat->fFrameSize+=2; 294 295 // 296 // Get bounds for the minimum and maximum length of a string that this 297 // pattern can match. Used to avoid looking for matches in strings that 298 // are too short. 299 // 300 fRXPat->fMinMatchLen = minMatchLength(3, fRXPat->fCompiledPat->size()-1); 301 302 // 303 // Optimization passes 304 // 305 matchStartType(); 306 stripNOPs(); 307 308 // 309 // Set up fast latin-1 range sets 310 // 311 int32_t numSets = fRXPat->fSets->size(); 312 fRXPat->fSets8 = new Regex8BitSet[numSets]; 313 // Null pointer check. 314 if (fRXPat->fSets8 == NULL) { 315 e = *fStatus = U_MEMORY_ALLOCATION_ERROR; 316 return; 317 } 318 int32_t i; 319 for (i=0; i<numSets; i++) { 320 UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(i); 321 fRXPat->fSets8[i].init(s); 322 } 323 324 } 325 326 327 328 329 330 //------------------------------------------------------------------------------ 331 // 332 // doParseAction Do some action during regex pattern parsing. 333 // Called by the parse state machine. 334 // 335 // Generation of the match engine PCode happens here, or 336 // in functions called from the parse actions defined here. 337 // 338 // 339 //------------------------------------------------------------------------------ 340 UBool RegexCompile::doParseActions(int32_t action) 341 { 342 UBool returnVal = TRUE; 343 344 switch ((Regex_PatternParseAction)action) { 345 346 case doPatStart: 347 // Start of pattern compiles to: 348 //0 SAVE 2 Fall back to position of FAIL 349 //1 jmp 3 350 //2 FAIL Stop if we ever reach here. 351 //3 NOP Dummy, so start of pattern looks the same as 352 // the start of an ( grouping. 353 //4 NOP Resreved, will be replaced by a save if there are 354 // OR | operators at the top level 355 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_STATE_SAVE, 2), *fStatus); 356 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_JMP, 3), *fStatus); 357 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_FAIL, 0), *fStatus); 358 359 // Standard open nonCapture paren action emits the two NOPs and 360 // sets up the paren stack frame. 361 doParseActions(doOpenNonCaptureParen); 362 break; 363 364 case doPatFinish: 365 // We've scanned to the end of the pattern 366 // The end of pattern compiles to: 367 // URX_END 368 // which will stop the runtime match engine. 369 // Encountering end of pattern also behaves like a close paren, 370 // and forces fixups of the State Save at the beginning of the compiled pattern 371 // and of any OR operations at the top level. 372 // 373 handleCloseParen(); 374 if (fParenStack.size() > 0) { 375 // Missing close paren in pattern. 376 error(U_REGEX_MISMATCHED_PAREN); 377 } 378 379 // add the END operation to the compiled pattern. 380 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_END, 0), *fStatus); 381 382 // Terminate the pattern compilation state machine. 383 returnVal = FALSE; 384 break; 385 386 387 388 case doOrOperator: 389 // Scanning a '|', as in (A|B) 390 { 391 // Insert a SAVE operation at the start of the pattern section preceding 392 // this OR at this level. This SAVE will branch the match forward 393 // to the right hand side of the OR in the event that the left hand 394 // side fails to match and backtracks. Locate the position for the 395 // save from the location on the top of the parentheses stack. 396 int32_t savePosition = fParenStack.popi(); 397 int32_t op = fRXPat->fCompiledPat->elementAti(savePosition); 398 U_ASSERT(URX_TYPE(op) == URX_NOP); // original contents of reserved location 399 op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+1); 400 fRXPat->fCompiledPat->setElementAt(op, savePosition); 401 402 // Append an JMP operation into the compiled pattern. The operand for 403 // the JMP will eventually be the location following the ')' for the 404 // group. This will be patched in later, when the ')' is encountered. 405 op = URX_BUILD(URX_JMP, 0); 406 fRXPat->fCompiledPat->addElement(op, *fStatus); 407 408 // Push the position of the newly added JMP op onto the parentheses stack. 409 // This registers if for fixup when this block's close paren is encountered. 410 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); 411 412 // Append a NOP to the compiled pattern. This is the slot reserved 413 // for a SAVE in the event that there is yet another '|' following 414 // this one. 415 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 416 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); 417 } 418 break; 419 420 421 case doOpenCaptureParen: 422 // Open Paren. 423 // Compile to a 424 // - NOP, which later may be replaced by a save-state if the 425 // parenthesized group gets a * quantifier, followed by 426 // - START_CAPTURE n where n is stack frame offset to the capture group variables. 427 // - NOP, which may later be replaced by a save-state if there 428 // is an '|' alternation within the parens. 429 // 430 // Each capture group gets three slots in the save stack frame: 431 // 0: Capture Group start position (in input string being matched.) 432 // 1: Capture Group end positino. 433 // 2: Start of Match-in-progress. 434 // The first two locations are for a completed capture group, and are 435 // referred to by back references and the like. 436 // The third location stores the capture start position when an START_CAPTURE is 437 // encountered. This will be promoted to a completed capture when (and if) the corresponding 438 // END_CAPure is encountered. 439 { 440 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 441 int32_t varsLoc = fRXPat->fFrameSize; // Reserve three slots in match stack frame. 442 fRXPat->fFrameSize += 3; 443 int32_t cop = URX_BUILD(URX_START_CAPTURE, varsLoc); 444 fRXPat->fCompiledPat->addElement(cop, *fStatus); 445 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 446 447 // On the Parentheses stack, start a new frame and add the postions 448 // of the two NOPs. Depending on what follows in the pattern, the 449 // NOPs may be changed to SAVE_STATE or JMP ops, with a target 450 // address of the end of the parenthesized group. 451 fParenStack.push(fModeFlags, *fStatus); // Match mode state 452 fParenStack.push(capturing, *fStatus); // Frame type. 453 fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus); // The first NOP location 454 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP loc 455 456 // Save the mapping from group number to stack frame variable position. 457 fRXPat->fGroupMap->addElement(varsLoc, *fStatus); 458 } 459 break; 460 461 case doOpenNonCaptureParen: 462 // Open non-caputuring (grouping only) Paren. 463 // Compile to a 464 // - NOP, which later may be replaced by a save-state if the 465 // parenthesized group gets a * quantifier, followed by 466 // - NOP, which may later be replaced by a save-state if there 467 // is an '|' alternation within the parens. 468 { 469 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 470 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 471 472 // On the Parentheses stack, start a new frame and add the postions 473 // of the two NOPs. 474 fParenStack.push(fModeFlags, *fStatus); // Match mode state 475 fParenStack.push(plain, *fStatus); // Begin a new frame. 476 fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 477 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP loc 478 } 479 break; 480 481 482 case doOpenAtomicParen: 483 // Open Atomic Paren. (?> 484 // Compile to a 485 // - NOP, which later may be replaced if the parenthesized group 486 // has a quantifier, followed by 487 // - STO_SP save state stack position, so it can be restored at the ")" 488 // - NOP, which may later be replaced by a save-state if there 489 // is an '|' alternation within the parens. 490 { 491 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 492 int32_t varLoc = fRXPat->fDataSize; // Reserve a data location for saving the 493 fRXPat->fDataSize += 1; // state stack ptr. 494 int32_t stoOp = URX_BUILD(URX_STO_SP, varLoc); 495 fRXPat->fCompiledPat->addElement(stoOp, *fStatus); 496 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 497 498 // On the Parentheses stack, start a new frame and add the postions 499 // of the two NOPs. Depending on what follows in the pattern, the 500 // NOPs may be changed to SAVE_STATE or JMP ops, with a target 501 // address of the end of the parenthesized group. 502 fParenStack.push(fModeFlags, *fStatus); // Match mode state 503 fParenStack.push(atomic, *fStatus); // Frame type. 504 fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus); // The first NOP 505 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP 506 } 507 break; 508 509 510 case doOpenLookAhead: 511 // Positive Look-ahead (?= stuff ) 512 // 513 // Note: Addition of transparent input regions, with the need to 514 // restore the original regions when failing out of a lookahead 515 // block, complicated this sequence. Some conbined opcodes 516 // might make sense - or might not, lookahead aren't that common. 517 // 518 // Caution: min match length optimization knows about this 519 // sequence; don't change without making updates there too. 520 // 521 // Compiles to 522 // 1 START_LA dataLoc Saves SP, Input Pos 523 // 2. STATE_SAVE 4 on failure of lookahead, goto 4 524 // 3 JMP 6 continue ... 525 // 526 // 4. LA_END Look Ahead failed. Restore regions. 527 // 5. BACKTRACK and back track again. 528 // 529 // 6. NOP reserved for use by quantifiers on the block. 530 // Look-ahead can't have quantifiers, but paren stack 531 // compile time conventions require the slot anyhow. 532 // 7. NOP may be replaced if there is are '|' ops in the block. 533 // 8. code for parenthesized stuff. 534 // 9. LA_END 535 // 536 // Two data slots are reserved, for saving the stack ptr and the input position. 537 { 538 int32_t dataLoc = fRXPat->fDataSize; 539 fRXPat->fDataSize += 2; 540 int32_t op = URX_BUILD(URX_LA_START, dataLoc); 541 fRXPat->fCompiledPat->addElement(op, *fStatus); 542 543 op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+ 2); 544 fRXPat->fCompiledPat->addElement(op, *fStatus); 545 546 op = URX_BUILD(URX_JMP, fRXPat->fCompiledPat->size()+ 3); 547 fRXPat->fCompiledPat->addElement(op, *fStatus); 548 549 op = URX_BUILD(URX_LA_END, dataLoc); 550 fRXPat->fCompiledPat->addElement(op, *fStatus); 551 552 op = URX_BUILD(URX_BACKTRACK, 0); 553 fRXPat->fCompiledPat->addElement(op, *fStatus); 554 555 op = URX_BUILD(URX_NOP, 0); 556 fRXPat->fCompiledPat->addElement(op, *fStatus); 557 fRXPat->fCompiledPat->addElement(op, *fStatus); 558 559 // On the Parentheses stack, start a new frame and add the postions 560 // of the NOPs. 561 fParenStack.push(fModeFlags, *fStatus); // Match mode state 562 fParenStack.push(lookAhead, *fStatus); // Frame type. 563 fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 564 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP location 565 } 566 break; 567 568 case doOpenLookAheadNeg: 569 // Negated Lookahead. (?! stuff ) 570 // Compiles to 571 // 1. START_LA dataloc 572 // 2. SAVE_STATE 7 // Fail within look-ahead block restores to this state, 573 // // which continues with the match. 574 // 3. NOP // Std. Open Paren sequence, for possible '|' 575 // 4. code for parenthesized stuff. 576 // 5. END_LA // Cut back stack, remove saved state from step 2. 577 // 6. BACKTRACK // code in block succeeded, so neg. lookahead fails. 578 // 7. END_LA // Restore match region, in case look-ahead was using 579 // an alternate (transparent) region. 580 { 581 int32_t dataLoc = fRXPat->fDataSize; 582 fRXPat->fDataSize += 2; 583 int32_t op = URX_BUILD(URX_LA_START, dataLoc); 584 fRXPat->fCompiledPat->addElement(op, *fStatus); 585 586 op = URX_BUILD(URX_STATE_SAVE, 0); // dest address will be patched later. 587 fRXPat->fCompiledPat->addElement(op, *fStatus); 588 589 op = URX_BUILD(URX_NOP, 0); 590 fRXPat->fCompiledPat->addElement(op, *fStatus); 591 592 // On the Parentheses stack, start a new frame and add the postions 593 // of the StateSave and NOP. 594 fParenStack.push(fModeFlags, *fStatus); // Match mode state 595 fParenStack.push(negLookAhead, *fStatus); // Frame type 596 fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The STATE_SAVE location 597 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP location 598 599 // BEGIN android-changed 600 // Apply important regex bug fixing for ticket#7224 601 // Instructions #5 - #7 will be added when the ')' is encountered. 602 // END android-changed 603 } 604 break; 605 606 case doOpenLookBehind: 607 { 608 // Compile a (?<= look-behind open paren. 609 // 610 // Compiles to 611 // 0 URX_LB_START dataLoc 612 // 1 URX_LB_CONT dataLoc 613 // 2 MinMatchLen 614 // 3 MaxMatchLen 615 // 4 URX_NOP Standard '(' boilerplate. 616 // 5 URX_NOP Reserved slot for use with '|' ops within (block). 617 // 6 <code for LookBehind expression> 618 // 7 URX_LB_END dataLoc # Check match len, restore input len 619 // 8 URX_LA_END dataLoc # Restore stack, input pos 620 // 621 // Allocate a block of matcher data, to contain (when running a match) 622 // 0: Stack ptr on entry 623 // 1: Input Index on entry 624 // 2: Start index of match current match attempt. 625 // 3: Original Input String len. 626 627 // Allocate data space 628 int32_t dataLoc = fRXPat->fDataSize; 629 fRXPat->fDataSize += 4; 630 631 // Emit URX_LB_START 632 int32_t op = URX_BUILD(URX_LB_START, dataLoc); 633 fRXPat->fCompiledPat->addElement(op, *fStatus); 634 635 // Emit URX_LB_CONT 636 op = URX_BUILD(URX_LB_CONT, dataLoc); 637 fRXPat->fCompiledPat->addElement(op, *fStatus); 638 fRXPat->fCompiledPat->addElement(0, *fStatus); // MinMatchLength. To be filled later. 639 fRXPat->fCompiledPat->addElement(0, *fStatus); // MaxMatchLength. To be filled later. 640 641 // Emit the NOP 642 op = URX_BUILD(URX_NOP, 0); 643 fRXPat->fCompiledPat->addElement(op, *fStatus); 644 fRXPat->fCompiledPat->addElement(op, *fStatus); 645 646 // On the Parentheses stack, start a new frame and add the postions 647 // of the URX_LB_CONT and the NOP. 648 fParenStack.push(fModeFlags, *fStatus); // Match mode state 649 fParenStack.push(lookBehind, *fStatus); // Frame type 650 fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 651 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The 2nd NOP location 652 653 // The final two instructions will be added when the ')' is encountered. 654 } 655 656 break; 657 658 case doOpenLookBehindNeg: 659 { 660 // Compile a (?<! negated look-behind open paren. 661 // 662 // Compiles to 663 // 0 URX_LB_START dataLoc # Save entry stack, input len 664 // 1 URX_LBN_CONT dataLoc # Iterate possible match positions 665 // 2 MinMatchLen 666 // 3 MaxMatchLen 667 // 4 continueLoc (9) 668 // 5 URX_NOP Standard '(' boilerplate. 669 // 6 URX_NOP Reserved slot for use with '|' ops within (block). 670 // 7 <code for LookBehind expression> 671 // 8 URX_LBN_END dataLoc # Check match len, cause a FAIL 672 // 9 ... 673 // 674 // Allocate a block of matcher data, to contain (when running a match) 675 // 0: Stack ptr on entry 676 // 1: Input Index on entry 677 // 2: Start index of match current match attempt. 678 // 3: Original Input String len. 679 680 // Allocate data space 681 int32_t dataLoc = fRXPat->fDataSize; 682 fRXPat->fDataSize += 4; 683 684 // Emit URX_LB_START 685 int32_t op = URX_BUILD(URX_LB_START, dataLoc); 686 fRXPat->fCompiledPat->addElement(op, *fStatus); 687 688 // Emit URX_LBN_CONT 689 op = URX_BUILD(URX_LBN_CONT, dataLoc); 690 fRXPat->fCompiledPat->addElement(op, *fStatus); 691 fRXPat->fCompiledPat->addElement(0, *fStatus); // MinMatchLength. To be filled later. 692 fRXPat->fCompiledPat->addElement(0, *fStatus); // MaxMatchLength. To be filled later. 693 fRXPat->fCompiledPat->addElement(0, *fStatus); // Continue Loc. To be filled later. 694 695 // Emit the NOP 696 op = URX_BUILD(URX_NOP, 0); 697 fRXPat->fCompiledPat->addElement(op, *fStatus); 698 fRXPat->fCompiledPat->addElement(op, *fStatus); 699 700 // On the Parentheses stack, start a new frame and add the postions 701 // of the URX_LB_CONT and the NOP. 702 fParenStack.push(fModeFlags, *fStatus); // Match mode state 703 fParenStack.push(lookBehindN, *fStatus); // Frame type 704 fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 705 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The 2nd NOP location 706 707 // The final two instructions will be added when the ')' is encountered. 708 } 709 break; 710 711 case doConditionalExpr: 712 // Conditionals such as (?(1)a:b) 713 case doPerlInline: 714 // Perl inline-condtionals. (?