| File: | out/../deps/icu-small/source/i18n/regexcmp.cpp |
| Warning: | line 406, column 21 Value stored to 'op' during its initialization is never read |
Press '?' to see keyboard shortcuts
Keyboard shortcuts:
| 1 | // © 2016 and later: Unicode, Inc. and others. |
| 2 | // License & terms of use: http://www.unicode.org/copyright.html |
| 3 | // |
| 4 | // file: regexcmp.cpp |
| 5 | // |
| 6 | // Copyright (C) 2002-2016 International Business Machines Corporation and others. |
| 7 | // All Rights Reserved. |
| 8 | // |
| 9 | // This file contains the ICU regular expression compiler, which is responsible |
| 10 | // for processing a regular expression pattern into the compiled form that |
| 11 | // is used by the match finding engine. |
| 12 | // |
| 13 | |
| 14 | #include "unicode/utypes.h" |
| 15 | |
| 16 | #if !UCONFIG_NO_REGULAR_EXPRESSIONS0 |
| 17 | |
| 18 | #include "unicode/ustring.h" |
| 19 | #include "unicode/unistr.h" |
| 20 | #include "unicode/uniset.h" |
| 21 | #include "unicode/uchar.h" |
| 22 | #include "unicode/uchriter.h" |
| 23 | #include "unicode/parsepos.h" |
| 24 | #include "unicode/parseerr.h" |
| 25 | #include "unicode/regex.h" |
| 26 | #include "unicode/utf.h" |
| 27 | #include "unicode/utf16.h" |
| 28 | #include "patternprops.h" |
| 29 | #include "putilimp.h" |
| 30 | #include "cmemory.h" |
| 31 | #include "cstr.h" |
| 32 | #include "cstring.h" |
| 33 | #include "uvectr32.h" |
| 34 | #include "uvectr64.h" |
| 35 | #include "uassert.h" |
| 36 | #include "uinvchar.h" |
| 37 | |
| 38 | #include "regeximp.h" |
| 39 | #include "regexcst.h" // Contains state table for the regex pattern parser. |
| 40 | // generated by a Perl script. |
| 41 | #include "regexcmp.h" |
| 42 | #include "regexst.h" |
| 43 | #include "regextxt.h" |
| 44 | |
| 45 | |
| 46 | |
| 47 | U_NAMESPACE_BEGINnamespace icu_71 { |
| 48 | |
| 49 | |
| 50 | //------------------------------------------------------------------------------ |
| 51 | // |
| 52 | // Constructor. |
| 53 | // |
| 54 | //------------------------------------------------------------------------------ |
| 55 | RegexCompile::RegexCompile(RegexPattern *rxp, UErrorCode &status) : |
| 56 | fParenStack(status), fSetStack(uprv_deleteUObjectuprv_deleteUObject_71, nullptr, status), fSetOpStack(status) |
| 57 | { |
| 58 | // Lazy init of all shared global sets (needed for init()'s empty text) |
| 59 | RegexStaticSets::initGlobals(&status); |
| 60 | |
| 61 | fStatus = &status; |
| 62 | |
| 63 | fRXPat = rxp; |
| 64 | fScanIndex = 0; |
| 65 | fLastChar = -1; |
| 66 | fPeekChar = -1; |
| 67 | fLineNum = 1; |
| 68 | fCharNum = 0; |
| 69 | fQuoteMode = FALSE0; |
| 70 | fInBackslashQuote = FALSE0; |
| 71 | fModeFlags = fRXPat->fFlags | 0x80000000; |
| 72 | fEOLComments = TRUE1; |
| 73 | |
| 74 | fMatchOpenParen = -1; |
| 75 | fMatchCloseParen = -1; |
| 76 | fCaptureName = NULL__null; |
| 77 | fLastSetLiteral = U_SENTINEL(-1); |
| 78 | |
| 79 | if (U_SUCCESS(status) && U_FAILURE(rxp->fDeferredStatus)) { |
| 80 | status = rxp->fDeferredStatus; |
| 81 | } |
| 82 | } |
| 83 | |
| 84 | static const UChar chAmp = 0x26; // '&' |
| 85 | static const UChar chDash = 0x2d; // '-' |
| 86 | |
| 87 | |
| 88 | //------------------------------------------------------------------------------ |
| 89 | // |
| 90 | // Destructor |
| 91 | // |
| 92 | //------------------------------------------------------------------------------ |
| 93 | RegexCompile::~RegexCompile() { |
| 94 | delete fCaptureName; // Normally will be NULL, but can exist if pattern |
| 95 | // compilation stops with a syntax error. |
| 96 | } |
| 97 | |
| 98 | static inline void addCategory(UnicodeSet *set, int32_t value, UErrorCode& ec) { |
| 99 | set->addAll(UnicodeSet().applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, value, ec)); |
| 100 | } |
| 101 | |
| 102 | //------------------------------------------------------------------------------ |
| 103 | // |
| 104 | // Compile regex pattern. The state machine for rexexp pattern parsing is here. |
| 105 | // The state tables are hand-written in the file regexcst.txt, |
| 106 | // and converted to the form used here by a perl |
| 107 | // script regexcst.pl |
| 108 | // |
| 109 | //------------------------------------------------------------------------------ |
| 110 | void RegexCompile::compile( |
| 111 | const UnicodeString &pat, // Source pat to be compiled. |
| 112 | UParseError &pp, // Error position info |
| 113 | UErrorCode &e) // Error Code |
| 114 | { |
| 115 | fRXPat->fPatternString = new UnicodeString(pat); |
| 116 | UText patternText = UTEXT_INITIALIZER{ UTEXT_MAGIC, 0, 0, sizeof(UText), 0, 0, 0, 0, 0, 0, __null, __null, __null, __null, __null, __null, __null, __null, 0, 0 , 0, 0, 0, 0 }; |
| 117 | utext_openConstUnicodeStringutext_openConstUnicodeString_71(&patternText, fRXPat->fPatternString, &e); |
| 118 | |
| 119 | if (U_SUCCESS(e)) { |
| 120 | compile(&patternText, pp, e); |
| 121 | utext_closeutext_close_71(&patternText); |
| 122 | } |
| 123 | } |
| 124 | |
| 125 | // |
| 126 | // compile, UText mode |
| 127 | // All the work is actually done here. |
| 128 | // |
| 129 | void RegexCompile::compile( |
| 130 | UText *pat, // Source pat to be compiled. |
| 131 | UParseError &pp, // Error position info |
| 132 | UErrorCode &e) // Error Code |
| 133 | { |
| 134 | fStatus = &e; |
| 135 | fParseErr = &pp; |
| 136 | fStackPtr = 0; |
| 137 | fStack[fStackPtr] = 0; |
| 138 | |
| 139 | if (U_FAILURE(*fStatus)) { |
| 140 | return; |
| 141 | } |
| 142 | |
| 143 | // There should be no pattern stuff in the RegexPattern object. They can not be reused. |
| 144 | U_ASSERT(fRXPat->fPattern == NULL || utext_nativeLength(fRXPat->fPattern) == 0)(void)0; |
| 145 | |
| 146 | // Prepare the RegexPattern object to receive the compiled pattern. |
| 147 | fRXPat->fPattern = utext_cloneutext_clone_71(fRXPat->fPattern, pat, FALSE0, TRUE1, fStatus); |
| 148 | if (U_FAILURE(*fStatus)) { |
| 149 | return; |
| 150 | } |
| 151 | |
| 152 | // Initialize the pattern scanning state machine |
| 153 | fPatternLength = utext_nativeLengthutext_nativeLength_71(pat); |
| 154 | uint16_t state = 1; |
| 155 | const RegexTableEl *tableEl; |
| 156 | |
| 157 | // UREGEX_LITERAL force entire pattern to be treated as a literal string. |
| 158 | if (fModeFlags & UREGEX_LITERAL) { |
| 159 | fQuoteMode = TRUE1; |
| 160 | } |
| 161 | |
| 162 | nextChar(fC); // Fetch the first char from the pattern string. |
| 163 | |
| 164 | // |
| 165 | // Main loop for the regex pattern parsing state machine. |
| 166 | // Runs once per state transition. |
| 167 | // Each time through optionally performs, depending on the state table, |
| 168 | // - an advance to the the next pattern char |
| 169 | // - an action to be performed. |
| 170 | // - pushing or popping a state to/from the local state return stack. |
| 171 | // file regexcst.txt is the source for the state table. The logic behind |
| 172 | // recongizing the pattern syntax is there, not here. |
| 173 | // |
| 174 | for (;;) { |
| 175 | // Bail out if anything has gone wrong. |
| 176 | // Regex pattern parsing stops on the first error encountered. |
| 177 | if (U_FAILURE(*fStatus)) { |
| 178 | break; |
| 179 | } |
| 180 | |
| 181 | U_ASSERT(state != 0)(void)0; |
| 182 | |
| 183 | // Find the state table element that matches the input char from the pattern, or the |
| 184 | // class of the input character. Start with the first table row for this |
| 185 | // state, then linearly scan forward until we find a row that matches the |
| 186 | // character. The last row for each state always matches all characters, so |
| 187 | // the search will stop there, if not before. |
| 188 | // |
| 189 | tableEl = &gRuleParseStateTable[state]; |
| 190 | REGEX_SCAN_DEBUG_PRINTF(("char, line, col = (\'%c\', %d, %d) state=%s ", |
| 191 | fC.fChar, fLineNum, fCharNum, RegexStateNames[state])); |
| 192 | |
| 193 | for (;;) { // loop through table rows belonging to this state, looking for one |
| 194 | // that matches the current input char. |
| 195 | REGEX_SCAN_DEBUG_PRINTF((".")); |
| 196 | if (tableEl->fCharClass < 127 && fC.fQuoted == FALSE0 && tableEl->fCharClass == fC.fChar) { |
| 197 | // Table row specified an individual character, not a set, and |
| 198 | // the input character is not quoted, and |
| 199 | // the input character matched it. |
| 200 | break; |
| 201 | } |
| 202 | if (tableEl->fCharClass == 255) { |
| 203 | // Table row specified default, match anything character class. |
| 204 | break; |
| 205 | } |
| 206 | if (tableEl->fCharClass == 254 && fC.fQuoted) { |
| 207 | // Table row specified "quoted" and the char was quoted. |
| 208 | break; |
| 209 | } |
| 210 | if (tableEl->fCharClass == 253 && fC.fChar == (UChar32)-1) { |
| 211 | // Table row specified eof and we hit eof on the input. |
| 212 | break; |
| 213 | } |
| 214 | |
| 215 | if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 && // Table specs a char class && |
| 216 | fC.fQuoted == FALSE0 && // char is not escaped && |
| 217 | fC.fChar != (UChar32)-1) { // char is not EOF |
| 218 | U_ASSERT(tableEl->fCharClass <= 137)(void)0; |
| 219 | if (RegexStaticSets::gStaticSets->fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) { |
| 220 | // Table row specified a character class, or set of characters, |
| 221 | // and the current char matches it. |
| 222 | break; |
| 223 | } |
| 224 | } |
| 225 | |
| 226 | // No match on this row, advance to the next row for this state, |
| 227 | tableEl++; |
| 228 | } |
| 229 | REGEX_SCAN_DEBUG_PRINTF(("\n")); |
| 230 | |
| 231 | // |
| 232 | // We've found the row of the state table that matches the current input |
| 233 | // character from the rules string. |
| 234 | // Perform any action specified by this row in the state table. |
| 235 | if (doParseActions(tableEl->fAction) == FALSE0) { |
| 236 | // Break out of the state machine loop if the |
| 237 | // the action signalled some kind of error, or |
| 238 | // the action was to exit, occurs on normal end-of-rules-input. |
| 239 | break; |
| 240 | } |
| 241 | |
| 242 | if (tableEl->fPushState != 0) { |
| 243 | fStackPtr++; |
| 244 | if (fStackPtr >= kStackSize) { |
| 245 | error(U_REGEX_INTERNAL_ERROR); |
| 246 | REGEX_SCAN_DEBUG_PRINTF(("RegexCompile::parse() - state stack overflow.\n")); |
| 247 | fStackPtr--; |
| 248 | } |
| 249 | fStack[fStackPtr] = tableEl->fPushState; |
| 250 | } |
| 251 | |
| 252 | // |
| 253 | // NextChar. This is where characters are actually fetched from the pattern. |
| 254 | // Happens under control of the 'n' tag in the state table. |
| 255 | // |
| 256 | if (tableEl->fNextChar) { |
| 257 | nextChar(fC); |
| 258 | } |
| 259 | |
| 260 | // Get the next state from the table entry, or from the |
| 261 | // state stack if the next state was specified as "pop". |
| 262 | if (tableEl->fNextState != 255) { |
| 263 | state = tableEl->fNextState; |
| 264 | } else { |
| 265 | state = fStack[fStackPtr]; |
| 266 | fStackPtr--; |
| 267 | if (fStackPtr < 0) { |
| 268 | // state stack underflow |
| 269 | // This will occur if the user pattern has mis-matched parentheses, |
| 270 | // with extra close parens. |
| 271 | // |
| 272 | fStackPtr++; |
| 273 | error(U_REGEX_MISMATCHED_PAREN); |
| 274 | } |
| 275 | } |
| 276 | |
| 277 | } |
| 278 | |
| 279 | if (U_FAILURE(*fStatus)) { |
| 280 | // Bail out if the pattern had errors. |
| 281 | return; |
| 282 | } |
| 283 | |
| 284 | // |
| 285 | // The pattern has now been read and processed, and the compiled code generated. |
| 286 | // |
| 287 | |
| 288 | // |
| 289 | // The pattern's fFrameSize so far has accumulated the requirements for |
| 290 | // storage for capture parentheses, counters, etc. that are encountered |
| 291 | // in the pattern. Add space for the two variables that are always |
| 292 | // present in the saved state: the input string position (int64_t) and |
| 293 | // the position in the compiled pattern. |
| 294 | // |
| 295 | allocateStackData(RESTACKFRAME_HDRCOUNT2); |
| 296 | |
| 297 | // |
| 298 | // Optimization pass 1: NOPs, back-references, and case-folding |
| 299 | // |
| 300 | stripNOPs(); |
| 301 | |
| 302 | // |
| 303 | // Get bounds for the minimum and maximum length of a string that this |
| 304 | // pattern can match. Used to avoid looking for matches in strings that |
| 305 | // are too short. |
| 306 | // |
| 307 | fRXPat->fMinMatchLen = minMatchLength(3, fRXPat->fCompiledPat->size()-1); |
| 308 | |
| 309 | // |
| 310 | // Optimization pass 2: match start type |
| 311 | // |
| 312 | matchStartType(); |
| 313 | |
| 314 | // |
| 315 | // Set up fast latin-1 range sets |
| 316 | // |
| 317 | int32_t numSets = fRXPat->fSets->size(); |
| 318 | fRXPat->fSets8 = new Regex8BitSet[numSets]; |
| 319 | // Null pointer check. |
| 320 | if (fRXPat->fSets8 == NULL__null) { |
| 321 | e = *fStatus = U_MEMORY_ALLOCATION_ERROR; |
| 322 | return; |
| 323 | } |
| 324 | int32_t i; |
| 325 | for (i=0; i<numSets; i++) { |
| 326 | UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(i); |
| 327 | fRXPat->fSets8[i].init(s); |
| 328 | } |
| 329 | |
| 330 | } |
| 331 | |
| 332 | |
| 333 | |
| 334 | |
| 335 | |
| 336 | //------------------------------------------------------------------------------ |
| 337 | // |
| 338 | // doParseAction Do some action during regex pattern parsing. |
| 339 | // Called by the parse state machine. |
| 340 | // |
| 341 | // Generation of the match engine PCode happens here, or |
| 342 | // in functions called from the parse actions defined here. |
| 343 | // |
| 344 | // |
| 345 | //------------------------------------------------------------------------------ |
| 346 | UBool RegexCompile::doParseActions(int32_t action) |
| 347 | { |
| 348 | UBool returnVal = TRUE1; |
| 349 | |
| 350 | switch ((Regex_PatternParseAction)action) { |
| 351 | |
| 352 | case doPatStart: |
| 353 | // Start of pattern compiles to: |
| 354 | //0 SAVE 2 Fall back to position of FAIL |
| 355 | //1 jmp 3 |
| 356 | //2 FAIL Stop if we ever reach here. |
| 357 | //3 NOP Dummy, so start of pattern looks the same as |
| 358 | // the start of an ( grouping. |
| 359 | //4 NOP Resreved, will be replaced by a save if there are |
| 360 | // OR | operators at the top level |
| 361 | appendOp(URX_STATE_SAVE, 2); |
| 362 | appendOp(URX_JMP, 3); |
| 363 | appendOp(URX_FAIL, 0); |
| 364 | |
| 365 | // Standard open nonCapture paren action emits the two NOPs and |
| 366 | // sets up the paren stack frame. |
| 367 | doParseActions(doOpenNonCaptureParen); |
| 368 | break; |
| 369 | |
| 370 | case doPatFinish: |
| 371 | // We've scanned to the end of the pattern |
| 372 | // The end of pattern compiles to: |
| 373 | // URX_END |
| 374 | // which will stop the runtime match engine. |
| 375 | // Encountering end of pattern also behaves like a close paren, |
| 376 | // and forces fixups of the State Save at the beginning of the compiled pattern |
| 377 | // and of any OR operations at the top level. |
| 378 | // |
| 379 | handleCloseParen(); |
| 380 | if (fParenStack.size() > 0) { |
| 381 | // Missing close paren in pattern. |
| 382 | error(U_REGEX_MISMATCHED_PAREN); |
| 383 | } |
| 384 | |
| 385 | // add the END operation to the compiled pattern. |
| 386 | appendOp(URX_END, 0); |
| 387 | |
| 388 | // Terminate the pattern compilation state machine. |
| 389 | returnVal = FALSE0; |
| 390 | break; |
| 391 | |
| 392 | |
| 393 | |
| 394 | case doOrOperator: |
| 395 | // Scanning a '|', as in (A|B) |
| 396 | { |
| 397 | // Generate code for any pending literals preceding the '|' |
| 398 | fixLiterals(FALSE0); |
| 399 | |
| 400 | // Insert a SAVE operation at the start of the pattern section preceding |
| 401 | // this OR at this level. This SAVE will branch the match forward |
| 402 | // to the right hand side of the OR in the event that the left hand |