{perl code}a|b) We're not perl, no way to do them. 715 error(U_REGEX_UNIMPLEMENTED); 716 break; 717 718 719 case doCloseParen: 720 handleCloseParen(); 721 if (fParenStack.size() <= 0) { 722 // Extra close paren, or missing open paren. 723 error(U_REGEX_MISMATCHED_PAREN); 724 } 725 break; 726 727 case doNOP: 728 break; 729 730 731 case doBadOpenParenType: 732 case doRuleError: 733 error(U_REGEX_RULE_SYNTAX); 734 break; 735 736 737 case doMismatchedParenErr: 738 error(U_REGEX_MISMATCHED_PAREN); 739 break; 740 741 case doPlus: 742 // Normal '+' compiles to 743 // 1. stuff to be repeated (already built) 744 // 2. jmp-sav 1 745 // 3. ... 746 // 747 // Or, if the item to be repeated can match a zero length string, 748 // 1. STO_INP_LOC data-loc 749 // 2. body of stuff to be repeated 750 // 3. JMP_SAV_X 2 751 // 4. ... 752 753 // 754 // Or, if the item to be repeated is simple 755 // 1. Item to be repeated. 756 // 2. LOOP_SR_I set number (assuming repeated item is a set ref) 757 // 3. LOOP_C stack location 758 { 759 int32_t topLoc = blockTopLoc(FALSE); // location of item #1 760 int32_t frameLoc; 761 762 // Check for simple constructs, which may get special optimized code. 763 if (topLoc == fRXPat->fCompiledPat->size() - 1) { 764 int32_t repeatedOp = fRXPat->fCompiledPat->elementAti(topLoc); 765 766 if (URX_TYPE(repeatedOp) == URX_SETREF) { 767 // Emit optimized code for [char set]+ 768 int32_t loopOpI = URX_BUILD(URX_LOOP_SR_I, URX_VAL(repeatedOp)); 769 fRXPat->fCompiledPat->addElement(loopOpI, *fStatus); 770 frameLoc = fRXPat->fFrameSize; 771 fRXPat->fFrameSize++; 772 int32_t loopOpC = URX_BUILD(URX_LOOP_C, frameLoc); 773 fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 774 break; 775 } 776 777 if (URX_TYPE(repeatedOp) == URX_DOTANY || 778 URX_TYPE(repeatedOp) == URX_DOTANY_ALL || 779 URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) { 780 // Emit Optimized code for .+ operations. 781 int32_t loopOpI = URX_BUILD(URX_LOOP_DOT_I, 0); 782 if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) { 783 // URX_LOOP_DOT_I operand is a flag indicating ". matches any" mode. 784 loopOpI |= 1; 785 } 786 if (fModeFlags & UREGEX_UNIX_LINES) { 787 loopOpI |= 2; 788 } 789 fRXPat->fCompiledPat->addElement(loopOpI, *fStatus); 790 frameLoc = fRXPat->fFrameSize; 791 fRXPat->fFrameSize++; 792 int32_t loopOpC = URX_BUILD(URX_LOOP_C, frameLoc); 793 fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 794 break; 795 } 796 797 } 798 799 // General case. 800 801 // Check for minimum match length of zero, which requires 802 // extra loop-breaking code. 803 if (minMatchLength(topLoc, fRXPat->fCompiledPat->size()-1) == 0) { 804 // Zero length match is possible. 805 // Emit the code sequence that can handle it. 806 insertOp(topLoc); 807 frameLoc = fRXPat->fFrameSize; 808 fRXPat->fFrameSize++; 809 810 int32_t op = URX_BUILD(URX_STO_INP_LOC, frameLoc); 811 fRXPat->fCompiledPat->setElementAt(op, topLoc); 812 813 op = URX_BUILD(URX_JMP_SAV_X, topLoc+1); 814 fRXPat->fCompiledPat->addElement(op, *fStatus); 815 } else { 816 // Simpler code when the repeated body must match something non-empty 817 int32_t jmpOp = URX_BUILD(URX_JMP_SAV, topLoc); 818 fRXPat->fCompiledPat->addElement(jmpOp, *fStatus); 819 } 820 } 821 break; 822 823 case doNGPlus: 824 // Non-greedy '+?' compiles to 825 // 1. stuff to be repeated (already built) 826 // 2. state-save 1 827 // 3. ... 828 { 829 int32_t topLoc = blockTopLoc(FALSE); 830 int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, topLoc); 831 fRXPat->fCompiledPat->addElement(saveStateOp, *fStatus); 832 } 833 break; 834 835 836 case doOpt: 837 // Normal (greedy) ? quantifier. 838 // Compiles to 839 // 1. state save 3 840 // 2. body of optional block 841 // 3. ... 842 // Insert the state save into the compiled pattern, and we're done. 843 { 844 int32_t saveStateLoc = blockTopLoc(TRUE); 845 int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()); 846 fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc); 847 } 848 break; 849 850 case doNGOpt: 851 // Non-greedy ?? quantifier 852 // compiles to 853 // 1. jmp 4 854 // 2. body of optional block 855 // 3 jmp 5 856 // 4. state save 2 857 // 5 ... 858 // This code is less than ideal, with two jmps instead of one, because we can only 859 // insert one instruction at the top of the block being iterated. 860 { 861 int32_t jmp1_loc = blockTopLoc(TRUE); 862 int32_t jmp2_loc = fRXPat->fCompiledPat->size(); 863 864 int32_t jmp1_op = URX_BUILD(URX_JMP, jmp2_loc+1); 865 fRXPat->fCompiledPat->setElementAt(jmp1_op, jmp1_loc); 866 867 int32_t jmp2_op = URX_BUILD(URX_JMP, jmp2_loc+2); 868 fRXPat->fCompiledPat->addElement(jmp2_op, *fStatus); 869 870 int32_t save_op = URX_BUILD(URX_STATE_SAVE, jmp1_loc+1); 871 fRXPat->fCompiledPat->addElement(save_op, *fStatus); 872 } 873 break; 874 875 876 case doStar: 877 // Normal (greedy) * quantifier. 878 // Compiles to 879 // 1. STATE_SAVE 4 880 // 2. body of stuff being iterated over 881 // 3. JMP_SAV 2 882 // 4. ... 883 // 884 // Or, if the body is a simple [Set], 885 // 1. LOOP_SR_I set number 886 // 2. LOOP_C stack location 887 // ... 888 // 889 // Or if this is a .* 890 // 1. LOOP_DOT_I (. matches all mode flag) 891 // 2. LOOP_C stack location 892 // 893 // Or, if the body can match a zero-length string, to inhibit infinite loops, 894 // 1. STATE_SAVE 5 895 // 2. STO_INP_LOC data-loc 896 // 3. body of stuff 897 // 4. JMP_SAV_X 2 898 // 5. ... 899 { 900 // location of item #1, the STATE_SAVE 901 int32_t topLoc = blockTopLoc(FALSE); 902 int32_t dataLoc = -1; 903 904 // Check for simple *, where the construct being repeated 905 // compiled to single opcode, and might be optimizable. 906 if (topLoc == fRXPat->fCompiledPat->size() - 1) { 907 int32_t repeatedOp = fRXPat->fCompiledPat->elementAti(topLoc); 908 909 if (URX_TYPE(repeatedOp) == URX_SETREF) { 910 // Emit optimized code for a [char set]* 911 int32_t loopOpI = URX_BUILD(URX_LOOP_SR_I, URX_VAL(repeatedOp)); 912 fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc); 913 dataLoc = fRXPat->fFrameSize; 914 fRXPat->fFrameSize++; 915 int32_t loopOpC = URX_BUILD(URX_LOOP_C, dataLoc); 916 fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 917 break; 918 } 919 920 if (URX_TYPE(repeatedOp) == URX_DOTANY || 921 URX_TYPE(repeatedOp) == URX_DOTANY_ALL || 922 URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) { 923 // Emit Optimized code for .* operations. 924 int32_t loopOpI = URX_BUILD(URX_LOOP_DOT_I, 0); 925 if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) { 926 // URX_LOOP_DOT_I operand is a flag indicating . matches any mode. 927 loopOpI |= 1; 928 } 929 if ((fModeFlags & UREGEX_UNIX_LINES) != 0) { 930 loopOpI |= 2; 931 } 932 fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc); 933 dataLoc = fRXPat->fFrameSize; 934 fRXPat->fFrameSize++; 935 int32_t loopOpC = URX_BUILD(URX_LOOP_C, dataLoc); 936 fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 937 break; 938 } 939 } 940 941 // Emit general case code for this * 942 // The optimizations did not apply. 943 944 int32_t saveStateLoc = blockTopLoc(TRUE); 945 int32_t jmpOp = URX_BUILD(URX_JMP_SAV, saveStateLoc+1); 946 947 // Check for minimum match length of zero, which requires 948 // extra loop-breaking code. 949 if (minMatchLength(saveStateLoc, fRXPat->fCompiledPat->size()-1) == 0) { 950 insertOp(saveStateLoc); 951 dataLoc = fRXPat->fFrameSize; 952 fRXPat->fFrameSize++; 953 954 int32_t op = URX_BUILD(URX_STO_INP_LOC, dataLoc); 955 fRXPat->fCompiledPat->setElementAt(op, saveStateLoc+1); 956 jmpOp = URX_BUILD(URX_JMP_SAV_X, saveStateLoc+2); 957 } 958 959 // Locate the position in the compiled pattern where the match will continue 960 // after completing the *. (4 or 5 in the comment above) 961 int32_t continueLoc = fRXPat->fCompiledPat->size()+1; 962 963 // Put together the save state op store it into the compiled code. 964 int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, continueLoc); 965 fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc); 966 967 // Append the URX_JMP_SAV or URX_JMPX operation to the compiled pattern. 968 fRXPat->fCompiledPat->addElement(jmpOp, *fStatus); 969 } 970 break; 971 972 case doNGStar: 973 // Non-greedy *? quantifier 974 // compiles to 975 // 1. JMP 3 976 // 2. body of stuff being iterated over 977 // 3. STATE_SAVE 2 978 // 4 ... 979 { 980 int32_t jmpLoc = blockTopLoc(TRUE); // loc 1. 981 int32_t saveLoc = fRXPat->fCompiledPat->size(); // loc 3. 982 int32_t jmpOp = URX_BUILD(URX_JMP, saveLoc); 983 int32_t stateSaveOp = URX_BUILD(URX_STATE_SAVE, jmpLoc+1); 984 fRXPat->fCompiledPat->setElementAt(jmpOp, jmpLoc); 985 fRXPat->fCompiledPat->addElement(stateSaveOp, *fStatus); 986 } 987 break; 988 989 990 case doIntervalInit: 991 // The '{' opening an interval quantifier was just scanned. 992 // Init the counter varaiables that will accumulate the values as the digits 993 // are scanned. 994 fIntervalLow = 0; 995 fIntervalUpper = -1; 996 break; 997 998 case doIntevalLowerDigit: 999 // Scanned a digit from the lower value of an {lower,upper} interval 1000 { 1001 int32_t digitValue = u_charDigitValue(fC.fChar); 1002 U_ASSERT(digitValue >= 0); 1003 fIntervalLow = fIntervalLow*10 + digitValue; 1004 if (fIntervalLow < 0) { 1005 error(U_REGEX_NUMBER_TOO_BIG); 1006 } 1007 } 1008 break; 1009 1010 case doIntervalUpperDigit: 1011 // Scanned a digit from the upper value of an {lower,upper} interval 1012 { 1013 if (fIntervalUpper < 0) { 1014 fIntervalUpper = 0; 1015 } 1016 int32_t digitValue = u_charDigitValue(fC.fChar); 1017 U_ASSERT(digitValue >= 0); 1018 fIntervalUpper = fIntervalUpper*10 + digitValue; 1019 if (fIntervalUpper < 0) { 1020 error(U_REGEX_NUMBER_TOO_BIG); 1021 } 1022 } 1023 break; 1024 1025 case doIntervalSame: 1026 // Scanned a single value interval like {27}. Upper = Lower. 1027 fIntervalUpper = fIntervalLow; 1028 break; 1029 1030 case doInterval: 1031 // Finished scanning a normal {lower,upper} interval. Generate the code for it. 1032 if (compileInlineInterval() == FALSE) { 1033 compileInterval(URX_CTR_INIT, URX_CTR_LOOP); 1034 } 1035 break; 1036 1037 case doPossessiveInterval: 1038 // Finished scanning a Possessive {lower,upper}+ interval. Generate the code for it. 1039 { 1040 // Remember the loc for the top of the block being looped over. 1041 // (Can not reserve a slot in the compiled pattern at this time, because 1042 // compileInterval needs to reserve also, and blockTopLoc can only reserve 1043 // once per block.) 1044 int32_t topLoc = blockTopLoc(FALSE); 1045 1046 // Produce normal looping code. 1047 compileInterval(URX_CTR_INIT, URX_CTR_LOOP); 1048 1049 // Surround the just-emitted normal looping code with a STO_SP ... LD_SP 1050 // just as if the loop was inclosed in atomic parentheses. 1051 1052 // First the STO_SP before the start of the loop 1053 insertOp(topLoc); 1054 int32_t varLoc = fRXPat->fDataSize; // Reserve a data location for saving the 1055 fRXPat->fDataSize += 1; // state stack ptr. 1056 int32_t op = URX_BUILD(URX_STO_SP, varLoc); 1057 fRXPat->fCompiledPat->setElementAt(op, topLoc); 1058 1059 int32_t loopOp = fRXPat->fCompiledPat->popi(); 1060 U_ASSERT(URX_TYPE(loopOp) == URX_CTR_LOOP && URX_VAL(loopOp) == topLoc); 1061 loopOp++; // point LoopOp after the just-inserted STO_SP 1062 fRXPat->fCompiledPat->push(loopOp, *fStatus); 1063 1064 // Then the LD_SP after the end of the loop 1065 op = URX_BUILD(URX_LD_SP, varLoc); 1066 fRXPat->fCompiledPat->addElement(op, *fStatus); 1067 } 1068 1069 break; 1070 1071 case doNGInterval: 1072 // Finished scanning a non-greedy {lower,upper}? interval. Generate the code for it. 1073 compileInterval(URX_CTR_INIT_NG, URX_CTR_LOOP_NG); 1074 break; 1075 1076 case doIntervalError: 1077 error(U_REGEX_BAD_INTERVAL); 1078 break; 1079 1080 case doLiteralChar: 1081 // We've just scanned a "normal" character from the pattern, 1082 literalChar(fC.fChar); 1083 break; 1084 1085 1086 case doEscapedLiteralChar: 1087 // We've just scanned an backslashed escaped character with no 1088 // special meaning. It represents itself. 1089 if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 && 1090 ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) || // in [A-Z] 1091 (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) { // in [a-z] 1092 error(U_REGEX_BAD_ESCAPE_SEQUENCE); 1093 } 1094 literalChar(fC.fChar); 1095 break; 1096 1097 1098 case doDotAny: 1099 // scanned a ".", match any single character. 1100 { 1101 int32_t op; 1102 if (fModeFlags & UREGEX_DOTALL) { 1103 op = URX_BUILD(URX_DOTANY_ALL, 0); 1104 } else if (fModeFlags & UREGEX_UNIX_LINES) { 1105 op = URX_BUILD(URX_DOTANY_UNIX, 0); 1106 } else { 1107 op = URX_BUILD(URX_DOTANY, 0); 1108 } 1109 fRXPat->fCompiledPat->addElement(op, *fStatus); 1110 } 1111 break; 1112 1113 case doCaret: 1114 { 1115 int32_t op = 0; 1116 if ( (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1117 op = URX_CARET; 1118 } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1119 op = URX_CARET_M; 1120 } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1121 op = URX_CARET; // Only testing true start of input. 1122 } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1123 op = URX_CARET_M_UNIX; 1124 } 1125 fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus); 1126 } 1127 break; 1128 1129 case doDollar: 1130 { 1131 int32_t op = 0; 1132 if ( (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1133 op = URX_DOLLAR; 1134 } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1135 op = URX_DOLLAR_M; 1136 } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1137 op = URX_DOLLAR_D; 1138 } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1139 op = URX_DOLLAR_MD; 1140 } 1141 fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus); 1142 } 1143 break; 1144 1145 case doBackslashA: 1146 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_CARET, 0), *fStatus); 1147 break; 1148 1149 case doBackslashB: 1150 { 1151 #if UCONFIG_NO_BREAK_ITERATION==1 1152 if (fModeFlags & UREGEX_UWORD) { 1153 error(U_UNSUPPORTED_ERROR); 1154 } 1155 #endif 1156 int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B; 1157 fRXPat->fCompiledPat->addElement(URX_BUILD(op, 1), *fStatus); 1158 } 1159 break; 1160 1161 case doBackslashb: 1162 { 1163 #if UCONFIG_NO_BREAK_ITERATION==1 1164 if (fModeFlags & UREGEX_UWORD) { 1165 error(U_UNSUPPORTED_ERROR); 1166 } 1167 #endif 1168 int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B; 1169 fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus); 1170 } 1171 break; 1172 1173 case doBackslashD: 1174 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_D, 1), *fStatus); 1175 break; 1176 1177 case doBackslashd: 1178 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_D, 0), *fStatus); 1179 break; 1180 1181 case doBackslashG: 1182 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_G, 0), *fStatus); 1183 break; 1184 1185 case doBackslashS: 1186 fRXPat->fCompiledPat->addElement( 1187 URX_BUILD(URX_STAT_SETREF_N, URX_ISSPACE_SET), *fStatus); 1188 break; 1189 1190 case doBackslashs: 1191 fRXPat->fCompiledPat->addElement( 1192 URX_BUILD(URX_STATIC_SETREF, URX_ISSPACE_SET), *fStatus); 1193 break; 1194 1195 case doBackslashW: 1196 fRXPat->fCompiledPat->addElement( 1197 URX_BUILD(URX_STAT_SETREF_N, URX_ISWORD_SET), *fStatus); 1198 break; 1199 1200 case doBackslashw: 1201 fRXPat->fCompiledPat->addElement( 1202 URX_BUILD(URX_STATIC_SETREF, URX_ISWORD_SET), *fStatus); 1203 break; 1204 1205 case doBackslashX: 1206 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_X, 0), *fStatus); 1207 break; 1208 1209 1210 case doBackslashZ: 1211 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_DOLLAR, 0), *fStatus); 1212 break; 1213 1214 case doBackslashz: 1215 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_Z, 0), *fStatus); 1216 break; 1217 1218 case doEscapeError: 1219 error(U_REGEX_BAD_ESCAPE_SEQUENCE); 1220 break; 1221 1222 case doExit: 1223 returnVal = FALSE; 1224 break; 1225 1226 case doProperty: 1227 { 1228 UnicodeSet *theSet = scanProp(); 1229 compileSet(theSet); 1230 } 1231 break; 1232 1233 case doNamedChar: 1234 { 1235 UChar32 c = scanNamedChar(); 1236 literalChar(c); 1237 } 1238 break; 1239 1240 1241 case doBackRef: 1242 // BackReference. Somewhat unusual in that the front-end can not completely parse 1243 // the regular expression, because the number of digits to be consumed 1244 // depends on the number of capture groups that have been defined. So 1245 // we have to do it here instead. 1246 { 1247 int32_t numCaptureGroups = fRXPat->fGroupMap->size(); 1248 int32_t groupNum = 0; 1249 UChar32 c = fC.fChar; 1250 1251 for (;;) { 1252 // Loop once per digit, for max allowed number of digits in a back reference. 1253 int32_t digit = u_charDigitValue(c); 1254 groupNum = groupNum * 10 + digit; 1255 if (groupNum >= numCaptureGroups) { 1256 break; 1257 } 1258 c = peekCharLL(); 1259 if (RegexStaticSets::gStaticSets->fRuleDigitsAlias->contains(c) == FALSE) { 1260 break; 1261 } 1262 nextCharLL(); 1263 } 1264 1265 // Scan of the back reference in the source regexp is complete. Now generate 1266 // the compiled code for it. 1267 // Because capture groups can be forward-referenced by back-references, 1268 // we fill the operand with the capture group number. At the end 1269 // of compilation, it will be changed to the variable's location. 1270 U_ASSERT(groupNum > 0); 1271 int32_t op; 1272 if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 1273 op = URX_BUILD(URX_BACKREF_I, groupNum); 1274 } else { 1275 op = URX_BUILD(URX_BACKREF, groupNum); 1276 } 1277 fRXPat->fCompiledPat->addElement(op, *fStatus); 1278 } 1279 break; 1280 1281 1282 case doPossessivePlus: 1283 // Possessive ++ quantifier. 1284 // Compiles to 1285 // 1. STO_SP 1286 // 2. body of stuff being iterated over 1287 // 3. STATE_SAVE 5 1288 // 4. JMP 2 1289 // 5. LD_SP 1290 // 6. ... 1291 // 1292 // Note: TODO: This is pretty inefficient. A mass of saved state is built up 1293 // then unconditionally discarded. Perhaps introduce a new opcode. Ticket 6056 1294 // 1295 { 1296 // Emit the STO_SP 1297 int32_t topLoc = blockTopLoc(TRUE); 1298 int32_t stoLoc = fRXPat->fDataSize; 1299 fRXPat->fDataSize++; // Reserve the data location for storing save stack ptr. 1300 int32_t op = URX_BUILD(URX_STO_SP, stoLoc); 1301 fRXPat->fCompiledPat->setElementAt(op, topLoc); 1302 1303 // Emit the STATE_SAVE 1304 op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+2); 1305 fRXPat->fCompiledPat->addElement(op, *fStatus); 1306 1307 // Emit the JMP 1308 op = URX_BUILD(URX_JMP, topLoc+1); 1309 fRXPat->fCompiledPat->addElement(op, *fStatus); 1310 1311 // Emit the LD_SP 1312 op = URX_BUILD(URX_LD_SP, stoLoc); 1313 fRXPat->fCompiledPat->addElement(op, *fStatus); 1314 } 1315 break; 1316 1317 case doPossessiveStar: 1318 // Possessive *+ quantifier. 1319 // Compiles to 1320 // 1. STO_SP loc 1321 // 2. STATE_SAVE 5 1322 // 3. body of stuff being iterated over 1323 // 4. JMP 2 1324 // 5. LD_SP loc 1325 // 6 ... 1326 // TODO: do something to cut back the state stack each time through the loop. 1327 { 1328 // Reserve two slots at the top of the block. 1329 int32_t topLoc = blockTopLoc(TRUE); 1330 insertOp(topLoc); 1331 1332 // emit STO_SP loc 1333 int32_t stoLoc = fRXPat->fDataSize; 1334 fRXPat->fDataSize++; // Reserve the data location for storing save stack ptr. 1335 int32_t op = URX_BUILD(URX_STO_SP, stoLoc); 1336 fRXPat->fCompiledPat->setElementAt(op, topLoc); 1337 1338 // Emit the SAVE_STATE 5 1339 int32_t L7 = fRXPat->fCompiledPat->size()+1; 1340 op = URX_BUILD(URX_STATE_SAVE, L7); 1341 fRXPat->fCompiledPat->setElementAt(op, topLoc+1); 1342 1343 // Append the JMP operation. 1344 op = URX_BUILD(URX_JMP, topLoc+1); 1345 fRXPat->fCompiledPat->addElement(op, *fStatus); 1346 1347 // Emit the LD_SP loc 1348 op = URX_BUILD(URX_LD_SP, stoLoc); 1349 fRXPat->fCompiledPat->addElement(op, *fStatus); 1350 } 1351 break; 1352 1353 case doPossessiveOpt: 1354 // Possessive ?+ quantifier. 1355 // Compiles to 1356 // 1. STO_SP loc 1357 // 2. SAVE_STATE 5 1358 // 3. body of optional block 1359 // 4. LD_SP loc 1360 // 5. ... 1361 // 1362 { 1363 // Reserve two slots at the top of the block. 1364 int32_t topLoc = blockTopLoc(TRUE); 1365 insertOp(topLoc); 1366 1367 // Emit the STO_SP 1368 int32_t stoLoc = fRXPat->fDataSize; 1369 fRXPat->fDataSize++; // Reserve the data location for storing save stack ptr. 1370 int32_t op = URX_BUILD(URX_STO_SP, stoLoc); 1371 fRXPat->fCompiledPat->setElementAt(op, topLoc); 1372 1373 // Emit the SAVE_STATE 1374 int32_t continueLoc = fRXPat->fCompiledPat->size()+1; 1375 op = URX_BUILD(URX_STATE_SAVE, continueLoc); 1376 fRXPat->fCompiledPat->setElementAt(op, topLoc+1); 1377 1378 // Emit the LD_SP 1379 op = URX_BUILD(URX_LD_SP, stoLoc); 1380 fRXPat->fCompiledPat->addElement(op, *fStatus); 1381 } 1382 break; 1383 1384 1385 case doBeginMatchMode: 1386 fNewModeFlags = fModeFlags; 1387 fSetModeFlag = TRUE; 1388 break; 1389 1390 case doMatchMode: // (?i) and similar 1391 { 1392 int32_t bit = 0; 1393 switch (fC.fChar) { 1394 case 0x69: /* 'i' */ bit = UREGEX_CASE_INSENSITIVE; break; 1395 case 0x64: /* 'd' */ bit = UREGEX_UNIX_LINES; break; 1396 case 0x6d: /* 'm' */ bit = UREGEX_MULTILINE; break; 1397 case 0x73: /* 's' */ bit = UREGEX_DOTALL; break; 1398 case 0x75: /* 'u' */ bit = 0; /* Unicode casing */ break; 1399 case 0x77: /* 'w' */ bit = UREGEX_UWORD; break; 1400 case 0x78: /* 'x' */ bit = UREGEX_COMMENTS; break; 1401 case 0x2d: /* '-' */ fSetModeFlag = FALSE; break; 1402 default: 1403 U_ASSERT(FALSE); // Should never happen. Other chars are filtered out 1404 // by the scanner. 1405 } 1406 if (fSetModeFlag) { 1407 fNewModeFlags |= bit; 1408 } else { 1409 fNewModeFlags &= ~bit; 1410 } 1411 } 1412 break; 1413 1414 case doSetMatchMode: 1415 // We've got a (?i) or similar. The match mode is being changed, but 1416 // the change is not scoped to a parenthesized block. 1417 U_ASSERT(fNewModeFlags < 0); 1418 fModeFlags = fNewModeFlags; 1419 1420 // Prevent any string from spanning across the change of match mode. 1421 // Otherwise the pattern "abc(?i)def" would make a single string of "abcdef" 1422 fixLiterals(); 1423 break; 1424 1425 1426 case doMatchModeParen: 1427 // We've got a (?i: or similar. Begin a parenthesized block, save old 1428 // mode flags so they can be restored at the close of the block. 1429 // 1430 // Compile to a 1431 // - NOP, which later may be replaced by a save-state if the 1432 // parenthesized group gets a * quantifier, followed by 1433 // - NOP, which may later be replaced by a save-state if there 1434 // is an '|' alternation within the parens. 1435 { 1436 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 1437 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 1438 1439 // On the Parentheses stack, start a new frame and add the postions 1440 // of the two NOPs (a normal non-capturing () frame, except for the 1441 // saving of the orignal mode flags.) 1442 fParenStack.push(fModeFlags, *fStatus); 1443 fParenStack.push(flags, *fStatus); // Frame Marker 1444 fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP 1445 fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP 1446 1447 // Set the current mode flags to the new values. 1448 U_ASSERT(fNewModeFlags < 0); 1449 fModeFlags = fNewModeFlags; 1450 } 1451 break; 1452 1453 case doBadModeFlag: 1454 error(U_REGEX_INVALID_FLAG); 1455 break; 1456 1457 case doSuppressComments: 1458 // We have just scanned a '(?'. We now need to prevent the character scanner from 1459 // treating a '#' as a to-the-end-of-line comment. 1460 // (This Perl compatibility just gets uglier and uglier to do...) 1461 fEOLComments = FALSE; 1462 break; 1463 1464 1465 case doSetAddAmp: 1466 { 1467 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1468 set->add(chAmp); 1469 } 1470 break; 1471 1472 case doSetAddDash: 1473 { 1474 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1475 set->add(chDash); 1476 } 1477 break; 1478 1479 case doSetBackslash_s: 1480 { 1481 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1482 set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]); 1483 break; 1484 } 1485 1486 case doSetBackslash_S: 1487 { 1488 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1489 UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]); 1490 SSet.complement(); 1491 set->addAll(SSet); 1492 break; 1493 } 1494 1495 case doSetBackslash_d: 1496 { 1497 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1498 // TODO - make a static set, ticket 6058. 1499 addCategory(set, U_GC_ND_MASK, *fStatus); 1500 break; 1501 } 1502 1503 case doSetBackslash_D: 1504 { 1505 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1506 UnicodeSet digits; 1507 // TODO - make a static set, ticket 6058. 1508 digits.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus); 1509 digits.complement(); 1510 set->addAll(digits); 1511 break; 1512 } 1513 1514 case doSetBackslash_w: 1515 { 1516 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1517 set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]); 1518 break; 1519 } 1520 1521 case doSetBackslash_W: 1522 { 1523 UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1524 UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]); 1525 SSet.complement(); 1526 set->addAll(SSet); 1527 break; 1528 } 1529 1530 case doSetBegin: 1531 fSetStack.push(new UnicodeSet(), *fStatus); 1532 fSetOpStack.push(setStart, *fStatus); 1533 if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1534 fSetOpStack.push(setCaseClose, *fStatus); 1535 } 1536 break; 1537 1538 case doSetBeginDifference1: 1539 // We have scanned something like [[abc]-[ 1540 // Set up a new UnicodeSet for the set beginning with the just-scanned '[' 1541 // Push a Difference operator, which will cause the new set to be subtracted from what 1542 // went before once it is created. 1543 setPushOp(setDifference1); 1544 fSetOpStack.push(setStart, *fStatus); 1545 if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1546 fSetOpStack.push(setCaseClose, *fStatus); 1547 } 1548 break; 1549 1550 case doSetBeginIntersection1: 1551 // We have scanned something like [[abc]&[ 1552 // Need both the '&' operator and the open '[' operator. 1553 setPushOp(setIntersection1); 1554 fSetOpStack.push(setStart, *fStatus); 1555 if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1556 fSetOpStack.push(setCaseClose, *fStatus); 1557 } 1558 break; 1559 1560 case doSetBeginUnion: 1561 // We have scanned something like [[abc][ 1562 // Need to handle the union operation explicitly [[abc] | [ 1563 setPushOp(setUnion); 1564 fSetOpStack.push(setStart, *fStatus); 1565 if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1566 fSetOpStack.push(setCaseClose, *fStatus); 1567 } 1568 break; 1569 1570 case doSetDifference2: 1571 // We have scanned something like [abc-- 1572 // Consider this to unambiguously be a set difference operator. 1573 setPushOp(setDifference2); 1574 break; 1575 1576 case doSetEnd: 1577 // Have encountered the ']' that closes a set. 1578 // Force the evaluation of any pending operations within this set, 1579 // leave the completed set on the top of the set stack. 1580 setEval(setEnd); 1581 U_ASSERT(fSetOpStack.peeki()==setStart); 1582 fSetOpStack.popi(); 1583 break; 1584 1585 case doSetFinish: 1586 { 1587 // Finished a complete set expression, including all nested sets. 1588 // The close bracket has already triggered clearing out pending set operators, 1589 // the operator stack should be empty and the operand stack should have just 1590 // one entry, the result set. 1591 U_ASSERT(fSetOpStack.empty()); 1592 UnicodeSet *theSet = (UnicodeSet *)fSetStack.pop(); 1593 U_ASSERT(fSetStack.empty()); 1594 compileSet(theSet); 1595 break; 1596 } 1597 1598 case doSetIntersection2: 1599 // Have scanned something like [abc&& 1600 setPushOp(setIntersection2); 1601 break; 1602 1603 case doSetLiteral: 1604 // Union the just-scanned literal character into the set being built. 1605 // This operation is the highest precedence set operation, so we can always do 1606 // it immediately, without waiting to see what follows. It is necessary to perform 1607 // any pending '-' or '&' operation first, because these have the same precedence 1608 // as union-ing in a literal' 1609 { 1610 setEval(setUnion); 1611 UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1612 s->add(fC.fChar); 1613 fLastSetLiteral = fC.fChar; 1614 break; 1615 } 1616 1617 case doSetLiteralEscaped: 1618 // A back-slash escaped literal character was encountered. 1619 // Processing is the same as with setLiteral, above, with the addition of 1620 // the optional check for errors on escaped ASCII letters. 1621 { 1622 if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 && 1623 ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) || // in [A-Z] 1624 (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) { // in [a-z] 1625 error(U_REGEX_BAD_ESCAPE_SEQUENCE); 1626 } 1627 setEval(setUnion); 1628 UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1629 s->add(fC.fChar); 1630 fLastSetLiteral = fC.fChar; 1631 break; 1632 } 1633 1634 case doSetNamedChar: 1635 // Scanning a \N{UNICODE CHARACTER NAME} 1636 // Aside from the source of the character, the processing is identical to doSetLiteral, 1637 // above. 1638 { 1639 UChar32 c = scanNamedChar(); 1640 setEval(setUnion); 1641 UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1642 s->add(c); 1643 fLastSetLiteral = c; 1644 break; 1645 } 1646 1647 case doSetNamedRange: 1648 // We have scanned literal-\N{CHAR NAME}. Add the range to the set. 1649 // The left character is already in the set, and is saved in fLastSetLiteral. 1650 // The right side needs to be picked up, the scan is at the 'N'. 1651 // Lower Limit > Upper limit being an error matches both Java 1652 // and ICU UnicodeSet behavior. 1653 { 1654 UChar32 c = scanNamedChar(); 1655 if (U_SUCCESS(*fStatus) && fLastSetLiteral > c) { 1656 error(U_REGEX_INVALID_RANGE); 1657 } 1658 UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1659 s->add(fLastSetLiteral, c); 1660 fLastSetLiteral = c; 1661 break; 1662 } 1663 1664 1665 case doSetNegate: 1666 // Scanned a '^' at the start of a set. 1667 // Push the negation operator onto the set op stack. 1668 // A twist for case-insensitive matching: 1669 // the case closure operation must happen _before_ negation. 1670 // But the case closure operation will already be on the stack if it's required. 1671 // This requires checking for case closure, and swapping the stack order 1672 // if it is present. 1673 { 1674 int32_t tosOp = fSetOpStack.peeki(); 1675 if (tosOp == setCaseClose) { 1676 fSetOpStack.popi(); 1677 fSetOpStack.push(setNegation, *fStatus); 1678 fSetOpStack.push(setCaseClose, *fStatus); 1679 } else { 1680 fSetOpStack.push(setNegation, *fStatus); 1681 } 1682 } 1683 break; 1684 1685 case doSetNoCloseError: 1686 error(U_REGEX_MISSING_CLOSE_BRACKET); 1687 break; 1688 1689 case doSetOpError: 1690 error(U_REGEX_RULE_SYNTAX); // -- or && at the end of a set. Illegal. 1691 break; 1692 1693 case doSetPosixProp: 1694 { 1695 UnicodeSet *s = scanPosixProp(); 1696 if (s != NULL) { 1697 UnicodeSet *tos = (UnicodeSet *)fSetStack.peek(); 1698 tos->addAll(*s); 1699 delete s; 1700 } // else error. scanProp() reported the error status already. 1701 } 1702 break; 1703 1704 case doSetProp: 1705 // Scanned a \p \P within [brackets]. 1706 { 1707 UnicodeSet *s = scanProp(); 1708 if (s != NULL) { 1709 UnicodeSet *tos = (UnicodeSet *)fSetStack.peek(); 1710 tos->addAll(*s); 1711 delete s; 1712 } // else error. scanProp() reported the error status already. 1713 } 1714 break; 1715 1716 1717 case doSetRange: 1718 // We have scanned literal-literal. Add the range to the set. 1719 // The left character is already in the set, and is saved in fLastSetLiteral. 1720 // The right side is the current character. 1721 // Lower Limit > Upper limit being an error matches both Java 1722 // and ICU UnicodeSet behavior. 1723 { 1724 if (fLastSetLiteral > fC.fChar) { 1725 error(U_REGEX_INVALID_RANGE); 1726 } 1727 UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1728 s->add(fLastSetLiteral, fC.fChar); 1729 break; 1730 } 1731 1732 1733 default: 1734 U_ASSERT(FALSE); 1735 error(U_REGEX_INTERNAL_ERROR); 1736 break; 1737 } 1738 1739 if (U_FAILURE(*fStatus)) { 1740 returnVal = FALSE; 1741 } 1742 1743 return returnVal; 1744 } 1745 1746 1747 1748 //------------------------------------------------------------------------------ 1749 // 1750 // literalChar We've encountered a literal character from the pattern, 1751 // or an escape sequence that reduces to a character. 1752 // Add it to the string containing all literal chars/strings from 1753 // the pattern. 1754 // If we are in a pattern string already, add the new char to it. 1755 // If we aren't in a pattern string, begin one now. 1756 // 1757 //------------------------------------------------------------------------------ 1758 void RegexCompile::literalChar(UChar32 c) { 1759 int32_t op; // An operation in the compiled pattern. 1760 int32_t opType; 1761 int32_t patternLoc; // A position in the compiled pattern. 