| 403 | // side fails to match and backtracks. Locate the position for the |
| 404 | // save from the location on the top of the parentheses stack. |
| 405 | int32_t savePosition = fParenStack.popi(); |
| 406 | int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(savePosition); |
Value stored to 'op' during its initialization is never read | |
| 407 | U_ASSERT(URX_TYPE(op) == URX_NOP)(void)0; // original contents of reserved location |
| 408 | op = buildOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+1); |
| 409 | fRXPat->fCompiledPat->setElementAt(op, savePosition); |
| 410 | |
| 411 | // Append an JMP operation into the compiled pattern. The operand for |
| 412 | // the JMP will eventually be the location following the ')' for the |
| 413 | // group. This will be patched in later, when the ')' is encountered. |
| 414 | appendOp(URX_JMP, 0); |
| 415 | |
| 416 | // Push the position of the newly added JMP op onto the parentheses stack. |
| 417 | // This registers if for fixup when this block's close paren is encountered. |
| 418 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); |
| 419 | |
| 420 | // Append a NOP to the compiled pattern. This is the slot reserved |
| 421 | // for a SAVE in the event that there is yet another '|' following |
| 422 | // this one. |
| 423 | appendOp(URX_NOP, 0); |
| 424 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); |
| 425 | } |
| 426 | break; |
| 427 | |
| 428 | |
| 429 | case doBeginNamedCapture: |
| 430 | // Scanning (?<letter. |
| 431 | // The first letter of the name will come through again under doConinueNamedCapture. |
| 432 | fCaptureName = new UnicodeString(); |
| 433 | if (fCaptureName == NULL__null) { |
| 434 | error(U_MEMORY_ALLOCATION_ERROR); |
| 435 | } |
| 436 | break; |
| 437 | |
| 438 | case doContinueNamedCapture: |
| 439 | fCaptureName->append(fC.fChar); |
| 440 | break; |
| 441 | |
| 442 | case doBadNamedCapture: |
| 443 | error(U_REGEX_INVALID_CAPTURE_GROUP_NAME); |
| 444 | break; |
| 445 | |
| 446 | case doOpenCaptureParen: |
| 447 | // Open Capturing Paren, possibly named. |
| 448 | // Compile to a |
| 449 | // - NOP, which later may be replaced by a save-state if the |
| 450 | // parenthesized group gets a * quantifier, followed by |
| 451 | // - START_CAPTURE n where n is stack frame offset to the capture group variables. |
| 452 | // - NOP, which may later be replaced by a save-state if there |
| 453 | // is an '|' alternation within the parens. |
| 454 | // |
| 455 | // Each capture group gets three slots in the save stack frame: |
| 456 | // 0: Capture Group start position (in input string being matched.) |
| 457 | // 1: Capture Group end position. |
| 458 | // 2: Start of Match-in-progress. |
| 459 | // The first two locations are for a completed capture group, and are |
| 460 | // referred to by back references and the like. |
| 461 | // The third location stores the capture start position when an START_CAPTURE is |
| 462 | // encountered. This will be promoted to a completed capture when (and if) the corresponding |
| 463 | // END_CAPTURE is encountered. |
| 464 | { |
| 465 | fixLiterals(); |
| 466 | appendOp(URX_NOP, 0); |
| 467 | int32_t varsLoc = allocateStackData(3); // Reserve three slots in match stack frame. |
| 468 | appendOp(URX_START_CAPTURE, varsLoc); |
| 469 | appendOp(URX_NOP, 0); |
| 470 | |
| 471 | // On the Parentheses stack, start a new frame and add the positions |
| 472 | // of the two NOPs. Depending on what follows in the pattern, the |
| 473 | // NOPs may be changed to SAVE_STATE or JMP ops, with a target |
| 474 | // address of the end of the parenthesized group. |
| 475 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 476 | fParenStack.push(capturing, *fStatus); // Frame type. |
| 477 | fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus); // The first NOP location |
| 478 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP loc |
| 479 | |
| 480 | // Save the mapping from group number to stack frame variable position. |
| 481 | fRXPat->fGroupMap->addElement(varsLoc, *fStatus); |
| 482 | |
| 483 | // If this is a named capture group, add the name->group number mapping. |
| 484 | if (fCaptureName != NULL__null) { |
| 485 | if (!fRXPat->initNamedCaptureMap()) { |
| 486 | if (U_SUCCESS(*fStatus)) { |
| 487 | error(fRXPat->fDeferredStatus); |
| 488 | } |
| 489 | break; |
| 490 | } |
| 491 | int32_t groupNumber = fRXPat->fGroupMap->size(); |
| 492 | int32_t previousMapping = uhash_putiuhash_puti_71(fRXPat->fNamedCaptureMap, fCaptureName, groupNumber, fStatus); |
| 493 | fCaptureName = NULL__null; // hash table takes ownership of the name (key) string. |
| 494 | if (previousMapping > 0 && U_SUCCESS(*fStatus)) { |
| 495 | error(U_REGEX_INVALID_CAPTURE_GROUP_NAME); |
| 496 | } |
| 497 | } |
| 498 | } |
| 499 | break; |
| 500 | |
| 501 | case doOpenNonCaptureParen: |
| 502 | // Open non-caputuring (grouping only) Paren. |
| 503 | // Compile to a |
| 504 | // - NOP, which later may be replaced by a save-state if the |
| 505 | // parenthesized group gets a * quantifier, followed by |
| 506 | // - NOP, which may later be replaced by a save-state if there |
| 507 | // is an '|' alternation within the parens. |
| 508 | { |
| 509 | fixLiterals(); |
| 510 | appendOp(URX_NOP, 0); |
| 511 | appendOp(URX_NOP, 0); |
| 512 | |
| 513 | // On the Parentheses stack, start a new frame and add the positions |
| 514 | // of the two NOPs. |
| 515 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 516 | fParenStack.push(plain, *fStatus); // Begin a new frame. |
| 517 | fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location |
| 518 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP loc |
| 519 | } |
| 520 | break; |
| 521 | |
| 522 | |
| 523 | case doOpenAtomicParen: |
| 524 | // Open Atomic Paren. (?> |
| 525 | // Compile to a |
| 526 | // - NOP, which later may be replaced if the parenthesized group |
| 527 | // has a quantifier, followed by |
| 528 | // - STO_SP save state stack position, so it can be restored at the ")" |
| 529 | // - NOP, which may later be replaced by a save-state if there |
| 530 | // is an '|' alternation within the parens. |
| 531 | { |
| 532 | fixLiterals(); |
| 533 | appendOp(URX_NOP, 0); |
| 534 | int32_t varLoc = allocateData(1); // Reserve a data location for saving the state stack ptr. |
| 535 | appendOp(URX_STO_SP, varLoc); |
| 536 | appendOp(URX_NOP, 0); |
| 537 | |
| 538 | // On the Parentheses stack, start a new frame and add the positions |
| 539 | // of the two NOPs. Depending on what follows in the pattern, the |
| 540 | // NOPs may be changed to SAVE_STATE or JMP ops, with a target |
| 541 | // address of the end of the parenthesized group. |
| 542 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 543 | fParenStack.push(atomic, *fStatus); // Frame type. |
| 544 | fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus); // The first NOP |
| 545 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP |
| 546 | } |
| 547 | break; |
| 548 | |
| 549 | |
| 550 | case doOpenLookAhead: |
| 551 | // Positive Look-ahead (?= stuff ) |
| 552 | // |
| 553 | // Note: Addition of transparent input regions, with the need to |
| 554 | // restore the original regions when failing out of a lookahead |
| 555 | // block, complicated this sequence. Some combined opcodes |
| 556 | // might make sense - or might not, lookahead aren't that common. |
| 557 | // |
| 558 | // Caution: min match length optimization knows about this |
| 559 | // sequence; don't change without making updates there too. |
| 560 | // |
| 561 | // Compiles to |
| 562 | // 1 LA_START dataLoc Saves SP, Input Pos, Active input region. |
| 563 | // 2. STATE_SAVE 4 on failure of lookahead, goto 4 |
| 564 | // 3 JMP 6 continue ... |
| 565 | // |
| 566 | // 4. LA_END Look Ahead failed. Restore regions. |
| 567 | // 5. BACKTRACK and back track again. |
| 568 | // |
| 569 | // 6. NOP reserved for use by quantifiers on the block. |
| 570 | // Look-ahead can't have quantifiers, but paren stack |
| 571 | // compile time conventions require the slot anyhow. |
| 572 | // 7. NOP may be replaced if there is are '|' ops in the block. |
| 573 | // 8. code for parenthesized stuff. |
| 574 | // 9. LA_END |
| 575 | // |
| 576 | // Four data slots are reserved, for saving state on entry to the look-around |
| 577 | // 0: stack pointer on entry. |
| 578 | // 1: input position on entry. |
| 579 | // 2: fActiveStart, the active bounds start on entry. |
| 580 | // 3: fActiveLimit, the active bounds limit on entry. |
| 581 | { |
| 582 | fixLiterals(); |
| 583 | int32_t dataLoc = allocateData(4); |
| 584 | appendOp(URX_LA_START, dataLoc); |
| 585 | appendOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+ 2); |
| 586 | appendOp(URX_JMP, fRXPat->fCompiledPat->size()+ 3); |
| 587 | appendOp(URX_LA_END, dataLoc); |
| 588 | appendOp(URX_BACKTRACK, 0); |
| 589 | appendOp(URX_NOP, 0); |
| 590 | appendOp(URX_NOP, 0); |
| 591 | |
| 592 | // On the Parentheses stack, start a new frame and add the positions |
| 593 | // of the NOPs. |
| 594 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 595 | fParenStack.push(lookAhead, *fStatus); // Frame type. |
| 596 | fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location |
| 597 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP location |
| 598 | } |
| 599 | break; |
| 600 | |
| 601 | case doOpenLookAheadNeg: |
| 602 | // Negated Lookahead. (?! stuff ) |
| 603 | // Compiles to |
| 604 | // 1. LA_START dataloc |
| 605 | // 2. SAVE_STATE 7 // Fail within look-ahead block restores to this state, |
| 606 | // // which continues with the match. |
| 607 | // 3. NOP // Std. Open Paren sequence, for possible '|' |
| 608 | // 4. code for parenthesized stuff. |
| 609 | // 5. LA_END // Cut back stack, remove saved state from step 2. |
| 610 | // 6. BACKTRACK // code in block succeeded, so neg. lookahead fails. |
| 611 | // 7. END_LA // Restore match region, in case look-ahead was using |
| 612 | // an alternate (transparent) region. |
| 613 | // Four data slots are reserved, for saving state on entry to the look-around |
| 614 | // 0: stack pointer on entry. |
| 615 | // 1: input position on entry. |
| 616 | // 2: fActiveStart, the active bounds start on entry. |
| 617 | // 3: fActiveLimit, the active bounds limit on entry. |
| 618 | { |
| 619 | fixLiterals(); |
| 620 | int32_t dataLoc = allocateData(4); |
| 621 | appendOp(URX_LA_START, dataLoc); |
| 622 | appendOp(URX_STATE_SAVE, 0); // dest address will be patched later. |
| 623 | appendOp(URX_NOP, 0); |
| 624 | |
| 625 | // On the Parentheses stack, start a new frame and add the positions |
| 626 | // of the StateSave and NOP. |
| 627 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 628 | fParenStack.push(negLookAhead, *fStatus); // Frame type |
| 629 | fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The STATE_SAVE location |
| 630 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP location |
| 631 | |
| 632 | // Instructions #5 - #7 will be added when the ')' is encountered. |
| 633 | } |
| 634 | break; |
| 635 | |
| 636 | case doOpenLookBehind: |
| 637 | { |
| 638 | // Compile a (?<= look-behind open paren. |
| 639 | // |
| 640 | // Compiles to |
| 641 | // 0 URX_LB_START dataLoc |
| 642 | // 1 URX_LB_CONT dataLoc |
| 643 | // 2 MinMatchLen |
| 644 | // 3 MaxMatchLen |
| 645 | // 4 URX_NOP Standard '(' boilerplate. |
| 646 | // 5 URX_NOP Reserved slot for use with '|' ops within (block). |
| 647 | // 6 <code for LookBehind expression> |
| 648 | // 7 URX_LB_END dataLoc # Check match len, restore input len |
| 649 | // 8 URX_LA_END dataLoc # Restore stack, input pos |
| 650 | // |
| 651 | // Allocate a block of matcher data, to contain (when running a match) |
| 652 | // 0: Stack ptr on entry |
| 653 | // 1: Input Index on entry |
| 654 | // 2: fActiveStart, the active bounds start on entry. |
| 655 | // 3: fActiveLimit, the active bounds limit on entry. |
| 656 | // 4: Start index of match current match attempt. |
| 657 | // The first four items must match the layout of data for LA_START / LA_END |
| 658 | |
| 659 | // Generate match code for any pending literals. |
| 660 | fixLiterals(); |
| 661 | |
| 662 | // Allocate data space |
| 663 | int32_t dataLoc = allocateData(5); |
| 664 | |
| 665 | // Emit URX_LB_START |
| 666 | appendOp(URX_LB_START, dataLoc); |
| 667 | |
| 668 | // Emit URX_LB_CONT |
| 669 | appendOp(URX_LB_CONT, dataLoc); |
| 670 | appendOp(URX_RESERVED_OP, 0); // MinMatchLength. To be filled later. |
| 671 | appendOp(URX_RESERVED_OP, 0); // MaxMatchLength. To be filled later. |
| 672 | |
| 673 | // Emit the NOPs |
| 674 | appendOp(URX_NOP, 0); |
| 675 | appendOp(URX_NOP, 0); |
| 676 | |
| 677 | // On the Parentheses stack, start a new frame and add the positions |
| 678 | // of the URX_LB_CONT and the NOP. |
| 679 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 680 | fParenStack.push(lookBehind, *fStatus); // Frame type |
| 681 | fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location |
| 682 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The 2nd NOP location |
| 683 | |
| 684 | // The final two instructions will be added when the ')' is encountered. |
| 685 | } |
| 686 | |
| 687 | break; |
| 688 | |
| 689 | case doOpenLookBehindNeg: |
| 690 | { |
| 691 | // Compile a (?<! negated look-behind open paren. |
| 692 | // |
| 693 | // Compiles to |
| 694 | // 0 URX_LB_START dataLoc # Save entry stack, input len |
| 695 | // 1 URX_LBN_CONT dataLoc # Iterate possible match positions |
| 696 | // 2 MinMatchLen |
| 697 | // 3 MaxMatchLen |
| 698 | // 4 continueLoc (9) |
| 699 | // 5 URX_NOP Standard '(' boilerplate. |
| 700 | // 6 URX_NOP Reserved slot for use with '|' ops within (block). |
| 701 | // 7 <code for LookBehind expression> |
| 702 | // 8 URX_LBN_END dataLoc # Check match len, cause a FAIL |
| 703 | // 9 ... |
| 704 | // |
| 705 | // Allocate a block of matcher data, to contain (when running a match) |
| 706 | // 0: Stack ptr on entry |
| 707 | // 1: Input Index on entry |
| 708 | // 2: fActiveStart, the active bounds start on entry. |
| 709 | // 3: fActiveLimit, the active bounds limit on entry. |
| 710 | // 4: Start index of match current match attempt. |
| 711 | // The first four items must match the layout of data for LA_START / LA_END |
| 712 | |
| 713 | // Generate match code for any pending literals. |
| 714 | fixLiterals(); |
| 715 | |
| 716 | // Allocate data space |
| 717 | int32_t dataLoc = allocateData(5); |
| 718 | |
| 719 | // Emit URX_LB_START |
| 720 | appendOp(URX_LB_START, dataLoc); |
| 721 | |
| 722 | // Emit URX_LBN_CONT |
| 723 | appendOp(URX_LBN_CONT, dataLoc); |
| 724 | appendOp(URX_RESERVED_OP, 0); // MinMatchLength. To be filled later. |
| 725 | appendOp(URX_RESERVED_OP, 0); // MaxMatchLength. To be filled later. |
| 726 | appendOp(URX_RESERVED_OP, 0); // Continue Loc. To be filled later. |
| 727 | |
| 728 | // Emit the NOPs |
| 729 | appendOp(URX_NOP, 0); |
| 730 | appendOp(URX_NOP, 0); |
| 731 | |
| 732 | // On the Parentheses stack, start a new frame and add the positions |
| 733 | // of the URX_LB_CONT and the NOP. |
| 734 | fParenStack.push(fModeFlags, *fStatus); // Match mode state |
| 735 | fParenStack.push(lookBehindN, *fStatus); // Frame type |
| 736 | fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location |
| 737 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The 2nd NOP location |
| 738 | |
| 739 | // The final two instructions will be added when the ')' is encountered. |
| 740 | } |
| 741 | break; |
| 742 | |
| 743 | case doConditionalExpr: |
| 744 | // Conditionals such as (?(1)a:b) |
| 745 | case doPerlInline: |
| 746 | // Perl inline-conditionals. (?{perl code}a|b) We're not perl, no way to do them. |
| 747 | error(U_REGEX_UNIMPLEMENTED); |