1762 int32_t stringLen; 1763 1764 1765 // If the last thing compiled into the pattern was not a literal char, 1766 // force this new literal char to begin a new string, and not append to the previous. 1767 op = fRXPat->fCompiledPat->lastElementi(); 1768 opType = URX_TYPE(op); 1769 if (!(opType == URX_STRING_LEN || opType == URX_ONECHAR || opType == URX_ONECHAR_I)) { 1770 fixLiterals(); 1771 } 1772 1773 if (fStringOpStart == -1) { 1774 // First char of a string in the pattern. 1775 // Emit a OneChar op into the compiled pattern. 1776 emitONE_CHAR(c); 1777 1778 // Also add it to the string pool, in case we get a second adjacent literal 1779 // and want to change form ONE_CHAR to STRING 1780 fStringOpStart = fRXPat->fLiteralText.length(); 1781 fRXPat->fLiteralText.append(c); 1782 return; 1783 } 1784 1785 // We are adding onto an existing string 1786 fRXPat->fLiteralText.append(c); 1787 1788 op = fRXPat->fCompiledPat->lastElementi(); 1789 opType = URX_TYPE(op); 1790 U_ASSERT(opType == URX_ONECHAR || opType == URX_ONECHAR_I || opType == URX_STRING_LEN); 1791 1792 // If the most recently emitted op is a URX_ONECHAR, 1793 if (opType == URX_ONECHAR || opType == URX_ONECHAR_I) { 1794 if (U16_IS_TRAIL(c) && U16_IS_LEAD(URX_VAL(op))) { 1795 // The most recently emitted op is a ONECHAR that was the first half 1796 // of a surrogate pair. Update the ONECHAR's operand to be the 1797 // supplementary code point resulting from both halves of the pair. 1798 c = U16_GET_SUPPLEMENTARY(URX_VAL(op), c); 1799 op = URX_BUILD(opType, c); 1800 patternLoc = fRXPat->fCompiledPat->size() - 1; 1801 fRXPat->fCompiledPat->setElementAt(op, patternLoc); 1802 return; 1803 } 1804 1805 // The most recently emitted op is a ONECHAR. 1806 // We've now received another adjacent char. Change the ONECHAR op 1807 // to a string op. 1808 if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 1809 op = URX_BUILD(URX_STRING_I, fStringOpStart); 1810 } else { 1811 op = URX_BUILD(URX_STRING, fStringOpStart); 1812 } 1813 patternLoc = fRXPat->fCompiledPat->size() - 1; 1814 fRXPat->fCompiledPat->setElementAt(op, patternLoc); 1815 op = URX_BUILD(URX_STRING_LEN, 0); 1816 fRXPat->fCompiledPat->addElement(op, *fStatus); 1817 } 1818 1819 // The pattern contains a URX_SRING / URX_STRING_LEN. Update the 1820 // string length to reflect the new char we just added to the string. 1821 stringLen = fRXPat->fLiteralText.length() - fStringOpStart; 1822 op = URX_BUILD(URX_STRING_LEN, stringLen); 1823 patternLoc = fRXPat->fCompiledPat->size() - 1; 1824 fRXPat->fCompiledPat->setElementAt(op, patternLoc); 1825 } 1826 1827 1828 1829 //------------------------------------------------------------------------------ 1830 // 1831 // emitONE_CHAR emit a ONE_CHAR op into the generated code. 1832 // Choose cased or uncased version, depending on the 1833 // match mode and whether the character itself is cased. 1834 // 1835 //------------------------------------------------------------------------------ 1836 void RegexCompile::emitONE_CHAR(UChar32 c) { 1837 int32_t op; 1838 if ((fModeFlags & UREGEX_CASE_INSENSITIVE) && 1839 u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) { 1840 // We have a cased character, and are in case insensitive matching mode. 1841 c = u_foldCase(c, U_FOLD_CASE_DEFAULT); 1842 op = URX_BUILD(URX_ONECHAR_I, c); 1843 } else { 1844 // Uncased char, or case sensitive match mode. 1845 // Either way, just generate a literal compare of the char. 1846 op = URX_BUILD(URX_ONECHAR, c); 1847 } 1848 fRXPat->fCompiledPat->addElement(op, *fStatus); 1849 } 1850 1851 1852 //------------------------------------------------------------------------------ 1853 // 1854 // fixLiterals When compiling something that can follow a literal 1855 // string in a pattern, we need to "fix" any preceding 1856 // string, which will cause any subsequent literals to 1857 // begin a new string, rather than appending to the 1858 // old one. 1859 // 1860 // Optionally, split the last char of the string off into 1861 // a single "ONE_CHAR" operation, so that quantifiers can 1862 // apply to that char alone. Example: abc* 1863 // The * must apply to the 'c' only. 1864 // 1865 //------------------------------------------------------------------------------ 1866 void RegexCompile::fixLiterals(UBool split) { 1867 int32_t stringStart = fStringOpStart; // start index of the current literal string 1868 int32_t op; // An op from/for the compiled pattern. 1869 int32_t opType; // An opcode type from the compiled pattern. 1870 int32_t stringLastCharIdx; 1871 UChar32 lastChar; 1872 int32_t stringNextToLastCharIdx; 1873 UChar32 nextToLastChar; 1874 int32_t stringLen; 1875 1876 fStringOpStart = -1; 1877 if (!split) { 1878 return; 1879 } 1880 1881 // Split: We need to ensure that the last item in the compiled pattern does 1882 // not refer to a literal string of more than one char. If it does, 1883 // separate the last char from the rest of the string. 1884 1885 // If the last operation from the compiled pattern is not a string, 1886 // nothing needs to be done 1887 op = fRXPat->fCompiledPat->lastElementi(); 1888 opType = URX_TYPE(op); 1889 if (opType != URX_STRING_LEN) { 1890 return; 1891 } 1892 stringLen = URX_VAL(op); 1893 1894 // 1895 // Find the position of the last code point in the string (might be a surrogate pair) 1896 // 1897 stringLastCharIdx = fRXPat->fLiteralText.length(); 1898 stringLastCharIdx = fRXPat->fLiteralText.moveIndex32(stringLastCharIdx, -1); 1899 lastChar = fRXPat->fLiteralText.char32At(stringLastCharIdx); 1900 1901 // The string should always be at least two code points long, meaning that there 1902 // should be something before the last char position that we just found. 1903 U_ASSERT(stringLastCharIdx > stringStart); 1904 stringNextToLastCharIdx = fRXPat->fLiteralText.moveIndex32(stringLastCharIdx, -1); 1905 U_ASSERT(stringNextToLastCharIdx >= stringStart); 1906 nextToLastChar = fRXPat->fLiteralText.char32At(stringNextToLastCharIdx); 1907 1908 if (stringNextToLastCharIdx > stringStart) { 1909 // The length of string remaining after removing one char is two or more. 1910 // Leave the string in the compiled pattern, shorten it by one char, 1911 // and append a URX_ONECHAR op for the last char. 1912 stringLen -= (fRXPat->fLiteralText.length() - stringLastCharIdx); 1913 op = URX_BUILD(URX_STRING_LEN, stringLen); 1914 fRXPat->fCompiledPat->setElementAt(op, fRXPat->fCompiledPat->size() -1); 1915 emitONE_CHAR(lastChar); 1916 } else { 1917 // The original string consisted of exactly two characters. Replace 1918 // the existing compiled URX_STRING/URX_STRING_LEN ops with a pair 1919 // of URX_ONECHARs. 1920 fRXPat->fCompiledPat->setSize(fRXPat->fCompiledPat->size() -2); 1921 emitONE_CHAR(nextToLastChar); 1922 emitONE_CHAR(lastChar); 1923 } 1924 } 1925 1926 1927 1928 1929 1930 1931 //------------------------------------------------------------------------------ 1932 // 1933 // insertOp() Insert a slot for a new opcode into the already 1934 // compiled pattern code. 1935 // 1936 // Fill the slot with a NOP. Our caller will replace it 1937 // with what they really wanted. 1938 // 1939 //------------------------------------------------------------------------------ 1940 void RegexCompile::insertOp(int32_t where) { 1941 UVector32 *code = fRXPat->fCompiledPat; 1942 U_ASSERT(where>0 && where < code->size()); 1943 1944 int32_t nop = URX_BUILD(URX_NOP, 0); 1945 code->insertElementAt(nop, where, *fStatus); 1946 1947 // Walk through the pattern, looking for any ops with targets that 1948 // were moved down by the insert. Fix them. 1949 int32_t loc; 1950 for (loc=0; loc<code->size(); loc++) { 1951 int32_t op = code->elementAti(loc); 1952 int32_t opType = URX_TYPE(op); 1953 int32_t opValue = URX_VAL(op); 1954 if ((opType == URX_JMP || 1955 opType == URX_JMPX || 1956 opType == URX_STATE_SAVE || 1957 opType == URX_CTR_LOOP || 1958 opType == URX_CTR_LOOP_NG || 1959 opType == URX_JMP_SAV || 1960 opType == URX_RELOC_OPRND) && opValue > where) { 1961 // Target location for this opcode is after the insertion point and 1962 // needs to be incremented to adjust for the insertion. 1963 opValue++; 1964 op = URX_BUILD(opType, opValue); 1965 code->setElementAt(op, loc); 1966 } 1967 } 1968 1969 // Now fix up the parentheses stack. All positive values in it are locations in 1970 // the compiled pattern. (Negative values are frame boundaries, and don't need fixing.) 1971 for (loc=0; loc<fParenStack.size(); loc++) { 1972 int32_t x = fParenStack.elementAti(loc); 1973 U_ASSERT(x < code->size()); 1974 if (x>where) { 1975 x++; 1976 fParenStack.setElementAt(x, loc); 1977 } 1978 } 1979 1980 if (fMatchCloseParen > where) { 1981 fMatchCloseParen++; 1982 } 1983 if (fMatchOpenParen > where) { 1984 fMatchOpenParen++; 1985 } 1986 } 1987 1988 1989 1990 //------------------------------------------------------------------------------ 1991 // 1992 // blockTopLoc() Find or create a location in the compiled pattern 1993 // at the start of the operation or block that has 1994 // just been compiled. Needed when a quantifier (* or 1995 // whatever) appears, and we need to add an operation 1996 // at the start of the thing being quantified. 1997 // 1998 // (Parenthesized Blocks) have a slot with a NOP that 1999 // is reserved for this purpose. .* or similar don't 2000 // and a slot needs to be added. 2001 // 2002 // parameter reserveLoc : TRUE - ensure that there is space to add an opcode 2003 // at the returned location. 2004 // FALSE - just return the address, 2005 // do not reserve a location there. 2006 // 2007 //------------------------------------------------------------------------------ 2008 int32_t RegexCompile::blockTopLoc(UBool reserveLoc) { 2009 int32_t theLoc; 2010 if (fRXPat->fCompiledPat->size() == fMatchCloseParen) 2011 { 2012 // The item just processed is a parenthesized block. 2013 theLoc = fMatchOpenParen; // A slot is already reserved for us. 2014 U_ASSERT(theLoc > 0); 2015 U_ASSERT(URX_TYPE(((uint32_t)fRXPat->fCompiledPat->elementAti(theLoc))) == URX_NOP); 2016 } 2017 else { 2018 // Item just compiled is a single thing, a ".", or a single char, or a set reference. 2019 // No slot for STATE_SAVE was pre-reserved in the compiled code. 2020 // We need to make space now. 2021 fixLiterals(TRUE); // If last item was a string, separate the last char. 2022 theLoc = fRXPat->fCompiledPat->size()-1; 2023 if (reserveLoc) { 2024 /*int32_t opAtTheLoc = fRXPat->fCompiledPat->elementAti(theLoc);*/ 2025 int32_t nop = URX_BUILD(URX_NOP, 0); 2026 fRXPat->fCompiledPat->insertElementAt(nop, theLoc, *fStatus); 2027 } 2028 } 2029 return theLoc; 2030 } 2031 2032 2033 2034 //------------------------------------------------------------------------------ 2035 // 2036 // handleCloseParen When compiling a close paren, we need to go back 2037 // and fix up any JMP or SAVE operations within the 2038 // parenthesized block that need to target the end 2039 // of the block. The locations of these are kept on 2040 // the paretheses stack. 2041 // 2042 // This function is called both when encountering a 2043 // real ) and at the end of the pattern. 2044 // 2045 //------------------------------------------------------------------------------ 2046 void RegexCompile::handleCloseParen() { 2047 int32_t patIdx; 2048 int32_t patOp; 2049 if (fParenStack.size() <= 0) { 2050 error(U_REGEX_MISMATCHED_PAREN); 2051 return; 2052 } 2053 2054 // Force any literal chars that may follow the close paren to start a new string, 2055 // and not attach to any preceding it. 2056 fixLiterals(FALSE); 2057 2058 // Fixup any operations within the just-closed parenthesized group 2059 // that need to reference the end of the (block). 2060 // (The first one popped from the stack is an unused slot for 2061 // alternation (OR) state save, but applying the fixup to it does no harm.) 2062 for (;;) { 2063 patIdx = fParenStack.popi(); 2064 if (patIdx < 0) { 2065 // value < 0 flags the start of the frame on the paren stack. 2066 break; 2067 } 2068 U_ASSERT(patIdx>0 && patIdx <= fRXPat->fCompiledPat->size()); 2069 patOp = fRXPat->fCompiledPat->elementAti(patIdx); 2070 U_ASSERT(URX_VAL(patOp) == 0); // Branch target for JMP should not be set. 2071 patOp |= fRXPat->fCompiledPat->size(); // Set it now. 2072 fRXPat->fCompiledPat->setElementAt(patOp, patIdx); 2073 fMatchOpenParen = patIdx; 2074 } 2075 2076 // At the close of any parenthesized block, restore the match mode flags to 2077 // the value they had at the open paren. Saved value is 2078 // at the top of the paren stack. 2079 fModeFlags = fParenStack.popi(); 2080 U_ASSERT(fModeFlags < 0); 2081 2082 // DO any additional fixups, depending on the specific kind of 2083 // parentesized grouping this is 2084 2085 switch (patIdx) { 2086 case plain: 2087 case flags: 2088 // No additional fixups required. 2089 // (Grouping-only parentheses) 2090 break; 2091 case capturing: 2092 // Capturing Parentheses. 2093 // Insert a End Capture op into the pattern. 2094 // The frame offset of the variables for this cg is obtained from the 2095 // start capture op and put it into the end-capture op. 2096 { 2097 int32_t captureOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1); 2098 U_ASSERT(URX_TYPE(captureOp) == URX_START_CAPTURE); 2099 2100 int32_t frameVarLocation = URX_VAL(captureOp); 2101 int32_t endCaptureOp = URX_BUILD(URX_END_CAPTURE, frameVarLocation); 2102 fRXPat->fCompiledPat->addElement(endCaptureOp, *fStatus); 2103 } 2104 break; 2105 case atomic: 2106 // Atomic Parenthesis. 2107 // Insert a LD_SP operation to restore the state stack to the position 2108 // it was when the atomic parens were entered. 2109 { 2110 int32_t stoOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1); 2111 U_ASSERT(URX_TYPE(stoOp) == URX_STO_SP); 2112 int32_t stoLoc = URX_VAL(stoOp); 2113 int32_t ldOp = URX_BUILD(URX_LD_SP, stoLoc); 2114 fRXPat->fCompiledPat->addElement(ldOp, *fStatus); 2115 } 2116 break; 2117 2118 case lookAhead: 2119 { 2120 int32_t startOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5); 2121 U_ASSERT(URX_TYPE(startOp) == URX_LA_START); 2122 int32_t dataLoc = URX_VAL(startOp); 2123 int32_t op = URX_BUILD(URX_LA_END, dataLoc); 2124 fRXPat->fCompiledPat->addElement(op, *fStatus); 2125 } 2126 break; 2127 2128 case negLookAhead: 2129 { 2130 // See comment at doOpenLookAheadNeg 2131 int32_t startOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen-1); 2132 U_ASSERT(URX_TYPE(startOp) == URX_LA_START); 2133 int32_t dataLoc = URX_VAL(startOp); 2134 int32_t op = URX_BUILD(URX_LA_END, dataLoc); 2135 fRXPat->fCompiledPat->addElement(op, *fStatus); 2136 op = URX_BUILD(URX_BACKTRACK, 0); 2137 fRXPat->fCompiledPat->addElement(op, *fStatus); 2138 // BEGIN android-changed 2139 // Apply important regex bug fixing for ticket#7224 2140 op = URX_BUILD(URX_LA_END, dataLoc); 2141 // END android-changed 2142 fRXPat->fCompiledPat->addElement(op, *fStatus); 2143 2144 // Patch the URX_SAVE near the top of the block. 2145 // The destination of the SAVE is the final LA_END that was just added. 2146 int32_t saveOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen); 2147 U_ASSERT(URX_TYPE(saveOp) == URX_STATE_SAVE); 2148 int32_t dest = fRXPat->fCompiledPat->size()-1; 2149 saveOp = URX_BUILD(URX_STATE_SAVE, dest); 2150 fRXPat->fCompiledPat->setElementAt(saveOp, fMatchOpenParen); 2151 } 2152 break; 2153 2154 case lookBehind: 2155 { 2156 // See comment at doOpenLookBehind. 2157 2158 // Append the URX_LB_END and URX_LA_END to the compiled pattern. 2159 int32_t startOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen-4); 2160 U_ASSERT(URX_TYPE(startOp) == URX_LB_START); 2161 int32_t dataLoc = URX_VAL(startOp); 2162 int32_t op = URX_BUILD(URX_LB_END, dataLoc); 2163 fRXPat->fCompiledPat->addElement(op, *fStatus); 2164 op = URX_BUILD(URX_LA_END, dataLoc); 2165 fRXPat->fCompiledPat->addElement(op, *fStatus); 2166 2167 // Determine the min and max bounds for the length of the 2168 // string that the pattern can match. 2169 // An unbounded upper limit is an error. 2170 int32_t patEnd = fRXPat->fCompiledPat->size() - 1; 2171 int32_t minML = minMatchLength(fMatchOpenParen, patEnd); 2172 int32_t maxML = maxMatchLength(fMatchOpenParen, patEnd); 2173 if (maxML == INT32_MAX) { 2174 error(U_REGEX_LOOK_BEHIND_LIMIT); 2175 break; 2176 } 2177 U_ASSERT(minML <= maxML); 2178 2179 // Insert the min and max match len bounds into the URX_LB_CONT op that 2180 // appears at the top of the look-behind block, at location fMatchOpenParen+1 2181 fRXPat->fCompiledPat->setElementAt(minML, fMatchOpenParen-2); 2182 fRXPat->fCompiledPat->setElementAt(maxML, fMatchOpenParen-1); 2183 2184 } 2185 break; 2186 2187 2188 2189 case lookBehindN: 2190 { 2191 // See comment at doOpenLookBehindNeg. 