| 748 | break; |
| 749 | |
| 750 | |
| 751 | case doCloseParen: |
| 752 | handleCloseParen(); |
| 753 | if (fParenStack.size() <= 0) { |
| 754 | // Extra close paren, or missing open paren. |
| 755 | error(U_REGEX_MISMATCHED_PAREN); |
| 756 | } |
| 757 | break; |
| 758 | |
| 759 | case doNOP: |
| 760 | break; |
| 761 | |
| 762 | |
| 763 | case doBadOpenParenType: |
| 764 | case doRuleError: |
| 765 | error(U_REGEX_RULE_SYNTAX); |
| 766 | break; |
| 767 | |
| 768 | |
| 769 | case doMismatchedParenErr: |
| 770 | error(U_REGEX_MISMATCHED_PAREN); |
| 771 | break; |
| 772 | |
| 773 | case doPlus: |
| 774 | // Normal '+' compiles to |
| 775 | // 1. stuff to be repeated (already built) |
| 776 | // 2. jmp-sav 1 |
| 777 | // 3. ... |
| 778 | // |
| 779 | // Or, if the item to be repeated can match a zero length string, |
| 780 | // 1. STO_INP_LOC data-loc |
| 781 | // 2. body of stuff to be repeated |
| 782 | // 3. JMP_SAV_X 2 |
| 783 | // 4. ... |
| 784 | |
| 785 | // |
| 786 | // Or, if the item to be repeated is simple |
| 787 | // 1. Item to be repeated. |
| 788 | // 2. LOOP_SR_I set number (assuming repeated item is a set ref) |
| 789 | // 3. LOOP_C stack location |
| 790 | { |
| 791 | int32_t topLoc = blockTopLoc(FALSE0); // location of item #1 |
| 792 | int32_t frameLoc; |
| 793 | |
| 794 | // Check for simple constructs, which may get special optimized code. |
| 795 | if (topLoc == fRXPat->fCompiledPat->size() - 1) { |
| 796 | int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc); |
| 797 | |
| 798 | if (URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_SETREF) { |
| 799 | // Emit optimized code for [char set]+ |
| 800 | appendOp(URX_LOOP_SR_I, URX_VAL(repeatedOp)((repeatedOp) & 0xffffff)); |
| 801 | frameLoc = allocateStackData(1); |
| 802 | appendOp(URX_LOOP_C, frameLoc); |
| 803 | break; |
| 804 | } |
| 805 | |
| 806 | if (URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY || |
| 807 | URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY_ALL || |
| 808 | URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY_UNIX) { |
| 809 | // Emit Optimized code for .+ operations. |
| 810 | int32_t loopOpI = buildOp(URX_LOOP_DOT_I, 0); |
| 811 | if (URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY_ALL) { |
| 812 | // URX_LOOP_DOT_I operand is a flag indicating ". matches any" mode. |
| 813 | loopOpI |= 1; |
| 814 | } |
| 815 | if (fModeFlags & UREGEX_UNIX_LINES) { |
| 816 | loopOpI |= 2; |
| 817 | } |
| 818 | appendOp(loopOpI); |
| 819 | frameLoc = allocateStackData(1); |
| 820 | appendOp(URX_LOOP_C, frameLoc); |
| 821 | break; |
| 822 | } |
| 823 | |
| 824 | } |
| 825 | |
| 826 | // General case. |
| 827 | |
| 828 | // Check for minimum match length of zero, which requires |
| 829 | // extra loop-breaking code. |
| 830 | if (minMatchLength(topLoc, fRXPat->fCompiledPat->size()-1) == 0) { |
| 831 | // Zero length match is possible. |
| 832 | // Emit the code sequence that can handle it. |
| 833 | insertOp(topLoc); |
| 834 | frameLoc = allocateStackData(1); |
| 835 | |
| 836 | int32_t op = buildOp(URX_STO_INP_LOC, frameLoc); |
| 837 | fRXPat->fCompiledPat->setElementAt(op, topLoc); |
| 838 | |
| 839 | appendOp(URX_JMP_SAV_X, topLoc+1); |
| 840 | } else { |
| 841 | // Simpler code when the repeated body must match something non-empty |
| 842 | appendOp(URX_JMP_SAV, topLoc); |
| 843 | } |
| 844 | } |
| 845 | break; |
| 846 | |
| 847 | case doNGPlus: |
| 848 | // Non-greedy '+?' compiles to |
| 849 | // 1. stuff to be repeated (already built) |
| 850 | // 2. state-save 1 |
| 851 | // 3. ... |
| 852 | { |
| 853 | int32_t topLoc = blockTopLoc(FALSE0); |
| 854 | appendOp(URX_STATE_SAVE, topLoc); |
| 855 | } |
| 856 | break; |
| 857 | |
| 858 | |
| 859 | case doOpt: |
| 860 | // Normal (greedy) ? quantifier. |
| 861 | // Compiles to |
| 862 | // 1. state save 3 |
| 863 | // 2. body of optional block |
| 864 | // 3. ... |
| 865 | // Insert the state save into the compiled pattern, and we're done. |
| 866 | { |
| 867 | int32_t saveStateLoc = blockTopLoc(TRUE1); |
| 868 | int32_t saveStateOp = buildOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()); |
| 869 | fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc); |
| 870 | } |
| 871 | break; |
| 872 | |
| 873 | case doNGOpt: |
| 874 | // Non-greedy ?? quantifier |
| 875 | // compiles to |
| 876 | // 1. jmp 4 |
| 877 | // 2. body of optional block |
| 878 | // 3 jmp 5 |
| 879 | // 4. state save 2 |
| 880 | // 5 ... |
| 881 | // This code is less than ideal, with two jmps instead of one, because we can only |
| 882 | // insert one instruction at the top of the block being iterated. |
| 883 | { |
| 884 | int32_t jmp1_loc = blockTopLoc(TRUE1); |
| 885 | int32_t jmp2_loc = fRXPat->fCompiledPat->size(); |
| 886 | |
| 887 | int32_t jmp1_op = buildOp(URX_JMP, jmp2_loc+1); |
| 888 | fRXPat->fCompiledPat->setElementAt(jmp1_op, jmp1_loc); |
| 889 | |
| 890 | appendOp(URX_JMP, jmp2_loc+2); |
| 891 | |
| 892 | appendOp(URX_STATE_SAVE, jmp1_loc+1); |
| 893 | } |
| 894 | break; |
| 895 | |
| 896 | |
| 897 | case doStar: |
| 898 | // Normal (greedy) * quantifier. |
| 899 | // Compiles to |
| 900 | // 1. STATE_SAVE 4 |
| 901 | // 2. body of stuff being iterated over |
| 902 | // 3. JMP_SAV 2 |
| 903 | // 4. ... |
| 904 | // |
| 905 | // Or, if the body is a simple [Set], |
| 906 | // 1. LOOP_SR_I set number |
| 907 | // 2. LOOP_C stack location |
| 908 | // ... |
| 909 | // |
| 910 | // Or if this is a .* |
| 911 | // 1. LOOP_DOT_I (. matches all mode flag) |
| 912 | // 2. LOOP_C stack location |
| 913 | // |
| 914 | // Or, if the body can match a zero-length string, to inhibit infinite loops, |
| 915 | // 1. STATE_SAVE 5 |
| 916 | // 2. STO_INP_LOC data-loc |
| 917 | // 3. body of stuff |
| 918 | // 4. JMP_SAV_X 2 |
| 919 | // 5. ... |
| 920 | { |
| 921 | // location of item #1, the STATE_SAVE |
| 922 | int32_t topLoc = blockTopLoc(FALSE0); |
| 923 | int32_t dataLoc = -1; |
| 924 | |
| 925 | // Check for simple *, where the construct being repeated |
| 926 | // compiled to single opcode, and might be optimizable. |
| 927 | if (topLoc == fRXPat->fCompiledPat->size() - 1) { |
| 928 | int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc); |
| 929 | |
| 930 | if (URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_SETREF) { |
| 931 | // Emit optimized code for a [char set]* |
| 932 | int32_t loopOpI = buildOp(URX_LOOP_SR_I, URX_VAL(repeatedOp)((repeatedOp) & 0xffffff)); |
| 933 | fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc); |
| 934 | dataLoc = allocateStackData(1); |
| 935 | appendOp(URX_LOOP_C, dataLoc); |
| 936 | break; |
| 937 | } |
| 938 | |
| 939 | if (URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY || |
| 940 | URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY_ALL || |
| 941 | URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY_UNIX) { |
| 942 | // Emit Optimized code for .* operations. |
| 943 | int32_t loopOpI = buildOp(URX_LOOP_DOT_I, 0); |
| 944 | if (URX_TYPE(repeatedOp)((uint32_t)(repeatedOp) >> 24) == URX_DOTANY_ALL) { |
| 945 | // URX_LOOP_DOT_I operand is a flag indicating . matches any mode. |
| 946 | loopOpI |= 1; |
| 947 | } |
| 948 | if ((fModeFlags & UREGEX_UNIX_LINES) != 0) { |
| 949 | loopOpI |= 2; |
| 950 | } |
| 951 | fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc); |
| 952 | dataLoc = allocateStackData(1); |
| 953 | appendOp(URX_LOOP_C, dataLoc); |
| 954 | break; |
| 955 | } |
| 956 | } |
| 957 | |
| 958 | // Emit general case code for this * |
| 959 | // The optimizations did not apply. |
| 960 | |
| 961 | int32_t saveStateLoc = blockTopLoc(TRUE1); |
| 962 | int32_t jmpOp = buildOp(URX_JMP_SAV, saveStateLoc+1); |
| 963 | |
| 964 | // Check for minimum match length of zero, which requires |
| 965 | // extra loop-breaking code. |
| 966 | if (minMatchLength(saveStateLoc, fRXPat->fCompiledPat->size()-1) == 0) { |
| 967 | insertOp(saveStateLoc); |