2192 2193 // Append the URX_LBN_END to the compiled pattern. 2194 int32_t startOp = fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5); 2195 U_ASSERT(URX_TYPE(startOp) == URX_LB_START); 2196 int32_t dataLoc = URX_VAL(startOp); 2197 int32_t op = URX_BUILD(URX_LBN_END, dataLoc); 2198 fRXPat->fCompiledPat->addElement(op, *fStatus); 2199 2200 // Determine the min and max bounds for the length of the 2201 // string that the pattern can match. 2202 // An unbounded upper limit is an error. 2203 int32_t patEnd = fRXPat->fCompiledPat->size() - 1; 2204 int32_t minML = minMatchLength(fMatchOpenParen, patEnd); 2205 int32_t maxML = maxMatchLength(fMatchOpenParen, patEnd); 2206 if (maxML == INT32_MAX) { 2207 error(U_REGEX_LOOK_BEHIND_LIMIT); 2208 break; 2209 } 2210 U_ASSERT(minML <= maxML); 2211 2212 // Insert the min and max match len bounds into the URX_LB_CONT op that 2213 // appears at the top of the look-behind block, at location fMatchOpenParen+1 2214 fRXPat->fCompiledPat->setElementAt(minML, fMatchOpenParen-3); 2215 fRXPat->fCompiledPat->setElementAt(maxML, fMatchOpenParen-2); 2216 2217 // Insert the pattern location to continue at after a successful match 2218 // as the last operand of the URX_LBN_CONT 2219 op = URX_BUILD(URX_RELOC_OPRND, fRXPat->fCompiledPat->size()); 2220 fRXPat->fCompiledPat->setElementAt(op, fMatchOpenParen-1); 2221 } 2222 break; 2223 2224 2225 2226 default: 2227 U_ASSERT(FALSE); 2228 } 2229 2230 // remember the next location in the compiled pattern. 2231 // The compilation of Quantifiers will look at this to see whether its looping 2232 // over a parenthesized block or a single item 2233 fMatchCloseParen = fRXPat->fCompiledPat->size(); 2234 } 2235 2236 2237 2238 //------------------------------------------------------------------------------ 2239 // 2240 // compileSet Compile the pattern operations for a reference to a 2241 // UnicodeSet. 2242 // 2243 //------------------------------------------------------------------------------ 2244 void RegexCompile::compileSet(UnicodeSet *theSet) 2245 { 2246 if (theSet == NULL) { 2247 return; 2248 } 2249 // Remove any strings from the set. 2250 // There shoudn't be any, but just in case. 2251 // (Case Closure can add them; if we had a simple case closure avaialble that 2252 // ignored strings, that would be better.) 2253 theSet->removeAllStrings(); 2254 int32_t setSize = theSet->size(); 2255 UChar32 firstSetChar = theSet->charAt(0); 2256 2257 switch (setSize) { 2258 case 0: 2259 { 2260 // Set of no elements. Always fails to match. 2261 fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKTRACK, 0), *fStatus); 2262 delete theSet; 2263 } 2264 break; 2265 2266 case 1: 2267 { 2268 // The set contains only a single code point. Put it into 2269 // the compiled pattern as a single char operation rather 2270 // than a set, and discard the set itself. 2271 literalChar(firstSetChar); 2272 delete theSet; 2273 } 2274 break; 2275 2276 default: 2277 { 2278 // The set contains two or more chars. (the normal case) 2279 // Put it into the compiled pattern as a set. 2280 int32_t setNumber = fRXPat->fSets->size(); 2281 fRXPat->fSets->addElement(theSet, *fStatus); 2282 int32_t setOp = URX_BUILD(URX_SETREF, setNumber); 2283 fRXPat->fCompiledPat->addElement(setOp, *fStatus); 2284 } 2285 } 2286 } 2287 2288 2289 //------------------------------------------------------------------------------ 2290 // 2291 // compileInterval Generate the code for a {min, max} style interval quantifier. 2292 // Except for the specific opcodes used, the code is the same 2293 // for all three types (greedy, non-greedy, possessive) of 2294 // intervals. The opcodes are supplied as parameters. 2295 // 2296 // The code for interval loops has this form: 2297 // 0 CTR_INIT counter loc (in stack frame) 2298 // 1 5 patt address of CTR_LOOP at bottom of block 2299 // 2 min count 2300 // 3 max count (-1 for unbounded) 2301 // 4 ... block to be iterated over 2302 // 5 CTR_LOOP 2303 // 2304 // In 2305 //------------------------------------------------------------------------------ 2306 void RegexCompile::compileInterval(int32_t InitOp, int32_t LoopOp) 2307 { 2308 // The CTR_INIT op at the top of the block with the {n,m} quantifier takes 2309 // four slots in the compiled code. Reserve them. 2310 int32_t topOfBlock = blockTopLoc(TRUE); 2311 insertOp(topOfBlock); 2312 insertOp(topOfBlock); 2313 insertOp(topOfBlock); 2314 2315 // The operands for the CTR_INIT opcode include the index in the matcher data 2316 // of the counter. Allocate it now. 2317 int32_t counterLoc = fRXPat->fFrameSize; 2318 fRXPat->fFrameSize++; 2319 2320 int32_t op = URX_BUILD(InitOp, counterLoc); 2321 fRXPat->fCompiledPat->setElementAt(op, topOfBlock); 2322 2323 // The second operand of CTR_INIT is the location following the end of the loop. 2324 // Must put in as a URX_RELOC_OPRND so that the value will be adjusted if the 2325 // compilation of something later on causes the code to grow and the target 2326 // position to move. 2327 int32_t loopEnd = fRXPat->fCompiledPat->size(); 2328 op = URX_BUILD(URX_RELOC_OPRND, loopEnd); 2329 fRXPat->fCompiledPat->setElementAt(op, topOfBlock+1); 2330 2331 // Followed by the min and max counts. 2332 fRXPat->fCompiledPat->setElementAt(fIntervalLow, topOfBlock+2); 2333 fRXPat->fCompiledPat->setElementAt(fIntervalUpper, topOfBlock+3); 2334 2335 // Apend the CTR_LOOP op. The operand is the location of the CTR_INIT op. 2336 // Goes at end of the block being looped over, so just append to the code so far. 2337 op = URX_BUILD(LoopOp, topOfBlock); 2338 fRXPat->fCompiledPat->addElement(op, *fStatus); 2339 2340 if ((fIntervalLow & 0xff000000) != 0 || 2341 fIntervalUpper > 0 && (fIntervalUpper & 0xff000000) != 0) { 2342 error(U_REGEX_NUMBER_TOO_BIG); 2343 } 2344 2345 if (fIntervalLow > fIntervalUpper && fIntervalUpper != -1) { 2346 error(U_REGEX_MAX_LT_MIN); 2347 } 2348 } 2349 2350 2351 2352 UBool RegexCompile::compileInlineInterval() { 2353 if (fIntervalUpper > 10 || fIntervalUpper < fIntervalLow) { 2354 // Too big to inline. Fail, which will cause looping code to be generated. 2355 // (Upper < Lower picks up unbounded upper and errors, both.) 2356 return FALSE; 2357 } 2358 2359 int32_t topOfBlock = blockTopLoc(FALSE); 2360 if (fIntervalUpper == 0) { 2361 // Pathological case. Attempt no matches, as if the block doesn't exist. 2362 fRXPat->fCompiledPat->setSize(topOfBlock); 2363 return TRUE; 2364 } 2365 2366 if (topOfBlock != fRXPat->fCompiledPat->size()-1 && fIntervalUpper != 1) { 2367 // The thing being repeated is not a single op, but some 2368 // more complex block. Do it as a loop, not inlines. 2369 // Note that things "repeated" a max of once are handled as inline, because 2370 // the one copy of the code already generated is just fine. 2371 return FALSE; 2372 } 2373 2374 // Pick up the opcode that is to be repeated 2375 // 2376 int32_t op = fRXPat->fCompiledPat->elementAti(topOfBlock); 2377 2378 // Compute the pattern location where the inline sequence 2379 // will end, and set up the state save op that will be needed. 2380 // 2381 int32_t endOfSequenceLoc = fRXPat->fCompiledPat->size()-1 2382 + fIntervalUpper + (fIntervalUpper-fIntervalLow); 2383 int32_t saveOp = URX_BUILD(URX_STATE_SAVE, endOfSequenceLoc); 2384 if (fIntervalLow == 0) { 2385 insertOp(topOfBlock); 2386 fRXPat->fCompiledPat->setElementAt(saveOp, topOfBlock); 2387 } 2388 2389 2390 2391 // Loop, emitting the op for the thing being repeated each time. 2392 // Loop starts at 1 because one instance of the op already exists in the pattern, 2393 // it was put there when it was originally encountered. 2394 int32_t i; 2395 for (i=1; i<fIntervalUpper; i++ ) { 2396 if (i == fIntervalLow) { 2397 fRXPat->fCompiledPat->addElement(saveOp, *fStatus); 2398 } 2399 if (i > fIntervalLow) { 2400 fRXPat->fCompiledPat->addElement(saveOp, *fStatus); 2401 } 2402 fRXPat->fCompiledPat->addElement(op, *fStatus); 2403 } 2404 return TRUE; 2405 } 2406 2407 2408 2409 //------------------------------------------------------------------------------ 2410 // 2411 // matchStartType Determine how a match can start. 2412 // Used to optimize find() operations. 2413 // 2414 // Operation is very similar to minMatchLength(). Walk the compiled 2415 // pattern, keeping an on-going minimum-match-length. For any 2416 // op where the min match coming in is zero, add that ops possible 2417 // starting matches to the possible starts for the overall pattern. 2418 // 2419 //------------------------------------------------------------------------------ 2420 void RegexCompile::matchStartType() { 2421 if (U_FAILURE(*fStatus)) { 2422 return; 2423 } 2424 2425 2426 int32_t loc; // Location in the pattern of the current op being processed. 2427 int32_t op; // The op being processed 2428 int32_t opType; // The opcode type of the op 2429 int32_t currentLen = 0; // Minimum length of a match to this point (loc) in the pattern 2430 int32_t numInitialStrings = 0; // Number of strings encountered that could match at start. 2431 2432 UBool atStart = TRUE; // True if no part of the pattern yet encountered 2433 // could have advanced the position in a match. 2434 // (Maximum match length so far == 0) 2435 2436 // forwardedLength is a vector holding minimum-match-length values that 2437 // are propagated forward in the pattern by JMP or STATE_SAVE operations. 2438 // It must be one longer than the pattern being checked because some ops 2439 // will jmp to a end-of-block+1 location from within a block, and we must 2440 // count those when checking the block. 2441 int32_t end = fRXPat->fCompiledPat->size(); 2442 UVector32 forwardedLength(end+1, *fStatus); 2443 forwardedLength.setSize(end+1); 2444 for (loc=3; loc<end; loc++) { 2445 forwardedLength.setElementAt(INT32_MAX, loc); 2446 } 2447 2448 for (loc = 3; loc<end; loc++) { 2449 op = fRXPat->fCompiledPat->elementAti(loc); 2450 opType = URX_TYPE(op); 2451 2452 // The loop is advancing linearly through the pattern. 2453 // If the op we are now at was the destination of a branch in the pattern, 2454 // and that path has a shorter minimum length than the current accumulated value, 2455 // replace the current accumulated value. 2456 if (forwardedLength.elementAti(loc) < currentLen) { 2457 currentLen = forwardedLength.elementAti(loc); 2458 U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); 2459 } 2460 2461 switch (opType) { 2462 // Ops that don't change the total length matched 2463 case URX_RESERVED_OP: 2464 case URX_END: 2465 case URX_FAIL: 2466 case URX_STRING_LEN: 2467 case URX_NOP: 2468 case URX_START_CAPTURE: 2469 case URX_END_CAPTURE: 2470 case URX_BACKSLASH_B: 2471 case URX_BACKSLASH_BU: 2472 case URX_BACKSLASH_G: 2473 case URX_BACKSLASH_Z: 2474 case URX_DOLLAR: 2475 case URX_DOLLAR_M: 2476 case URX_DOLLAR_D: 2477 case URX_DOLLAR_MD: 2478 case URX_RELOC_OPRND: 2479 case URX_STO_INP_LOC: 2480 case URX_BACKREF: // BackRef. Must assume that it might be a zero length match 2481 case URX_BACKREF_I: 2482 2483 case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. 2484 case URX_LD_SP: 2485 break; 2486 2487 case URX_CARET: 2488 if (atStart) { 2489 fRXPat->fStartType = START_START; 2490 } 2491 break; 2492 2493 case URX_CARET_M: 2494 case URX_CARET_M_UNIX: 2495 if (atStart) { 2496 fRXPat->fStartType = START_LINE; 2497 } 2498 break; 2499 2500 case URX_ONECHAR: 2501 if (currentLen == 0) { 2502 // This character could appear at the start of a match. 2503 // Add it to the set of possible starting characters. 2504 fRXPat->fInitialChars->add(URX_VAL(op)); 2505 numInitialStrings += 2; 2506 } 2507 currentLen++; 2508 atStart = FALSE; 2509 break; 2510 2511 2512 case URX_SETREF: 2513 if (currentLen == 0) { 2514 int32_t sn = URX_VAL(op); 2515 U_ASSERT(sn > 0 && sn < fRXPat->fSets->size()); 2516 const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn); 2517 fRXPat->fInitialChars->addAll(*s); 2518 numInitialStrings += 2; 2519 } 2520 currentLen++; 2521 atStart = FALSE; 2522 break; 2523 2524 case URX_LOOP_SR_I: 2525 // [Set]*, like a SETREF, above, in what it can match, 2526 // but may not match at all, so currentLen is not incremented. 2527 if (currentLen == 0) { 2528 int32_t sn = URX_VAL(op); 2529 U_ASSERT(sn > 0 && sn < fRXPat->fSets->size()); 2530 const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn); 2531 fRXPat->fInitialChars->addAll(*s); 2532 numInitialStrings += 2; 2533 } 2534 atStart = FALSE; 2535 break; 2536 2537 case URX_LOOP_DOT_I: 2538 if (currentLen == 0) { 2539 // .* at the start of a pattern. 2540 // Any character can begin the match. 2541 fRXPat->fInitialChars->clear(); 2542 fRXPat->fInitialChars->complement(); 2543 numInitialStrings += 2; 2544 } 2545 atStart = FALSE; 2546 break; 2547 2548 2549 case URX_STATIC_SETREF: 2550 if (currentLen == 0) { 2551 int32_t sn = URX_VAL(op); 2552 U_ASSERT(sn>0 && sn<URX_LAST_SET); 2553 const UnicodeSet *s = fRXPat->fStaticSets[sn]; 2554 fRXPat->fInitialChars->addAll(*s); 2555 numInitialStrings += 2; 2556 } 2557 currentLen++; 2558 atStart = FALSE; 2559 break; 2560 2561 2562 2563 case URX_STAT_SETREF_N: 2564 if (currentLen == 0) { 2565 int32_t sn = URX_VAL(op); 2566 const UnicodeSet *s = fRXPat->fStaticSets[sn]; 2567 UnicodeSet sc(*s); 2568 sc.complement(); 2569 fRXPat->fInitialChars->addAll(sc); 2570 numInitialStrings += 2; 2571 } 2572 currentLen++; 2573 atStart = FALSE; 2574 break; 2575 2576 2577 2578 case URX_BACKSLASH_D: 2579 // Digit Char 2580 if (currentLen == 0) { 2581 UnicodeSet s; 2582 s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus); 2583 if (URX_VAL(op) != 0) { 2584 s.complement(); 2585 } 2586 fRXPat->fInitialChars->addAll(s); 2587 numInitialStrings += 2; 2588 } 2589 currentLen++; 2590 atStart = FALSE; 2591 break; 2592 2593 2594 case URX_ONECHAR_I: 2595 // Case Insensitive Single Character. 2596 if (currentLen == 0) { 2597 UChar32 c = URX_VAL(op); 2598 if (u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) { 2599 // character may have distinct cased forms. Add all of them 2600 // to the set of possible starting match chars. 2601 UnicodeSet s(c, c); 2602 s.closeOver(USET_CASE_INSENSITIVE); 2603 fRXPat->fInitialChars->addAll(s); 2604 } else { 2605 // Char has no case variants. Just add it as-is to the 2606 // set of possible starting chars. 2607 fRXPat->fInitialChars->add(c); 2608 } 2609 numInitialStrings += 2; 2610 } 2611 currentLen++; 2612 atStart = FALSE; 2613 break; 2614 2615 2616 case URX_BACKSLASH_X: // Grahpeme Cluster. Minimum is 1, max unbounded. 2617 case URX_DOTANY_ALL: // . matches one or two. 2618 case URX_DOTANY: 2619 case URX_DOTANY_UNIX: 2620 if (currentLen == 0) { 2621 // These constructs are all bad news when they appear at the start 2622 // of a match. Any character can begin the match. 2623 fRXPat->fInitialChars->clear(); 2624 fRXPat->fInitialChars->complement(); 2625 numInitialStrings += 2; 2626 } 2627 currentLen++; 2628 atStart = FALSE; 2629 break; 2630 2631 2632 case URX_JMPX: 2633 loc++; // Except for extra operand on URX_JMPX, same as URX_JMP. 2634 case URX_JMP: 2635 { 2636 int32_t jmpDest = URX_VAL(op); 2637 if (jmpDest < loc) { 2638 // Loop of some kind. Can safely ignore, the worst that will happen 2639 // is that we understate the true minimum length 2640 currentLen = forwardedLength.elementAti(loc+1); 2641 2642 } else { 2643 // Forward jump. Propagate the current min length to the target loc of the jump. 2644 U_ASSERT(jmpDest <= end+1); 2645 if (forwardedLength.elementAti(jmpDest) > currentLen) { 2646 forwardedLength.setElementAt(currentLen, jmpDest); 2647 } 2648 } 2649 } 2650 atStart = FALSE; 2651 break; 2652 2653 case URX_JMP_SAV: 2654 case URX_JMP_SAV_X: 2655 // Combo of state save to the next loc, + jmp backwards. 2656 // Net effect on min. length computation is nothing. 2657 atStart = FALSE; 2658 break; 2659 2660 case URX_BACKTRACK: 2661 // Fails are kind of like a branch, except that the min length was 2662 // propagated already, by the state save. 2663 currentLen = forwardedLength.elementAti(loc+1); 2664 atStart = FALSE; 2665 break; 2666 2667 2668 case URX_STATE_SAVE: 2669 { 2670 // State Save, for forward jumps, propagate the current minimum. 2671 // of the state save. 2672 int32_t jmpDest = URX_VAL(op); 2673 if (jmpDest > loc) { 2674 if (currentLen < forwardedLength.elementAti(jmpDest)) { 2675 forwardedLength.setElementAt(currentLen, jmpDest); 2676 } 2677 } 2678 } 2679 atStart = FALSE; 2680 break; 2681 2682 2683 2684 2685 case URX_STRING: 2686 { 2687 loc++; 2688 int32_t stringLenOp = fRXPat->fCompiledPat->elementAti(loc); 2689 int32_t stringLen = URX_VAL(stringLenOp); 2690 U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN); 2691 U_ASSERT(stringLenOp >= 2); 2692 if (currentLen == 0) { 2693 // Add the starting character of this string to the set of possible starting 2694 // characters for this pattern. 2695 int32_t stringStartIdx = URX_VAL(op); 2696 UChar32 c = fRXPat->fLiteralText.char32At(stringStartIdx); 2697 fRXPat->fInitialChars->add(c); 2698 2699 // Remember this string. After the entire pattern has been checked, 2700 // if nothing else is identified that can start a match, we'll use it. 2701 numInitialStrings++; 2702 fRXPat->fInitialStringIdx = stringStartIdx; 2703 fRXPat->fInitialStringLen = stringLen; 2704 } 2705 2706 currentLen += stringLen; 2707 atStart = FALSE; 2708 } 2709 break; 2710 2711 case URX_STRING_I: 2712 { 2713 // Case-insensitive string. Unlike exact-match strings, we won't 2714 // attempt a string search for possible match positions. But we 2715 // do update the set of possible starting characters. 2716 loc++; 2717 int32_t stringLenOp = fRXPat->fCompiledPat->elementAti(loc); 2718 int32_t stringLen = URX_VAL(stringLenOp); 2719 U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN); 2720 U_ASSERT(stringLenOp >= 2); 2721 if (currentLen == 0) { 2722 // Add the starting character of this string to the set of possible starting 2723 // characters for this pattern. 2724 int32_t stringStartIdx = URX_VAL(op); 2725 UChar32 c = fRXPat->fLiteralText.char32At(stringStartIdx); 2726 UnicodeSet s(c, c); 2727 s.closeOver(USET_CASE_INSENSITIVE); 2728 fRXPat->fInitialChars->addAll(s); 2729 numInitialStrings += 2; // Matching on an initial string not possible. 2730 } 2731 currentLen += stringLen; 2732 atStart = FALSE; 2733 } 2734 break; 2735 2736 case URX_CTR_INIT: 2737 case URX_CTR_INIT_NG: 2738 { 2739 // Loop Init Ops. These don't change the min length, but they are 4 word ops 2740 // so location must be updated accordingly. 2741 // Loop Init Ops. 2742 // If the min loop count == 0 2743 // move loc forwards to the end of the loop, skipping over the body. 2744 // If the min count is > 0, 2745 // continue normal processing of the body of the loop. 2746 int32_t loopEndLoc = fRXPat->fCompiledPat->elementAti(loc+1); 2747 loopEndLoc = URX_VAL(loopEndLoc); 2748 int32_t minLoopCount = fRXPat->fCompiledPat->elementAti(loc+2); 2749 if (minLoopCount == 0) { 2750 // Min Loop Count of 0, treat like a forward branch and 2751 // move the current minimum length up to the target 2752 // (end of loop) location. 2753 U_ASSERT(loopEndLoc <= end+1); 2754 if (forwardedLength.elementAti(loopEndLoc) > currentLen) { 2755 forwardedLength.setElementAt(currentLen, loopEndLoc); 2756 } 2757 } 2758 loc+=3; // Skips over operands of CTR_INIT 2759 } 2760 atStart = FALSE; 2761 break; 2762 2763 2764 case URX_CTR_LOOP: 2765 case URX_CTR_LOOP_NG: 2766 // Loop ops. 2767 // The jump is conditional, backwards only. 2768 atStart = FALSE; 2769 break; 2770 2771 case URX_LOOP_C: 2772 // More loop ops. These state-save to themselves. 2773 // don't change the minimum match 2774 atStart = FALSE; 2775 break; 2776 2777 2778 case URX_LA_START: 2779 case URX_LB_START: 2780 { 2781 // Look-around. Scan forward until the matching look-ahead end, 2782 // without processing the look-around block. This is overly pessimistic. 2783 2784 // Keep track of the nesting depth of look-around blocks. Boilerplate code for 2785 // lookahead contains two LA_END instructions, so count goes up by two 2786 // for each LA_START. 2787 int32_t depth = (opType == URX_LA_START? 2: 1); 2788 for (;;) { 2789 loc++; 2790 op = fRXPat->fCompiledPat->elementAti(loc); 2791 if (URX_TYPE(op) == URX_LA_START) { 2792 depth+=2; 2793 } 2794 if (URX_TYPE(op) == URX_LB_START) { 2795 depth++; 2796 } 2797 if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) { 2798 depth--; 2799 if (depth == 0) { 2800 break; 2801 } 2802 } 2803 if (URX_TYPE(op) == URX_STATE_SAVE) { 2804 // Need this because neg lookahead blocks will FAIL to outside 2805 // of the block. 2806 int32_t jmpDest = URX_VAL(op); 2807 if (jmpDest > loc) { 2808 if (currentLen < forwardedLength.elementAti(jmpDest)) { 2809 forwardedLength.setElementAt(currentLen, jmpDest); 2810 } 2811 } 2812 } 2813 U_ASSERT(loc <= end); 2814 } 2815 } 2816 break; 2817 2818 case URX_LA_END: 2819 case URX_LB_CONT: 2820 case URX_LB_END: 2821 case URX_LBN_CONT: 2822 case URX_LBN_END: 2823 U_ASSERT(FALSE); // Shouldn't get here. These ops should be 2824 // consumed by the scan in URX_LA_START and LB_START 2825 2826 break; 2827 2828 default: 2829 U_ASSERT(FALSE); 2830 } 2831 2832 } 2833 2834 2835 // We have finished walking through the ops. Check whether some forward jump 2836 // propagated a shorter length to location end+1. 2837 if (forwardedLength.elementAti(end+1) < currentLen) { 2838 currentLen = forwardedLength.elementAti(end+1); 2839 } 2840 2841 2842 fRXPat->fInitialChars8->init(fRXPat->fInitialChars); 2843 2844 2845 // Sort out what we should check for when looking for candidate match start positions. 2846 // In order of preference, 2847 // 1. Start of input text buffer. 2848 // 2. A literal string. 2849 // 3. Start of line in multi-line mode. 2850 // 4. A single literal character. 2851 // 5. A character from a set of characters. 2852 // 2853 if (fRXPat->fStartType == START_START) { 2854 // Match only at the start of an input text string. 2855 // start type is already set. We're done. 2856 } else if (numInitialStrings == 1 && fRXPat->fMinMatchLen > 0) { 2857 // Match beginning only with a literal string. 2858 UChar32 c = fRXPat->fLiteralText.char32At(fRXPat->fInitialStringIdx); 2859 U_ASSERT(fRXPat->fInitialChars->contains(c)); 2860 fRXPat->fStartType = START_STRING; 2861 fRXPat->fInitialChar = c; 2862 } else if (fRXPat->fStartType == START_LINE) { 2863 // Match at start of line in Multi-Line mode. 2864 // Nothing to do here; everything is already set. 2865 } else if (fRXPat->fMinMatchLen == 0) { 2866 // Zero length match possible. We could start anywhere. 2867 fRXPat->fStartType = START_NO_INFO; 2868 } else if (fRXPat->fInitialChars->size() == 1) { 2869 // All matches begin with the same char. 2870 fRXPat->fStartType = START_CHAR; 2871 fRXPat->fInitialChar = fRXPat->fInitialChars->charAt(0); 2872 U_ASSERT(fRXPat->fInitialChar != (UChar32)-1); 2873 } else if (fRXPat->fInitialChars->contains((UChar32)0, (UChar32)0x10ffff) == FALSE && 2874 fRXPat->fMinMatchLen > 0) { 2875 // Matches start with a set of character smaller than the set of all chars. 2876 fRXPat->fStartType = START_SET; 2877 } else { 2878 // Matches can start with anything 2879 fRXPat->fStartType = START_NO_INFO; 2880 } 2881 2882 return; 2883 } 2884 2885 2886 2887 //------------------------------------------------------------------------------ 2888 // 2889 // minMatchLength Calculate the length of the shortest string that could 2890 // match the specified pattern. 2891 // Length is in 16 bit code units, not code points. 2892 // 2893 // The calculated length may not be exact. The returned 2894 // value may be shorter than the actual minimum; it must 2895 // never be longer. 2896 // 2897 // start and end are the range of p-code operations to be 2898 // examined. The endpoints are included in the range. 2899 // 2900 //------------------------------------------------------------------------------ 2901 int32_t RegexCompile::minMatchLength(int32_t start, int32_t end) { 2902 if (U_FAILURE(*fStatus)) { 2903 return 0; 2904 } 2905 2906 U_ASSERT(start <= end); 2907 U_ASSERT(end < fRXPat->fCompiledPat->size()); 2908 2909 2910 int32_t loc; 2911 int32_t op; 2912 int32_t opType; 2913 int32_t currentLen = 0; 2914 2915 2916 // forwardedLength is a vector holding minimum-match-length values that 2917 // are propagated forward in the pattern by JMP or STATE_SAVE operations. 2918 // It must be one longer than the pattern being checked because some ops 2919 // will jmp to a end-of-block+1 location from within a block, and we must 2920 // count those when checking the block. 2921 UVector32 forwardedLength(end+2, *fStatus); 2922 forwardedLength.setSize(end+2); 2923 for (loc=start; loc<=end+1; loc++) { 2924 forwardedLength.setElementAt(INT32_MAX, loc); 2925 } 2926 2927 for (loc = start; loc<=end; loc++) { 2928 op = fRXPat->fCompiledPat->elementAti(loc); 2929 opType = URX_TYPE(op); 2930 2931 // The loop is advancing linearly through the pattern. 2932 // If the op we are now at was the destination of a branch in the pattern, 2933 // and that path has a shorter minimum length than the current accumulated value, 2934 // replace the current accumulated value. 2935 // U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); // MinLength == INT32_MAX for some 2936 // no-match-possible cases. 2937 if (forwardedLength.elementAti(loc) < currentLen) { 2938 currentLen = forwardedLength.elementAti(loc); 2939 U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); 2940 } 2941 2942 switch (opType) { 2943 // Ops that don't change the total length matched 2944 case URX_RESERVED_OP: 2945 case URX_END: 2946 case URX_STRING_LEN: 2947 case URX_NOP: 2948 case URX_START_CAPTURE: 2949 case URX_END_CAPTURE: 2950 case URX_BACKSLASH_B: 2951 case URX_BACKSLASH_BU: 2952 case URX_BACKSLASH_G: 2953 case URX_BACKSLASH_Z: 2954 case URX_CARET: 2955 case URX_DOLLAR: 2956 case URX_DOLLAR_M: 2957 case URX_DOLLAR_D: 2958 case URX_DOLLAR_MD: 2959 case URX_RELOC_OPRND: 2960 case URX_STO_INP_LOC: 2961 case URX_CARET_M: 2962 case URX_CARET_M_UNIX: 2963 case URX_BACKREF: // BackRef. Must assume that it might be a zero length match 2964 case URX_BACKREF_I: 2965 2966 case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. 2967 case URX_LD_SP: 2968 2969 case URX_JMP_SAV: 2970 case URX_JMP_SAV_X: 2971 break; 2972 2973 2974 // Ops that match a minimum of one character (one or two 16 bit code units.) 2975 // 2976 case URX_ONECHAR: 2977 case URX_STATIC_SETREF: 2978 case URX_STAT_SETREF_N: 2979 case URX_SETREF: 2980 case URX_BACKSLASH_D: 2981 case URX_ONECHAR_I: 2982 case URX_BACKSLASH_X: // Grahpeme Cluster. Minimum is 1, max unbounded. 2983 case URX_DOTANY_ALL: // . matches one or two. 2984 case URX_DOTANY: 2985 case URX_DOTANY_UNIX: 2986 currentLen++; 2987 break; 2988 2989 2990 case URX_JMPX: 2991 loc++; // URX_JMPX has an extra operand, ignored here, 2992 // otherwise processed identically to URX_JMP. 2993 case URX_JMP: 2994 { 2995 int32_t jmpDest = URX_VAL(op); 2996 if (jmpDest < loc) { 2997 // Loop of some kind. Can safely ignore, the worst that will happen 2998 // is that we understate the true minimum length 2999 currentLen = forwardedLength.elementAti(loc+1); 3000 } else { 3001 // Forward jump. Propagate the current min length to the target loc of the jump. 3002 U_ASSERT(jmpDest <= end+1); 3003 if (forwardedLength.elementAti(jmpDest) > currentLen) { 3004 forwardedLength.setElementAt(currentLen, jmpDest); 3005 } 3006 } 3007 } 3008 break; 3009 3010 case URX_BACKTRACK: 3011 { 3012 // Back-tracks are kind of like a branch, except that the min length was 3013 // propagated already, by the state save. 3014 currentLen = forwardedLength.elementAti(loc+1); 3015 } 3016 break; 3017 3018 3019 case URX_STATE_SAVE: 3020 { 3021 // State Save, for forward jumps, propagate the current minimum. 3022 // of the state save. 3023 int32_t jmpDest = URX_VAL(op); 3024 if (jmpDest > loc) { 3025 if (currentLen < forwardedLength.elementAti(jmpDest)) { 3026 forwardedLength.setElementAt(currentLen, jmpDest); 3027 } 3028 } 3029 } 3030 break; 3031 3032 3033 case URX_STRING: 3034 case URX_STRING_I: 3035 { 3036 loc++; 3037 int32_t stringLenOp = fRXPat->fCompiledPat->elementAti(loc); 3038 currentLen += URX_VAL(stringLenOp); 3039 } 3040 break; 3041 3042 3043 case URX_CTR_INIT: 3044 case URX_CTR_INIT_NG: 3045 { 3046 // Loop Init Ops. 3047 // If the min loop count == 0 3048 // move loc forwards to the end of the loop, skipping over the body. 3049 // If the min count is > 0, 3050 // continue normal processing of the body of the loop. 3051 int32_t loopEndLoc = fRXPat->fCompiledPat->elementAti(loc+1); 3052 loopEndLoc = URX_VAL(loopEndLoc); 3053 int32_t minLoopCount = fRXPat->fCompiledPat->elementAti(loc+2); 3054 if (minLoopCount == 0) { 3055 loc = loopEndLoc; 3056 } else { 3057 loc+=3; // Skips over operands of CTR_INIT 3058 } 3059 } 3060 break; 3061 3062 3063 case URX_CTR_LOOP: 3064 case URX_CTR_LOOP_NG: 3065 // Loop ops. 3066 // The jump is conditional, backwards only. 3067 break; 3068 3069 case URX_LOOP_SR_I: 3070 case URX_LOOP_DOT_I: 3071 case URX_LOOP_C: 3072 // More loop ops. These state-save to themselves. 3073 // don't change the minimum match - could match nothing at all. 3074 break; 3075 3076 3077 case URX_LA_START: 3078 case URX_LB_START: 3079 { 3080 // Look-around. Scan forward until the matching look-ahead end, 3081 // without processing the look-around block. This is overly pessimistic for look-ahead, 3082 // it assumes that the look-ahead match might be zero-length. 3083 // TODO: Positive lookahead could recursively do the block, then continue 3084 // with the longer of the block or the value coming in. Ticket 6060 3085 int32_t depth = (opType == URX_LA_START? 2: 1);; 3086 for (;;) { 3087 loc++; 3088 op = fRXPat->fCompiledPat->elementAti(loc); 3089 if (URX_TYPE(op) == URX_LA_START) { 3090 // The boilerplate for look-ahead includes two LA_END insturctions, 3091 // Depth will be decremented by each one when it is seen. 3092 depth += 2; 3093 } 3094 if (URX_TYPE(op) == URX_LB_START) { 3095 depth++; 3096 } 3097 if (URX_TYPE(op) == URX_LA_END) { 3098 depth--; 3099 if (depth == 0) { 3100 break; 3101 } 3102 } 3103 if (URX_TYPE(op)==URX_LBN_END) { 3104 depth--; 3105 if (depth == 0) { 3106 break; 3107 } 3108 } 3109 if (URX_TYPE(op) == URX_STATE_SAVE) { 3110 // Need this because neg lookahead blocks will FAIL to outside 3111 // of the block. 3112 int32_t jmpDest = URX_VAL(op); 3113 if (jmpDest > loc) { 3114 if (currentLen < forwardedLength.elementAti(jmpDest)) { 3115 forwardedLength.setElementAt(currentLen, jmpDest); 3116 } 3117 } 3118 } 3119 U_ASSERT(loc <= end); 3120 } 3121 } 3122 break; 3123 3124 case URX_LA_END: 3125 case URX_LB_CONT: 3126 case URX_LB_END: 3127 case URX_LBN_CONT: 3128 case URX_LBN_END: 3129 // Only come here if the matching URX_LA_START or URX_LB_START was not in the 3130 // range being sized, which happens when measuring size of look-behind blocks. 3131 break; 3132 3133 default: 3134 U_ASSERT(FALSE); 3135 } 3136 3137 } 3138 3139 // We have finished walking through the ops. Check whether some forward jump 3140 // propagated a shorter length to location end+1. 3141 if (forwardedLength.elementAti(end+1) < currentLen) { 3142 currentLen = forwardedLength.elementAti(end+1); 3143 U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); 3144 } 3145 3146 return currentLen; 3147 } 3148 3149 3150 3151 //------------------------------------------------------------------------------ 3152 // 3153 // maxMatchLength Calculate the length of the longest string that could 3154 // match the specified pattern. 3155 // Length is in 16 bit code units, not code points. 3156 // 3157 // The calculated length may not be exact. The returned 3158 // value may be longer than the actual maximum; it must 3159 // never be shorter. 3160 // 3161 //------------------------------------------------------------------------------ 3162 int32_t RegexCompile::maxMatchLength(int32_t start, int32_t end) { 3163 if (U_FAILURE(*fStatus)) { 3164 return 0; 3165 } 3166 U_ASSERT(start <= end); 3167 U_ASSERT(end < fRXPat->fCompiledPat->size()); 3168 3169 3170 int32_t loc; 3171 int32_t op; 3172 int32_t opType; 3173 int32_t currentLen = 0; 3174 UVector32 forwardedLength(end+1, *fStatus); 3175 forwardedLength.setSize(end+1); 3176 3177 for (loc=start; loc<=end; loc++) { 3178 forwardedLength.setElementAt(0, loc); 3179 } 3180 3181 for (loc = start; loc<=end; loc++) { 3182 op = fRXPat->fCompiledPat->elementAti(loc); 3183 opType = URX_TYPE(op); 3184 3185 // The loop is advancing linearly through the pattern. 3186 // If the op we are now at was the destination of a branch in the pattern, 3187 // and that path has a longer maximum length than the current accumulated value, 3188 // replace the current accumulated value. 