| 968 | dataLoc = allocateStackData(1); |
| 969 | |
| 970 | int32_t op = buildOp(URX_STO_INP_LOC, dataLoc); |
| 971 | fRXPat->fCompiledPat->setElementAt(op, saveStateLoc+1); |
| 972 | jmpOp = buildOp(URX_JMP_SAV_X, saveStateLoc+2); |
| 973 | } |
| 974 | |
| 975 | // Locate the position in the compiled pattern where the match will continue |
| 976 | // after completing the *. (4 or 5 in the comment above) |
| 977 | int32_t continueLoc = fRXPat->fCompiledPat->size()+1; |
| 978 | |
| 979 | // Put together the save state op and store it into the compiled code. |
| 980 | int32_t saveStateOp = buildOp(URX_STATE_SAVE, continueLoc); |
| 981 | fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc); |
| 982 | |
| 983 | // Append the URX_JMP_SAV or URX_JMPX operation to the compiled pattern. |
| 984 | appendOp(jmpOp); |
| 985 | } |
| 986 | break; |
| 987 | |
| 988 | case doNGStar: |
| 989 | // Non-greedy *? quantifier |
| 990 | // compiles to |
| 991 | // 1. JMP 3 |
| 992 | // 2. body of stuff being iterated over |
| 993 | // 3. STATE_SAVE 2 |
| 994 | // 4 ... |
| 995 | { |
| 996 | int32_t jmpLoc = blockTopLoc(TRUE1); // loc 1. |
| 997 | int32_t saveLoc = fRXPat->fCompiledPat->size(); // loc 3. |
| 998 | int32_t jmpOp = buildOp(URX_JMP, saveLoc); |
| 999 | fRXPat->fCompiledPat->setElementAt(jmpOp, jmpLoc); |
| 1000 | appendOp(URX_STATE_SAVE, jmpLoc+1); |
| 1001 | } |
| 1002 | break; |
| 1003 | |
| 1004 | |
| 1005 | case doIntervalInit: |
| 1006 | // The '{' opening an interval quantifier was just scanned. |
| 1007 | // Init the counter variables that will accumulate the values as the digits |
| 1008 | // are scanned. |
| 1009 | fIntervalLow = 0; |
| 1010 | fIntervalUpper = -1; |
| 1011 | break; |
| 1012 | |
| 1013 | case doIntevalLowerDigit: |
| 1014 | // Scanned a digit from the lower value of an {lower,upper} interval |
| 1015 | { |
| 1016 | int32_t digitValue = u_charDigitValueu_charDigitValue_71(fC.fChar); |
| 1017 | U_ASSERT(digitValue >= 0)(void)0; |
| 1018 | int64_t val = (int64_t)fIntervalLow*10 + digitValue; |
| 1019 | if (val > INT32_MAX(2147483647)) { |
| 1020 | error(U_REGEX_NUMBER_TOO_BIG); |
| 1021 | } else { |
| 1022 | fIntervalLow = (int32_t)val; |
| 1023 | } |
| 1024 | } |
| 1025 | break; |
| 1026 | |
| 1027 | case doIntervalUpperDigit: |
| 1028 | // Scanned a digit from the upper value of an {lower,upper} interval |
| 1029 | { |
| 1030 | if (fIntervalUpper < 0) { |
| 1031 | fIntervalUpper = 0; |
| 1032 | } |
| 1033 | int32_t digitValue = u_charDigitValueu_charDigitValue_71(fC.fChar); |
| 1034 | U_ASSERT(digitValue >= 0)(void)0; |
| 1035 | int64_t val = (int64_t)fIntervalUpper*10 + digitValue; |
| 1036 | if (val > INT32_MAX(2147483647)) { |
| 1037 | error(U_REGEX_NUMBER_TOO_BIG); |
| 1038 | } else { |
| 1039 | fIntervalUpper = (int32_t)val; |
| 1040 | } |
| 1041 | } |
| 1042 | break; |
| 1043 | |
| 1044 | case doIntervalSame: |
| 1045 | // Scanned a single value interval like {27}. Upper = Lower. |
| 1046 | fIntervalUpper = fIntervalLow; |
| 1047 | break; |
| 1048 | |
| 1049 | case doInterval: |
| 1050 | // Finished scanning a normal {lower,upper} interval. Generate the code for it. |
| 1051 | if (compileInlineInterval() == FALSE0) { |
| 1052 | compileInterval(URX_CTR_INIT, URX_CTR_LOOP); |
| 1053 | } |
| 1054 | break; |
| 1055 | |
| 1056 | case doPossessiveInterval: |
| 1057 | // Finished scanning a Possessive {lower,upper}+ interval. Generate the code for it. |
| 1058 | { |
| 1059 | // Remember the loc for the top of the block being looped over. |
| 1060 | // (Can not reserve a slot in the compiled pattern at this time, because |
| 1061 | // compileInterval needs to reserve also, and blockTopLoc can only reserve |
| 1062 | // once per block.) |
| 1063 | int32_t topLoc = blockTopLoc(FALSE0); |
| 1064 | |
| 1065 | // Produce normal looping code. |
| 1066 | compileInterval(URX_CTR_INIT, URX_CTR_LOOP); |
| 1067 | |
| 1068 | // Surround the just-emitted normal looping code with a STO_SP ... LD_SP |
| 1069 | // just as if the loop was inclosed in atomic parentheses. |
| 1070 | |
| 1071 | // First the STO_SP before the start of the loop |
| 1072 | insertOp(topLoc); |
| 1073 | |
| 1074 | int32_t varLoc = allocateData(1); // Reserve a data location for saving the |
| 1075 | int32_t op = buildOp(URX_STO_SP, varLoc); |
| 1076 | fRXPat->fCompiledPat->setElementAt(op, topLoc); |
| 1077 | |
| 1078 | int32_t loopOp = (int32_t)fRXPat->fCompiledPat->popi(); |
| 1079 | U_ASSERT(URX_TYPE(loopOp) == URX_CTR_LOOP && URX_VAL(loopOp) == topLoc)(void)0; |
| 1080 | loopOp++; // point LoopOp after the just-inserted STO_SP |
| 1081 | fRXPat->fCompiledPat->push(loopOp, *fStatus); |
| 1082 | |
| 1083 | // Then the LD_SP after the end of the loop |
| 1084 | appendOp(URX_LD_SP, varLoc); |
| 1085 | } |
| 1086 | |
| 1087 | break; |
| 1088 | |
| 1089 | case doNGInterval: |
| 1090 | // Finished scanning a non-greedy {lower,upper}? interval. Generate the code for it. |
| 1091 | compileInterval(URX_CTR_INIT_NG, URX_CTR_LOOP_NG); |
| 1092 | break; |
| 1093 | |
| 1094 | case doIntervalError: |
| 1095 | error(U_REGEX_BAD_INTERVAL); |
| 1096 | break; |
| 1097 | |
| 1098 | case doLiteralChar: |
| 1099 | // We've just scanned a "normal" character from the pattern, |
| 1100 | literalChar(fC.fChar); |
| 1101 | break; |
| 1102 | |
| 1103 | |
| 1104 | case doEscapedLiteralChar: |
| 1105 | // We've just scanned an backslashed escaped character with no |
| 1106 | // special meaning. It represents itself. |
| 1107 | if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 && |
| 1108 | ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) || // in [A-Z] |
| 1109 | (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) { // in [a-z] |
| 1110 | error(U_REGEX_BAD_ESCAPE_SEQUENCE); |
| 1111 | } |
| 1112 | literalChar(fC.fChar); |
| 1113 | break; |
| 1114 | |
| 1115 | |
| 1116 | case doDotAny: |
| 1117 | // scanned a ".", match any single character. |
| 1118 | { |
| 1119 | fixLiterals(FALSE0); |
| 1120 | if (fModeFlags & UREGEX_DOTALL) { |
| 1121 | appendOp(URX_DOTANY_ALL, 0); |
| 1122 | } else if (fModeFlags & UREGEX_UNIX_LINES) { |
| 1123 | appendOp(URX_DOTANY_UNIX, 0); |
| 1124 | } else { |
| 1125 | appendOp(URX_DOTANY, 0); |
| 1126 | } |
| 1127 | } |
| 1128 | break; |
| 1129 | |
| 1130 | case doCaret: |
| 1131 | { |
| 1132 | fixLiterals(FALSE0); |
| 1133 | if ( (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { |
| 1134 | appendOp(URX_CARET, 0); |
| 1135 | } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { |
| 1136 | appendOp(URX_CARET_M, 0); |
| 1137 | } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { |
| 1138 | appendOp(URX_CARET, 0); // Only testing true start of input. |
| 1139 | } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { |
| 1140 | appendOp(URX_CARET_M_UNIX, 0); |
| 1141 | } |
| 1142 | } |
| 1143 | break; |
| 1144 | |
| 1145 | case doDollar: |
| 1146 | { |
| 1147 | fixLiterals(FALSE0); |
| 1148 | if ( (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { |
| 1149 | appendOp(URX_DOLLAR, 0); |
| 1150 | } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { |
| 1151 | appendOp(URX_DOLLAR_M, 0); |
| 1152 | } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { |
| 1153 | appendOp(URX_DOLLAR_D, 0); |
| 1154 | } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { |
| 1155 | appendOp(URX_DOLLAR_MD, 0); |
| 1156 | } |
| 1157 | } |
| 1158 | break; |
| 1159 | |
| 1160 | case doBackslashA: |
| 1161 | fixLiterals(FALSE0); |
| 1162 | appendOp(URX_CARET, 0); |
| 1163 | break; |
| 1164 | |
| 1165 | case doBackslashB: |
| 1166 | { |
| 1167 | #if UCONFIG_NO_BREAK_ITERATION0==1 |
| 1168 | if (fModeFlags & UREGEX_UWORD) { |
| 1169 | error(U_UNSUPPORTED_ERROR); |
| 1170 | } |
| 1171 | #endif |
| 1172 | fixLiterals(FALSE0); |
| 1173 | int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B; |
| 1174 | appendOp(op, 1); |
| 1175 | } |
| 1176 | break; |
| 1177 | |
| 1178 | case doBackslashb: |
| 1179 | { |
| 1180 | #if UCONFIG_NO_BREAK_ITERATION0==1 |
| 1181 | if (fModeFlags & UREGEX_UWORD) { |
| 1182 | error(U_UNSUPPORTED_ERROR); |
| 1183 | } |
| 1184 | #endif |
| 1185 | fixLiterals(FALSE0); |
| 1186 | int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B; |
| 1187 | appendOp(op, 0); |
| 1188 | } |
| 1189 | break; |
| 1190 | |
| 1191 | case doBackslashD: |
| 1192 | fixLiterals(FALSE0); |
| 1193 | appendOp(URX_BACKSLASH_D, 1); |
| 1194 | break; |
| 1195 | |
| 1196 | case doBackslashd: |
| 1197 | fixLiterals(FALSE0); |
| 1198 | appendOp(URX_BACKSLASH_D, 0); |
| 1199 | break; |
| 1200 | |
| 1201 | case doBackslashG: |
| 1202 | fixLiterals(FALSE0); |
| 1203 | appendOp(URX_BACKSLASH_G, 0); |
| 1204 | break; |
| 1205 | |
| 1206 | case doBackslashH: |
| 1207 | fixLiterals(FALSE0); |
| 1208 | appendOp(URX_BACKSLASH_H, 1); |
| 1209 | break; |
| 1210 | |
| 1211 | case doBackslashh: |
| 1212 | fixLiterals(FALSE0); |
| 1213 | appendOp(URX_BACKSLASH_H, 0); |
| 1214 | break; |
| 1215 | |
| 1216 | case doBackslashR: |
| 1217 | fixLiterals(FALSE0); |
| 1218 | appendOp(URX_BACKSLASH_R, 0); |
| 1219 | break; |
| 1220 | |
| 1221 | case doBackslashS: |
| 1222 | fixLiterals(FALSE0); |
| 1223 | appendOp(URX_STAT_SETREF_N, URX_ISSPACE_SET); |
| 1224 | break; |
| 1225 | |
| 1226 | case doBackslashs: |
| 1227 | fixLiterals(FALSE0); |
| 1228 | appendOp(URX_STATIC_SETREF, URX_ISSPACE_SET); |
| 1229 | break; |
| 1230 | |
| 1231 | case doBackslashV: |
| 1232 | fixLiterals(FALSE0); |
| 1233 | appendOp(URX_BACKSLASH_V, 1); |
| 1234 | break; |
| 1235 | |
| 1236 | case doBackslashv: |
| 1237 | fixLiterals(FALSE0); |
| 1238 | appendOp(URX_BACKSLASH_V, 0); |
| 1239 | break; |
| 1240 | |
| 1241 | case doBackslashW: |
| 1242 | fixLiterals(FALSE0); |
| 1243 | appendOp(URX_STAT_SETREF_N, URX_ISWORD_SET); |
| 1244 | break; |
| 1245 | |
| 1246 | case doBackslashw: |
| 1247 | fixLiterals(FALSE0); |
| 1248 | appendOp(URX_STATIC_SETREF, URX_ISWORD_SET); |
| 1249 | break; |
| 1250 | |
| 1251 | case doBackslashX: |
| 1252 | #if UCONFIG_NO_BREAK_ITERATION0==1 |
| 1253 | // Grapheme Cluster Boundary requires ICU break iteration. |
| 1254 | error(U_UNSUPPORTED_ERROR); |
| 1255 | #endif |
| 1256 | fixLiterals(FALSE0); |
| 1257 | appendOp(URX_BACKSLASH_X, 0); |
| 1258 | break; |
| 1259 | |
| 1260 | case doBackslashZ: |
| 1261 | fixLiterals(FALSE0); |
| 1262 | appendOp(URX_DOLLAR, 0); |
| 1263 | break; |
| 1264 | |
| 1265 | case doBackslashz: |
| 1266 | fixLiterals(FALSE0); |
| 1267 | appendOp(URX_BACKSLASH_Z, 0); |
| 1268 | break; |
| 1269 | |
| 1270 | case doEscapeError: |
| 1271 | error(U_REGEX_BAD_ESCAPE_SEQUENCE); |
| 1272 | break; |
| 1273 | |
| 1274 | case doExit: |
| 1275 | fixLiterals(FALSE0); |
| 1276 | returnVal = FALSE0; |
| 1277 | break; |
| 1278 | |
| 1279 | case doProperty: |
| 1280 | { |
| 1281 | fixLiterals(FALSE0); |
| 1282 | UnicodeSet *theSet = scanProp(); |
| 1283 | compileSet(theSet); |
| 1284 | } |
| 1285 | break; |
| 1286 | |
| 1287 | case doNamedChar: |
| 1288 | { |
| 1289 | UChar32 c = scanNamedChar(); |
| 1290 | literalChar(c); |
| 1291 | } |
| 1292 | break; |
| 1293 | |
| 1294 | |
| 1295 | case doBackRef: |
| 1296 | // BackReference. Somewhat unusual in that the front-end can not completely parse |
| 1297 | // the regular expression, because the number of digits to be consumed |
| 1298 | // depends on the number of capture groups that have been defined. So |
| 1299 | // we have to do it here instead. |
| 1300 | { |
| 1301 | int32_t numCaptureGroups = fRXPat->fGroupMap->size(); |
| 1302 | int32_t groupNum = 0; |
| 1303 | UChar32 c = fC.fChar; |
| 1304 | |
| 1305 | for (;;) { |
| 1306 | // Loop once per digit, for max allowed number of digits in a back reference. |
| 1307 | int32_t digit = u_charDigitValueu_charDigitValue_71(c); |
| 1308 | groupNum = groupNum * 10 + digit; |
| 1309 | if (groupNum >= numCaptureGroups) { |
| 1310 | break; |
| 1311 | } |
| 1312 | c = peekCharLL(); |
| 1313 | if (RegexStaticSets::gStaticSets->fRuleDigitsAlias->contains(c) == FALSE0) { |
| 1314 | break; |
| 1315 | } |
| 1316 | nextCharLL(); |
| 1317 | } |
| 1318 | |
| 1319 | // Scan of the back reference in the source regexp is complete. Now generate |
| 1320 | // the compiled code for it. |
| 1321 | // Because capture groups can be forward-referenced by back-references, |
| 1322 | // we fill the operand with the capture group number. At the end |
| 1323 | // of compilation, it will be changed to the variable's location. |
| 1324 | U_ASSERT(groupNum > 0)(void)0; // Shouldn't happen. '\0' begins an octal escape sequence, |
| 1325 | // and shouldn't enter this code path at all. |
| 1326 | fixLiterals(FALSE0); |
| 1327 | if (fModeFlags & UREGEX_CASE_INSENSITIVE) { |
| 1328 | appendOp(URX_BACKREF_I, groupNum); |
| 1329 | } else { |
| 1330 | appendOp(URX_BACKREF, groupNum); |
| 1331 | } |
| 1332 | } |
| 1333 | break; |
| 1334 | |
| 1335 | case doBeginNamedBackRef: |
| 1336 | U_ASSERT(fCaptureName == NULL)(void)0; |
| 1337 | fCaptureName = new UnicodeString; |
| 1338 | if (fCaptureName == NULL__null) { |
| 1339 | error(U_MEMORY_ALLOCATION_ERROR); |
| 1340 | } |
| 1341 | break; |
| 1342 | |
| 1343 | case doContinueNamedBackRef: |
| 1344 | fCaptureName->append(fC.fChar); |
| 1345 | break; |
| 1346 | |
| 1347 | case doCompleteNamedBackRef: |
| 1348 | { |
| 1349 | int32_t groupNumber = |
| 1350 | fRXPat->fNamedCaptureMap ? uhash_getiuhash_geti_71(fRXPat->fNamedCaptureMap, fCaptureName) : 0; |
| 1351 | if (groupNumber == 0) { |
| 1352 | // Group name has not been defined. |
| 1353 | // Could be a forward reference. If we choose to support them at some |
| 1354 | // future time, extra mechanism will be required at this point. |
| 1355 | error(U_REGEX_INVALID_CAPTURE_GROUP_NAME); |
| 1356 | } else { |
| 1357 | // Given the number, handle identically to a \n numbered back reference. |
| 1358 | // See comments above, under doBackRef |
| 1359 | fixLiterals(FALSE0); |
| 1360 | if (fModeFlags & UREGEX_CASE_INSENSITIVE) { |
| 1361 | appendOp(URX_BACKREF_I, groupNumber); |
| 1362 | } else { |
| 1363 | appendOp(URX_BACKREF, groupNumber); |
| 1364 | } |
| 1365 | } |
| 1366 | delete fCaptureName; |
| 1367 | fCaptureName = NULL__null; |
| 1368 | break; |
| 1369 | } |
| 1370 | |
| 1371 | case doPossessivePlus: |
| 1372 | // Possessive ++ quantifier. |
| 1373 | // Compiles to |
| 1374 | // 1. STO_SP |
| 1375 | // 2. body of stuff being iterated over |
| 1376 | // 3. STATE_SAVE 5 |
| 1377 | // 4. JMP 2 |
| 1378 | // 5. LD_SP |
| 1379 | // 6. ... |
| 1380 | // |
| 1381 | // Note: TODO: This is pretty inefficient. A mass of saved state is built up |
| 1382 | // then unconditionally discarded. Perhaps introduce a new opcode. Ticket 6056 |
| 1383 | // |
| 1384 | { |
| 1385 | // Emit the STO_SP |
| 1386 | int32_t topLoc = blockTopLoc(TRUE1); |
| 1387 | int32_t stoLoc = allocateData(1); // Reserve the data location for storing save stack ptr. |
| 1388 | int32_t op = buildOp(URX_STO_SP, stoLoc); |
| 1389 | fRXPat->fCompiledPat->setElementAt(op, topLoc); |
| 1390 | |
| 1391 | // Emit the STATE_SAVE |
| 1392 | appendOp(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+2); |
| 1393 | |
| 1394 | // Emit the JMP |
| 1395 | appendOp(URX_JMP, topLoc+1); |
| 1396 | |
| 1397 | // Emit the LD_SP |
| 1398 | appendOp(URX_LD_SP, stoLoc); |
| 1399 | } |
| 1400 | break; |
| 1401 | |
| 1402 | case doPossessiveStar: |
| 1403 | // Possessive *+ quantifier. |
| 1404 | // Compiles to |
| 1405 | // 1. STO_SP loc |
| 1406 | // 2. STATE_SAVE 5 |
| 1407 | // 3. body of stuff being iterated over |
| 1408 | // 4. JMP 2 |
| 1409 | // 5. LD_SP loc |
| 1410 | // 6 ... |
| 1411 | // TODO: do something to cut back the state stack each time through the loop. |
| 1412 | { |
| 1413 | // Reserve two slots at the top of the block. |
| 1414 | int32_t topLoc = blockTopLoc(TRUE1); |
| 1415 | insertOp(topLoc); |
| 1416 | |