3189 if (forwardedLength.elementAti(loc) > currentLen) { 3190 currentLen = forwardedLength.elementAti(loc); 3191 } 3192 3193 switch (opType) { 3194 // Ops that don't change the total length matched 3195 case URX_RESERVED_OP: 3196 case URX_END: 3197 case URX_STRING_LEN: 3198 case URX_NOP: 3199 case URX_START_CAPTURE: 3200 case URX_END_CAPTURE: 3201 case URX_BACKSLASH_B: 3202 case URX_BACKSLASH_BU: 3203 case URX_BACKSLASH_G: 3204 case URX_BACKSLASH_Z: 3205 case URX_CARET: 3206 case URX_DOLLAR: 3207 case URX_DOLLAR_M: 3208 case URX_DOLLAR_D: 3209 case URX_DOLLAR_MD: 3210 case URX_RELOC_OPRND: 3211 case URX_STO_INP_LOC: 3212 case URX_CARET_M: 3213 case URX_CARET_M_UNIX: 3214 3215 case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. 3216 case URX_LD_SP: 3217 3218 case URX_LB_END: 3219 case URX_LB_CONT: 3220 case URX_LBN_CONT: 3221 case URX_LBN_END: 3222 break; 3223 3224 3225 // Ops that increase that cause an unbounded increase in the length 3226 // of a matched string, or that increase it a hard to characterize way. 3227 // Call the max length unbounded, and stop further checking. 3228 case URX_BACKREF: // BackRef. Must assume that it might be a zero length match 3229 case URX_BACKREF_I: 3230 case URX_BACKSLASH_X: // Grahpeme Cluster. Minimum is 1, max unbounded. 3231 currentLen = INT32_MAX; 3232 break; 3233 3234 3235 // Ops that match a max of one character (possibly two 16 bit code units.) 3236 // 3237 case URX_STATIC_SETREF: 3238 case URX_STAT_SETREF_N: 3239 case URX_SETREF: 3240 case URX_BACKSLASH_D: 3241 case URX_ONECHAR_I: 3242 case URX_DOTANY_ALL: 3243 case URX_DOTANY: 3244 case URX_DOTANY_UNIX: 3245 currentLen+=2; 3246 break; 3247 3248 // Single literal character. Increase current max length by one or two, 3249 // depending on whether the char is in the supplementary range. 3250 case URX_ONECHAR: 3251 currentLen++; 3252 if (URX_VAL(op) > 0x10000) { 3253 currentLen++; 3254 } 3255 break; 3256 3257 // Jumps. 3258 // 3259 case URX_JMP: 3260 case URX_JMPX: 3261 case URX_JMP_SAV: 3262 case URX_JMP_SAV_X: 3263 { 3264 int32_t jmpDest = URX_VAL(op); 3265 if (jmpDest < loc) { 3266 // Loop of some kind. Max match length is unbounded. 3267 currentLen = INT32_MAX; 3268 } else { 3269 // Forward jump. Propagate the current min length to the target loc of the jump. 3270 if (forwardedLength.elementAti(jmpDest) < currentLen) { 3271 forwardedLength.setElementAt(currentLen, jmpDest); 3272 } 3273 currentLen = 0; 3274 } 3275 } 3276 break; 3277 3278 case URX_BACKTRACK: 3279 // back-tracks are kind of like a branch, except that the max length was 3280 // propagated already, by the state save. 3281 currentLen = forwardedLength.elementAti(loc+1); 3282 break; 3283 3284 3285 case URX_STATE_SAVE: 3286 { 3287 // State Save, for forward jumps, propagate the current minimum. 3288 // of the state save. 3289 // For backwards jumps, they create a loop, maximum 3290 // match length is unbounded. 3291 int32_t jmpDest = URX_VAL(op); 3292 if (jmpDest > loc) { 3293 if (currentLen > forwardedLength.elementAti(jmpDest)) { 3294 forwardedLength.setElementAt(currentLen, jmpDest); 3295 } 3296 } else { 3297 currentLen = INT32_MAX; 3298 } 3299 } 3300 break; 3301 3302 3303 3304 3305 case URX_STRING: 3306 case URX_STRING_I: 3307 { 3308 loc++; 3309 int32_t stringLenOp = fRXPat->fCompiledPat->elementAti(loc); 3310 currentLen += URX_VAL(stringLenOp); 3311 } 3312 break; 3313 3314 3315 case URX_CTR_INIT: 3316 case URX_CTR_INIT_NG: 3317 case URX_CTR_LOOP: 3318 case URX_CTR_LOOP_NG: 3319 case URX_LOOP_SR_I: 3320 case URX_LOOP_DOT_I: 3321 case URX_LOOP_C: 3322 // For anything to do with loops, make the match length unbounded. 3323 // Note: INIT instructions are multi-word. Can ignore because 3324 // INT32_MAX length will stop the per-instruction loop. 3325 currentLen = INT32_MAX; 3326 break; 3327 3328 3329 3330 case URX_LA_START: 3331 case URX_LA_END: 3332 // Look-ahead. Just ignore, treat the look-ahead block as if 3333 // it were normal pattern. Gives a too-long match length, 3334 // but good enough for now. 3335 break; 3336 3337 // End of look-ahead ops should always be consumed by the processing at 3338 // the URX_LA_START op. 3339 // U_ASSERT(FALSE); 3340 // break; 3341 3342 case URX_LB_START: 3343 { 3344 // Look-behind. Scan forward until the matching look-around end, 3345 // without processing the look-behind block. 3346 int32_t depth = 0; 3347 for (;;) { 3348 loc++; 3349 op = fRXPat->fCompiledPat->elementAti(loc); 3350 if (URX_TYPE(op) == URX_LA_START || URX_TYPE(op) == URX_LB_START) { 3351 depth++; 3352 } 3353 if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) { 3354 if (depth == 0) { 3355 break; 3356 } 3357 depth--; 3358 } 3359 U_ASSERT(loc < end); 3360 } 3361 } 3362 break; 3363 3364 default: 3365 U_ASSERT(FALSE); 3366 } 3367 3368 3369 if (currentLen == INT32_MAX) { 3370 // The maximum length is unbounded. 3371 // Stop further processing of the pattern. 3372 break; 3373 } 3374 3375 } 3376 return currentLen; 3377 3378 } 3379 3380 3381 //------------------------------------------------------------------------------ 3382 // 3383 // stripNOPs Remove any NOP operations from the compiled pattern code. 3384 // Extra NOPs are inserted for some constructs during the initial 3385 // code generation to provide locations that may be patched later. 3386 // Many end up unneeded, and are removed by this function. 3387 // 3388 //------------------------------------------------------------------------------ 3389 void RegexCompile::stripNOPs() { 3390 3391 if (U_FAILURE(*fStatus)) { 3392 return; 3393 } 3394 3395 int32_t end = fRXPat->fCompiledPat->size(); 3396 UVector32 deltas(end, *fStatus); 3397 3398 // Make a first pass over the code, computing the amount that things 3399 // will be offset at each location in the original code. 3400 int32_t loc; 3401 int32_t d = 0; 3402 for (loc=0; loc<end; loc++) { 3403 deltas.addElement(d, *fStatus); 3404 int32_t op = fRXPat->fCompiledPat->elementAti(loc); 3405 if (URX_TYPE(op) == URX_NOP) { 3406 d++; 3407 } 3408 } 3409 3410 // Make a second pass over the code, removing the NOPs by moving following 3411 // code up, and patching operands that refer to code locations that 3412 // are being moved. The array of offsets from the first step is used 3413 // to compute the new operand values. 3414 int32_t src; 3415 int32_t dst = 0; 3416 for (src=0; src<end; src++) { 3417 int32_t op = fRXPat->fCompiledPat->elementAti(src); 3418 int32_t opType = URX_TYPE(op); 3419 switch (opType) { 3420 case URX_NOP: 3421 break; 3422 3423 case URX_STATE_SAVE: 3424 case URX_JMP: 3425 case URX_CTR_LOOP: 3426 case URX_CTR_LOOP_NG: 3427 case URX_RELOC_OPRND: 3428 case URX_JMPX: 3429 case URX_JMP_SAV: 3430 case URX_JMP_SAV_X: 3431 // These are instructions with operands that refer to code locations. 3432 { 3433 int32_t operandAddress = URX_VAL(op); 3434 U_ASSERT(operandAddress>=0 && operandAddress<deltas.size()); 3435 int32_t fixedOperandAddress = operandAddress - deltas.elementAti(operandAddress); 3436 op = URX_BUILD(opType, fixedOperandAddress); 3437 fRXPat->fCompiledPat->setElementAt(op, dst); 3438 dst++; 3439 break; 3440 } 3441 3442 case URX_RESERVED_OP: 3443 case URX_RESERVED_OP_N: 3444 case URX_BACKTRACK: 3445 case URX_END: 3446 case URX_ONECHAR: 3447 case URX_STRING: 3448 case URX_STRING_LEN: 3449 case URX_START_CAPTURE: 3450 case URX_END_CAPTURE: 3451 case URX_STATIC_SETREF: 3452 case URX_STAT_SETREF_N: 3453 case URX_SETREF: 3454 case URX_DOTANY: 3455 case URX_FAIL: 3456 case URX_BACKSLASH_B: 3457 case URX_BACKSLASH_BU: 3458 case URX_BACKSLASH_G: 3459 case URX_BACKSLASH_X: 3460 case URX_BACKSLASH_Z: 3461 case URX_DOTANY_ALL: 3462 case URX_BACKSLASH_D: 3463 case URX_CARET: 3464 case URX_DOLLAR: 3465 case URX_CTR_INIT: 3466 case URX_CTR_INIT_NG: 3467 case URX_DOTANY_UNIX: 3468 case URX_STO_SP: 3469 case URX_LD_SP: 3470 case URX_BACKREF: 3471 case URX_STO_INP_LOC: 3472 case URX_LA_START: 3473 case URX_LA_END: 3474 case URX_ONECHAR_I: 3475 case URX_STRING_I: 3476 case URX_BACKREF_I: 3477 case URX_DOLLAR_M: 3478 case URX_CARET_M: 3479 case URX_CARET_M_UNIX: 3480 case URX_LB_START: 3481 case URX_LB_CONT: 3482 case URX_LB_END: 3483 case URX_LBN_CONT: 3484 case URX_LBN_END: 3485 case URX_LOOP_SR_I: 3486 case URX_LOOP_DOT_I: 3487 case URX_LOOP_C: 3488 case URX_DOLLAR_D: 3489 case URX_DOLLAR_MD: 3490 // These instructions are unaltered by the relocation. 3491 fRXPat->fCompiledPat->setElementAt(op, dst); 3492 dst++; 3493 break; 3494 3495 default: 3496 // Some op is unaccounted for. 3497 U_ASSERT(FALSE); 3498 error(U_REGEX_INTERNAL_ERROR); 3499 } 3500 } 3501 3502 fRXPat->fCompiledPat->setSize(dst); 3503 } 3504 3505 3506 3507 3508 //------------------------------------------------------------------------------ 3509 // 3510 // Error Report a rule parse error. 3511 // Only report it if no previous error has been recorded. 3512 // 3513 //------------------------------------------------------------------------------ 3514 void RegexCompile::error(UErrorCode e) { 3515 if (U_SUCCESS(*fStatus)) { 3516 *fStatus = e; 3517 fParseErr->line = fLineNum; 3518 fParseErr->offset = fCharNum; 3519 3520 // Fill in the context. 3521 // Note: extractBetween() pins supplied indicies to the string bounds. 3522 uprv_memset(fParseErr->preContext, 0, sizeof(fParseErr->preContext)); 3523 uprv_memset(fParseErr->postContext, 0, sizeof(fParseErr->postContext)); 3524 fRXPat->fPattern.extractBetween(fScanIndex-U_PARSE_CONTEXT_LEN+1, fScanIndex, 3525 fParseErr->preContext, 0); 3526 fRXPat->fPattern.extractBetween(fScanIndex, fScanIndex+U_PARSE_CONTEXT_LEN-1, 3527 fParseErr->postContext, 0); 3528 } 3529 } 3530 3531 3532 // 3533 // Assorted Unicode character constants. 3534 // Numeric because there is no portable way to enter them as literals. 3535 // (Think EBCDIC). 3536 // 3537 static const UChar chCR = 0x0d; // New lines, for terminating comments. 3538 static const UChar chLF = 0x0a; // Line Feed 3539 static const UChar chPound = 0x23; // '#', introduces a comment. 3540 static const UChar chDigit0 = 0x30; // '0' 3541 static const UChar chDigit7 = 0x37; // '9' 3542 static const UChar chColon = 0x3A; // ':' 3543 static const UChar chE = 0x45; // 'E' 3544 static const UChar chQ = 0x51; // 'Q' 3545 static const UChar chN = 0x4E; // 'N' 3546 static const UChar chP = 0x50; // 'P' 3547 static const UChar chBackSlash = 0x5c; // '\' introduces a char escape 3548 static const UChar chLBracket = 0x5b; // '[' 3549 static const UChar chRBracket = 0x5d; // ']' 3550 static const UChar chUp = 0x5e; // '^' 3551 static const UChar chLowerP = 0x70; 3552 static const UChar chLBrace = 0x7b; // '{' 3553 static const UChar chRBrace = 0x7d; // '}' 3554 static const UChar chNEL = 0x85; // NEL newline variant 3555 static const UChar chLS = 0x2028; // Unicode Line Separator 3556 3557 3558 //------------------------------------------------------------------------------ 3559 // 3560 // nextCharLL Low Level Next Char from the regex pattern. 3561 // Get a char from the string, keep track of input position 3562 // for error reporting. 3563 // 3564 //------------------------------------------------------------------------------ 3565 UChar32 RegexCompile::nextCharLL() { 3566 UChar32 ch; 3567 UnicodeString &pattern = fRXPat->fPattern; 3568 3569 if (fPeekChar != -1) { 3570 ch = fPeekChar; 3571 fPeekChar = -1; 3572 return ch; 3573 } 3574 if (fPatternLength==0 || fNextIndex >= fPatternLength) { 3575 return (UChar32)-1; 3576 } 3577 ch = pattern.char32At(fNextIndex); 3578 fNextIndex = pattern.moveIndex32(fNextIndex, 1); 3579 3580 if (ch == chCR || 3581 ch == chNEL || 3582 ch == chLS || 3583 ch == chLF && fLastChar != chCR) { 3584 // Character is starting a new line. Bump up the line number, and 3585 // reset the column to 0. 3586 fLineNum++; 3587 fCharNum=0; 3588 } 3589 else { 3590 // Character is not starting a new line. Except in the case of a 3591 // LF following a CR, increment the column position. 3592 if (ch != chLF) { 3593 fCharNum++; 3594 } 3595 } 3596 fLastChar = ch; 3597 return ch; 3598 } 3599 3600 //------------------------------------------------------------------------------ 3601 // 3602 // peekCharLL Low Level Character Scanning, sneak a peek at the next 3603 // character without actually getting it. 3604 // 3605 //------------------------------------------------------------------------------ 3606 UChar32 RegexCompile::peekCharLL() { 3607 if (fPeekChar == -1) { 3608 fPeekChar = nextCharLL(); 3609 } 3610 return fPeekChar; 3611 } 3612 3613 3614 //------------------------------------------------------------------------------ 3615 // 3616 // nextChar for pattern scanning. At this level, we handle stripping 3617 // out comments and processing some backslash character escapes. 3618 // The rest of the pattern grammar is handled at the next level up. 3619 // 3620 //------------------------------------------------------------------------------ 3621 void RegexCompile::nextChar(RegexPatternChar &c) { 3622 3623 fScanIndex = fNextIndex; 3624 c.fChar = nextCharLL(); 3625 c.fQuoted = FALSE; 3626 3627 if (fQuoteMode) { 3628 c.fQuoted = TRUE; 3629 if ((c.fChar==chBackSlash && peekCharLL()==chE) || c.fChar == (UChar32)-1) { 3630 fQuoteMode = FALSE; // Exit quote mode, 3631 nextCharLL(); // discard the E 3632 nextChar(c); // recurse to get the real next char 3633 } 3634 } 3635 else if (fInBackslashQuote) { 3636 // The current character immediately follows a '\' 3637 // Don't check for any further escapes, just return it as-is. 3638 // Don't set c.fQuoted, because that would prevent the state machine from 3639 // dispatching on the character. 3640 fInBackslashQuote = FALSE; 3641 } 3642 else 3643 { 3644 // We are not in a \Q quoted region \E of the source. 3645 // 3646 if (fModeFlags & UREGEX_COMMENTS) { 3647 // 3648 // We are in free-spacing and comments mode. 3649 // Scan through any white space and comments, until we 3650 // reach a significant character or the end of inut. 3651 for (;;) { 3652 if (c.fChar == (UChar32)-1) { 3653 break; // End of Input 3654 } 3655 if (c.fChar == chPound && fEOLComments == TRUE) { 3656 // Start of a comment. Consume the rest of it, until EOF or a new line 3657 for (;;) { 3658 c.fChar = nextCharLL(); 3659 if (c.fChar == (UChar32)-1 || // EOF 3660 c.fChar == chCR || 3661 c.fChar == chLF || 3662 c.fChar == chNEL || 3663 c.fChar == chLS) { 3664 break; 3665 } 3666 } 3667 } 3668 // TODO: check what Java & Perl do with non-ASCII white spaces. Ticket 6061. 3669 if (uprv_isRuleWhiteSpace(c.fChar) == FALSE) { 3670 break; 3671 } 3672 c.fChar = nextCharLL(); 3673 } 3674 } 3675 3676 // 3677 // check for backslash escaped characters. 3678 // 3679 if (c.fChar == chBackSlash) { 3680 int32_t startX = fNextIndex; // start and end positions of the 3681 int32_t endX = fNextIndex; // sequence following the '\' 3682 if (RegexStaticSets::gStaticSets->fUnescapeCharSet.contains(peekCharLL())) { 3683 // 3684 // A '\' sequence that is handled by ICU's standard unescapeAt function. 3685 // Includes \uxxxx, \n, \r, many others. 3686 // Return the single equivalent character. 3687 // 3688 nextCharLL(); // get & discard the peeked char. 3689 c.fQuoted = TRUE; 3690 c.fChar = fRXPat->fPattern.unescapeAt(endX); 3691 if (startX == endX) { 3692 error(U_REGEX_BAD_ESCAPE_SEQUENCE); 3693 } 3694 fCharNum += endX - startX; 3695 fNextIndex = endX; 3696 } 3697 else if (peekCharLL() == chDigit0) { 3698 // Octal Escape, using Java Regexp Conventions 3699 // which are \0 followed by 1-3 octal digits. 3700 // Different from ICU Unescape handling of Octal, which does not 3701 // require the leading 0. 3702 // Java also has the convention of only consuning 2 octal digits if 3703 // the three digit number would be > 0xff 3704 // 3705 c.fChar = 0; 3706 nextCharLL(); // Consume the initial 0. 3707 int index; 3708 for (index=0; index<3; index++) { 3709 int32_t ch = peekCharLL(); 3710 if (ch<chDigit0 || ch>chDigit7) { 3711 if (index==0) { 3712 // \0 is not followed by any octal digits. 3713 error(U_REGEX_BAD_ESCAPE_SEQUENCE); 3714 } 3715 break; 3716 } 3717 c.fChar <<= 3; 3718 c.fChar += ch&7; 3719 if (c.fChar <= 255) { 3720 nextCharLL(); 3721 } else { 3722 // The last digit made the number too big. Forget we saw it. 3723 c.fChar >>= 3; 3724 } 3725 } 3726 c.fQuoted = TRUE; 3727 } 3728 else if (peekCharLL() == chQ) { 3729 // "\Q" enter quote mode, which will continue until "\E" 3730 fQuoteMode = TRUE; 3731 nextCharLL(); // discard the 'Q'. 3732 nextChar(c); // recurse to get the real next char. 3733 } 3734 else 3735 { 3736 // We are in a '\' escape that will be handled by the state table scanner. 3737 // Just return the backslash, but remember that the following char is to 3738 // be taken literally. 3739 fInBackslashQuote = TRUE; 3740 } 3741 } 3742 } 3743 3744 // re-enable # to end-of-line comments, in case they were disabled. 3745 // They are disabled by the parser upon seeing '(?', but this lasts for 3746 // the fetching of the next character only. 3747 fEOLComments = TRUE; 3748 3749 // putc(c.fChar, stdout); 3750 } 3751 3752 3753 3754 //------------------------------------------------------------------------------ 3755 // 3756 // scanNamedChar 3757 // Get a UChar32 from a \N{UNICODE CHARACTER NAME} in the pattern. 3758 // 3759 // The scan position will be at the 'N'. On return 3760 // the scan position should be just after the '}' 3761 // 3762 // Return the UChar32 3763 // 3764 //------------------------------------------------------------------------------ 3765 UChar32 RegexCompile::scanNamedChar() { 3766 if (U_FAILURE(*fStatus)) { 3767 return 0; 3768 } 3769 3770 nextChar(fC); 3771 if (fC.fChar != chLBrace) { 3772 error(U_REGEX_PROPERTY_SYNTAX); 3773 return 0; 3774 } 3775 3776 UnicodeString charName; 3777 for (;;) { 3778 nextChar(fC); 3779 if (fC.fChar == chRBrace) { 3780 break; 3781 } 3782 if (fC.fChar == -1) { 3783 error(U_REGEX_PROPERTY_SYNTAX); 3784 return 0; 3785 } 3786 charName.append(fC.fChar); 3787 } 3788 3789 char name[100]; 3790 if (!uprv_isInvariantUString(charName.getBuffer(), charName.length()) || 3791 (uint32_t)charName.length()>=sizeof(name)) { 3792 // All Unicode character names have only invariant characters. 3793 // The API to get a character, given a name, accepts only char *, forcing us to convert, 3794 // which requires this error check 3795 error(U_REGEX_PROPERTY_SYNTAX); 3796 return 0; 3797 } 3798 charName.extract(0, charName.length(), name, sizeof(name), US_INV); 3799 3800 UChar32 theChar = u_charFromName(U_UNICODE_CHAR_NAME, name, fStatus); 3801 if (U_FAILURE(*fStatus)) { 3802 error(U_REGEX_PROPERTY_SYNTAX); 3803 } 3804 3805 nextChar(fC); // Continue overall regex pattern processing with char after the '}' 3806 return theChar; 3807 } 3808 3809 //------------------------------------------------------------------------------ 3810 // 3811 // scanProp Construct a UnicodeSet from the text at the current scan 3812 // position, which will be of the form \p{whaterver} 3813 // 3814 // The scan position will be at the 'p' or 'P'. On return 3815 // the scan position should be just after the '}' 3816 // 3817 // Return a UnicodeSet, constructed from the \P pattern, 3818 // or NULL if the pattern is invalid. 3819 // 3820 //------------------------------------------------------------------------------ 3821 UnicodeSet *RegexCompile::scanProp() { 3822 UnicodeSet *uset = NULL; 3823 3824 if (U_FAILURE(*fStatus)) { 3825 return NULL; 3826 } 3827 U_ASSERT(fC.fChar == chLowerP || fC.fChar == chP); 3828 UBool negated = (fC.fChar == chP); 3829 3830 UnicodeString propertyName; 3831 nextChar(fC); 3832 if (fC.fChar != chLBrace) { 3833 error(U_REGEX_PROPERTY_SYNTAX); 3834 return NULL; 3835 } 3836 for (;;) { 3837 nextChar(fC); 3838 if (fC.fChar == chRBrace) { 3839 break; 3840 } 3841 if (fC.fChar == -1) { 3842 // Hit the end of the input string without finding the closing '}' 3843 error(U_REGEX_PROPERTY_SYNTAX); 3844 return NULL; 3845 } 3846 propertyName.append(fC.fChar); 3847 } 3848 uset = createSetForProperty(propertyName, negated); 3849 nextChar(fC); // Move input scan to position following the closing '}' 3850 return uset; 3851 } 3852 3853 //------------------------------------------------------------------------------ 3854 // 3855 // scanPosixProp Construct a UnicodeSet from the text at the current scan 3856 // position, which is expected be of the form [:property expression:] 3857 // 3858 // The scan position will be at the opening ':'. On return 3859 // the scan position must be on the closing ']' 3860 // 3861 // Return a UnicodeSet constructed from the pattern, 3862 // or NULL if this is not a valid POSIX-style set expression. 3863 // If not a property expression, restore the initial scan position 3864 // (to the opening ':') 3865 // 3866 // Note: the opening '[:' is not sufficient to guarantee that 3867 // this is a [:property:] expression. 3868 // [:'+=,] is a perfectly good ordinary set expression that 3869 // happens to include ':' as one of its characters. 3870 // 3871 //------------------------------------------------------------------------------ 3872 UnicodeSet *RegexCompile::scanPosixProp() { 3873 UnicodeSet *uset = NULL; 3874 3875 if (U_FAILURE(*fStatus)) { 3876 return NULL; 3877 } 3878 3879 U_ASSERT(fC.fChar == chColon); 3880 3881 // Save the scanner state. 3882 // TODO: move this into the scanner, with the state encapsulated in some way. Ticket 6062 3883 int32_t savedScanIndex = fScanIndex; 3884 int32_t savedNextIndex = fNextIndex; 3885 UBool savedQuoteMode = fQuoteMode; 3886 UBool savedInBackslashQuote = fInBackslashQuote; 3887 UBool savedEOLComments = fEOLComments; 3888 int32_t savedLineNum = fLineNum; 3889 int32_t savedCharNum = fCharNum; 3890 UChar32 savedLastChar = fLastChar; 3891 UChar32 savedPeekChar = fPeekChar; 3892 RegexPatternChar savedfC = fC; 3893 3894 // Scan for a closing ]. A little tricky because there are some perverse 3895 // edge cases possible. "[:abc\Qdef:] \E]" is a valid non-property expression, 3896 // ending on the second closing ]. 3897 3898 UnicodeString propName; 3899 UBool negated = FALSE; 3900 3901 // Check for and consume the '^' in a negated POSIX property, e.g. [:^Letter:] 3902 nextChar(fC); 3903 if (fC.fChar == chUp) { 3904 negated = TRUE; 3905 nextChar(fC); 3906 } 3907 3908 // Scan for the closing ":]", collecting the property name along the way. 3909 UBool sawPropSetTerminator = FALSE; 3910 for (;;) { 3911 propName.append(fC.fChar); 3912 nextChar(fC); 3913 if (fC.fQuoted || fC.fChar == -1) { 3914 // Escaped characters or end of input - either says this isn't a [:Property:] 3915 break; 3916 } 3917 if (fC.fChar == chColon) { 3918 nextChar(fC); 3919 if (fC.fChar == chRBracket) { 3920 sawPropSetTerminator = TRUE; 3921 } 3922 break; 3923 } 3924 } 3925 3926 if (sawPropSetTerminator) { 3927 uset = createSetForProperty(propName, negated); 3928 } 3929 else 3930 { 3931 // No closing ":]". 3932 // Restore the original scan position. 3933 // The main scanner will retry the input as a normal set expression, 3934 // not a [:Property:] expression. 3935 fScanIndex = savedScanIndex; 3936 fNextIndex = savedNextIndex; 3937 fQuoteMode = savedQuoteMode; 3938 fInBackslashQuote = savedInBackslashQuote; 3939 fEOLComments = savedEOLComments; 3940 fLineNum = savedLineNum; 3941 fCharNum = savedCharNum; 3942 fLastChar = savedLastChar; 3943 fPeekChar = savedPeekChar; 3944 fC = savedfC; 3945 } 3946 return uset; 3947 } 3948 3949 static inline void addIdentifierIgnorable(UnicodeSet *set, UErrorCode& ec) { 3950 set->add(0, 8).add(0x0e, 0x1b).add(0x7f, 0x9f); 3951 addCategory(set, U_GC_CF_MASK, ec); 3952 } 3953 3954 // 3955 // Create a Unicode Set from a Unicode Property expression. 3956 // This is common code underlying both \p{...} ane [:...:] expressions. 3957 // Includes trying the Java "properties" that aren't supported as 3958 // normal ICU UnicodeSet properties 3959 // 3960 static const UChar posSetPrefix[] = {0x5b, 0x5c, 0x70, 0x7b, 0}; // "[\p{" 3961 static const UChar negSetPrefix[] = {0x5b, 0x5c, 0x50, 0x7b, 0}; // "[\P{" 3962 UnicodeSet *RegexCompile::createSetForProperty(const UnicodeString &propName, UBool negated) { 3963 UnicodeString setExpr; 3964 UnicodeSet *set; 3965 uint32_t usetFlags = 0; 3966 3967 if (U_FAILURE(*fStatus)) { 3968 return NULL; 3969 } 3970 3971 // 3972 // First try the property as we received it 3973 // 3974 if (negated) { 3975 setExpr.append(negSetPrefix, -1); 3976 } else { 3977 setExpr.append(posSetPrefix, -1); 3978 } 3979 setExpr.append(propName); 3980 setExpr.append(chRBrace); 3981 setExpr.append(chRBracket); 3982 if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 3983 usetFlags |= USET_CASE_INSENSITIVE; 3984 } 3985 set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus); 3986 if (U_SUCCESS(*fStatus)) { 3987 return set; 3988 } 3989 delete set; 3990 set = NULL; 3991 3992 // 3993 // The property as it was didn't work. 3994 // Do emergency fixes - 3995 // InGreek -> InGreek or Coptic, that being the official Unicode name for that block. 3996 // InCombiningMarksforSymbols -> InCombiningDiacriticalMarksforSymbols. 3997 // 3998 // Note on Spaces: either "InCombiningMarksForSymbols" or "InCombining Marks for Symbols" 3999 // is accepted by Java. The property part of the name is compared 4000 // case-insenstively. The spaces must be exactly as shown, either 4001 // all there, or all omitted, with exactly one at each position 4002 // if they are present. From checking against JDK 1.6 4003 // 4004 // This code should be removed when ICU properties support the Java compatibility names 4005 // (ICU 4.0?) 4006 // 4007 UnicodeString mPropName = propName; 4008 if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InGreek"), 0) == 0) { 4009 mPropName = UNICODE_STRING_SIMPLE("InGreek and Coptic"); 4010 } 4011 if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombining Marks for Symbols"), 0) == 0 || 4012 mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombiningMarksforSymbols"), 0) == 0) { 4013 mPropName = UNICODE_STRING_SIMPLE("InCombining Diacritical Marks for Symbols"); 4014 } 4015 else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) { 4016 mPropName = UNICODE_STRING_SIMPLE("javaValidCodePoint"); 4017 } 4018 4019 // See if the property looks like a Java "InBlockName", which 4020 // we will recast as "Block=BlockName" 4021 // 4022 static const UChar IN[] = {0x49, 0x6E, 0}; // "In" 4023 static const UChar BLOCK[] = {0x42, 0x6C, 0x6f, 0x63, 0x6b, 0x3d, 00}; // "Block=" 4024 if (mPropName.startsWith(IN, 2) && propName.length()>=3) { 4025 setExpr.truncate(4); // Leaves "[\p{", or "[\P{" 4026 setExpr.append(BLOCK, -1); 4027 setExpr.append(UnicodeString(mPropName, 2)); // Property with the leading "In" removed. 4028 setExpr.append(chRBrace); 4029 setExpr.append(chRBracket); 4030 *fStatus = U_ZERO_ERROR; 4031 set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus); 4032 if (U_SUCCESS(*fStatus)) { 4033 return set; 4034 } 4035 delete set; 4036 set = NULL; 4037 } 4038 4039 if (propName.startsWith(UNICODE_STRING_SIMPLE("java")) || 4040 propName.compare(UNICODE_STRING_SIMPLE("all")) == 0) 4041 { 4042 UErrorCode localStatus = U_ZERO_ERROR; 4043 //setExpr.remove(); 4044 set = new UnicodeSet(); 4045 // 4046 // Try the various Java specific properties. 4047 // These all begin with "java" 4048 // 4049 if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDefined")) == 0) { 4050 addCategory(set, U_GC_CN_MASK, localStatus); 4051 set->complement(); 4052 } 4053 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDigit")) == 0) { 4054 addCategory(set, U_GC_ND_MASK, localStatus); 4055 } 4056 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaIdentifierIgnorable")) == 0) { 4057 addIdentifierIgnorable(set, localStatus); 4058 } 4059 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaISOControl")) == 0) { 4060 set->add(0, 0x1F).add(0x7F, 0x9F); 4061 } 4062 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierPart")) == 0) { 4063 addCategory(set, U_GC_L_MASK, localStatus); 4064 addCategory(set, U_GC_SC_MASK, localStatus); 4065 addCategory(set, U_GC_PC_MASK, localStatus); 4066 addCategory(set, U_GC_ND_MASK, localStatus); 4067 addCategory(set, U_GC_NL_MASK, localStatus); 4068 addCategory(set, U_GC_MC_MASK, localStatus); 4069 addCategory(set, U_GC_MN_MASK, localStatus); 4070 addIdentifierIgnorable(set, localStatus); 4071 } 4072 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierStart")) == 0) { 4073 addCategory(set, U_GC_L_MASK, localStatus); 4074 addCategory(set, U_GC_NL_MASK, localStatus); 4075 addCategory(set, U_GC_SC_MASK, localStatus); 4076 addCategory(set, U_GC_PC_MASK, localStatus); 4077 } 4078 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetter")) == 0) { 4079 addCategory(set, U_GC_L_MASK, localStatus); 4080 } 4081 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetterOrDigit")) == 0) { 4082 addCategory(set, U_GC_L_MASK, localStatus); 4083 addCategory(set, U_GC_ND_MASK, localStatus); 4084 } 4085 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLowerCase")) == 0) { 4086 addCategory(set, U_GC_LL_MASK, localStatus); 4087 } 4088 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaMirrored")) == 0) { 4089 set->applyIntPropertyValue(UCHAR_BIDI_MIRRORED, 1, localStatus); 4090 } 4091 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSpaceChar")) == 0) { 4092 addCategory(set, U_GC_Z_MASK, localStatus); 4093 } 4094 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSupplementaryCodePoint")) == 0) { 4095 set->add(0x10000, UnicodeSet::MAX_VALUE); 4096 } 4097 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaTitleCase")) == 0) { 4098 addCategory(set, U_GC_LT_MASK, localStatus); 4099 } 4100 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierStart")) == 0) { 4101 addCategory(set, U_GC_L_MASK, localStatus); 4102 addCategory(set, U_GC_NL_MASK, localStatus); 4103 } 4104 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierPart")) == 0) { 4105 addCategory(set, U_GC_L_MASK, localStatus); 4106 addCategory(set, U_GC_PC_MASK, localStatus); 4107 addCategory(set, U_GC_ND_MASK, localStatus); 4108 addCategory(set, U_GC_NL_MASK, localStatus); 4109 addCategory(set, U_GC_MC_MASK, localStatus); 4110 addCategory(set, U_GC_MN_MASK, localStatus); 4111 addIdentifierIgnorable(set, localStatus); 4112 } 4113 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUpperCase")) == 0) { 4114 addCategory(set, U_GC_LU_MASK, localStatus); 4115 } 4116 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaValidCodePoint")) == 0) { 4117 set->add(0, UnicodeSet::MAX_VALUE); 4118 } 4119 else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaWhitespace")) == 0) { 4120 addCategory(set, U_GC_Z_MASK, localStatus); 4121 set->removeAll(UnicodeSet().add(0xa0).add(0x2007).add(0x202f)); 4122 set->add(9, 0x0d).add(0x1c, 0x1f); 4123 } 4124 else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) { 4125 set->add(0, UnicodeSet::MAX_VALUE); 4126 } 4127 4128 if (U_SUCCESS(localStatus) && !set->isEmpty()) { 4129 *fStatus = U_ZERO_ERROR; 4130 if (usetFlags & USET_CASE_INSENSITIVE) { 4131 set->closeOver(USET_CASE_INSENSITIVE); 4132 } 4133 if (negated) { 4134 set->complement(); 4135 } 4136 return set; 4137 } 4138 delete set; 4139 set = NULL; 4140 } 4141 error(*fStatus); 4142 return NULL; 4143 } 4144 4145 4146 4147 // 4148 // SetEval Part of the evaluation of [set expressions]. 4149 // Perform any pending (stacked) operations with precedence 4150 // equal or greater to that of the next operator encountered 4151 // in the expression. 4152 // 4153 void RegexCompile::setEval(int32_t nextOp) { 4154 UnicodeSet *rightOperand = NULL; 4155 UnicodeSet *leftOperand = NULL; 4156 for (;;) { 4157 U_ASSERT(fSetOpStack.empty()==FALSE); 4158 int32_t pendingSetOperation = fSetOpStack.peeki(); 4159 if ((pendingSetOperation&0xffff0000) < (nextOp&0xffff0000)) { 4160 break; 4161 } 4162 fSetOpStack.popi(); 4163 U_ASSERT(fSetStack.empty() == FALSE); 4164 rightOperand = (UnicodeSet *)fSetStack.peek(); 4165 switch (pendingSetOperation) { 4166 case setNegation: 4167 rightOperand->complement(); 4168 break; 4169 case setCaseClose: 4170 // TODO: need a simple close function. Ticket 6065 4171 rightOperand->closeOver(USET_CASE_INSENSITIVE); 4172 rightOperand->removeAllStrings(); 4173 break; 4174 case setDifference1: 4175 case setDifference2: 4176 fSetStack.pop(); 4177 leftOperand = (UnicodeSet *)fSetStack.peek(); 4178 leftOperand->removeAll(*rightOperand); 4179 delete rightOperand; 4180 break; 4181 case setIntersection1: 4182 case setIntersection2: 4183 fSetStack.pop(); 4184 leftOperand = (UnicodeSet *)fSetStack.peek(); 4185 leftOperand->retainAll(*rightOperand); 4186 delete rightOperand; 4187 break; 4188 case setUnion: 4189 fSetStack.pop(); 4190 leftOperand = (UnicodeSet *)fSetStack.peek(); 4191 leftOperand->addAll(*rightOperand); 4192 delete rightOperand; 4193 break; 4194 default: 4195 U_ASSERT(FALSE); 4196 break; 4197 } 4198 } 4199 } 4200 4201 void RegexCompile::setPushOp(int32_t op) { 4202 setEval(op); 4203 fSetOpStack.push(op, *fStatus); 4204 fSetStack.push(new UnicodeSet(), *fStatus); 4205 } 4206 4207 U_NAMESPACE_END 4208 #endif // !UCONFIG_NO_REGULAR_EXPRESSIONS 4209 4210