| 1417 | // emit STO_SP loc |
| 1418 | int32_t stoLoc = allocateData(1); // Reserve the data location for storing save stack ptr. |
| 1419 | int32_t op = buildOp(URX_STO_SP, stoLoc); |
| 1420 | fRXPat->fCompiledPat->setElementAt(op, topLoc); |
| 1421 | |
| 1422 | // Emit the SAVE_STATE 5 |
| 1423 | int32_t L7 = fRXPat->fCompiledPat->size()+1; |
| 1424 | op = buildOp(URX_STATE_SAVE, L7); |
| 1425 | fRXPat->fCompiledPat->setElementAt(op, topLoc+1); |
| 1426 | |
| 1427 | // Append the JMP operation. |
| 1428 | appendOp(URX_JMP, topLoc+1); |
| 1429 | |
| 1430 | // Emit the LD_SP loc |
| 1431 | appendOp(URX_LD_SP, stoLoc); |
| 1432 | } |
| 1433 | break; |
| 1434 | |
| 1435 | case doPossessiveOpt: |
| 1436 | // Possessive ?+ quantifier. |
| 1437 | // Compiles to |
| 1438 | // 1. STO_SP loc |
| 1439 | // 2. SAVE_STATE 5 |
| 1440 | // 3. body of optional block |
| 1441 | // 4. LD_SP loc |
| 1442 | // 5. ... |
| 1443 | // |
| 1444 | { |
| 1445 | // Reserve two slots at the top of the block. |
| 1446 | int32_t topLoc = blockTopLoc(TRUE1); |
| 1447 | insertOp(topLoc); |
| 1448 | |
| 1449 | // Emit the STO_SP |
| 1450 | int32_t stoLoc = allocateData(1); // Reserve the data location for storing save stack ptr. |
| 1451 | int32_t op = buildOp(URX_STO_SP, stoLoc); |
| 1452 | fRXPat->fCompiledPat->setElementAt(op, topLoc); |
| 1453 | |
| 1454 | // Emit the SAVE_STATE |
| 1455 | int32_t continueLoc = fRXPat->fCompiledPat->size()+1; |
| 1456 | op = buildOp(URX_STATE_SAVE, continueLoc); |
| 1457 | fRXPat->fCompiledPat->setElementAt(op, topLoc+1); |
| 1458 | |
| 1459 | // Emit the LD_SP |
| 1460 | appendOp(URX_LD_SP, stoLoc); |
| 1461 | } |
| 1462 | break; |
| 1463 | |
| 1464 | |
| 1465 | case doBeginMatchMode: |
| 1466 | fNewModeFlags = fModeFlags; |
| 1467 | fSetModeFlag = TRUE1; |
| 1468 | break; |
| 1469 | |
| 1470 | case doMatchMode: // (?i) and similar |
| 1471 | { |
| 1472 | int32_t bit = 0; |
| 1473 | switch (fC.fChar) { |
| 1474 | case 0x69: /* 'i' */ bit = UREGEX_CASE_INSENSITIVE; break; |
| 1475 | case 0x64: /* 'd' */ bit = UREGEX_UNIX_LINES; break; |
| 1476 | case 0x6d: /* 'm' */ bit = UREGEX_MULTILINE; break; |
| 1477 | case 0x73: /* 's' */ bit = UREGEX_DOTALL; break; |
| 1478 | case 0x75: /* 'u' */ bit = 0; /* Unicode casing */ break; |
| 1479 | case 0x77: /* 'w' */ bit = UREGEX_UWORD; break; |
| 1480 | case 0x78: /* 'x' */ bit = UREGEX_COMMENTS; break; |
| 1481 | case 0x2d: /* '-' */ fSetModeFlag = FALSE0; break; |
| 1482 | default: |
| 1483 | UPRV_UNREACHABLE_EXITabort(); // Should never happen. Other chars are filtered out |
| 1484 | // by the scanner. |
| 1485 | } |
| 1486 | if (fSetModeFlag) { |
| 1487 | fNewModeFlags |= bit; |
| 1488 | } else { |
| 1489 | fNewModeFlags &= ~bit; |
| 1490 | } |
| 1491 | } |
| 1492 | break; |
| 1493 | |
| 1494 | case doSetMatchMode: |
| 1495 | // Emit code to match any pending literals, using the not-yet changed match mode. |
| 1496 | fixLiterals(); |
| 1497 | |
| 1498 | // We've got a (?i) or similar. The match mode is being changed, but |
| 1499 | // the change is not scoped to a parenthesized block. |
| 1500 | U_ASSERT(fNewModeFlags < 0)(void)0; |
| 1501 | fModeFlags = fNewModeFlags; |
| 1502 | |
| 1503 | break; |
| 1504 | |
| 1505 | |
| 1506 | case doMatchModeParen: |
| 1507 | // We've got a (?i: or similar. Begin a parenthesized block, save old |
| 1508 | // mode flags so they can be restored at the close of the block. |
| 1509 | // |
| 1510 | // Compile to a |
| 1511 | // - NOP, which later may be replaced by a save-state if the |
| 1512 | // parenthesized group gets a * quantifier, followed by |
| 1513 | // - NOP, which may later be replaced by a save-state if there |
| 1514 | // is an '|' alternation within the parens. |
| 1515 | { |
| 1516 | fixLiterals(FALSE0); |
| 1517 | appendOp(URX_NOP, 0); |
| 1518 | appendOp(URX_NOP, 0); |
| 1519 | |
| 1520 | // On the Parentheses stack, start a new frame and add the positions |
| 1521 | // of the two NOPs (a normal non-capturing () frame, except for the |
| 1522 | // saving of the original mode flags.) |
| 1523 | fParenStack.push(fModeFlags, *fStatus); |
| 1524 | fParenStack.push(flags, *fStatus); // Frame Marker |
| 1525 | fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP |
| 1526 | fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP |
| 1527 | |
| 1528 | // Set the current mode flags to the new values. |
| 1529 | U_ASSERT(fNewModeFlags < 0)(void)0; |
| 1530 | fModeFlags = fNewModeFlags; |
| 1531 | } |
| 1532 | break; |
| 1533 | |
| 1534 | case doBadModeFlag: |
| 1535 | error(U_REGEX_INVALID_FLAG); |
| 1536 | break; |
| 1537 | |
| 1538 | case doSuppressComments: |
| 1539 | // We have just scanned a '(?'. We now need to prevent the character scanner from |
| 1540 | // treating a '#' as a to-the-end-of-line comment. |
| 1541 | // (This Perl compatibility just gets uglier and uglier to do...) |
| 1542 | fEOLComments = FALSE0; |
| 1543 | break; |
| 1544 | |
| 1545 | |
| 1546 | case doSetAddAmp: |
| 1547 | { |
| 1548 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1549 | set->add(chAmp); |
| 1550 | } |
| 1551 | break; |
| 1552 | |
| 1553 | case doSetAddDash: |
| 1554 | { |
| 1555 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1556 | set->add(chDash); |
| 1557 | } |
| 1558 | break; |
| 1559 | |
| 1560 | case doSetBackslash_s: |
| 1561 | { |
| 1562 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1563 | set->addAll(RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]); |
| 1564 | break; |
| 1565 | } |
| 1566 | |
| 1567 | case doSetBackslash_S: |
| 1568 | { |
| 1569 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1570 | UnicodeSet SSet; |
| 1571 | SSet.addAll(RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]).complement(); |
| 1572 | set->addAll(SSet); |
| 1573 | break; |
| 1574 | } |
| 1575 | |
| 1576 | case doSetBackslash_d: |
| 1577 | { |
| 1578 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1579 | // TODO - make a static set, ticket 6058. |
| 1580 | addCategory(set, U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), *fStatus); |
| 1581 | break; |
| 1582 | } |
| 1583 | |
| 1584 | case doSetBackslash_D: |
| 1585 | { |
| 1586 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1587 | UnicodeSet digits; |
| 1588 | // TODO - make a static set, ticket 6058. |
| 1589 | digits.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), *fStatus); |
| 1590 | digits.complement(); |
| 1591 | set->addAll(digits); |
| 1592 | break; |
| 1593 | } |
| 1594 | |
| 1595 | case doSetBackslash_h: |
| 1596 | { |
| 1597 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1598 | UnicodeSet h; |
| 1599 | h.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ZS_MASK((uint32_t)1<<(U_SPACE_SEPARATOR)), *fStatus); |
| 1600 | h.add((UChar32)9); // Tab |
| 1601 | set->addAll(h); |
| 1602 | break; |
| 1603 | } |
| 1604 | |
| 1605 | case doSetBackslash_H: |
| 1606 | { |
| 1607 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1608 | UnicodeSet h; |
| 1609 | h.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ZS_MASK((uint32_t)1<<(U_SPACE_SEPARATOR)), *fStatus); |
| 1610 | h.add((UChar32)9); // Tab |
| 1611 | h.complement(); |
| 1612 | set->addAll(h); |
| 1613 | break; |
| 1614 | } |
| 1615 | |
| 1616 | case doSetBackslash_v: |
| 1617 | { |
| 1618 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1619 | set->add((UChar32)0x0a, (UChar32)0x0d); // add range |
| 1620 | set->add((UChar32)0x85); |
| 1621 | set->add((UChar32)0x2028, (UChar32)0x2029); |
| 1622 | break; |
| 1623 | } |
| 1624 | |
| 1625 | case doSetBackslash_V: |
| 1626 | { |
| 1627 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1628 | UnicodeSet v; |
| 1629 | v.add((UChar32)0x0a, (UChar32)0x0d); // add range |
| 1630 | v.add((UChar32)0x85); |
| 1631 | v.add((UChar32)0x2028, (UChar32)0x2029); |
| 1632 | v.complement(); |
| 1633 | set->addAll(v); |
| 1634 | break; |
| 1635 | } |
| 1636 | |
| 1637 | case doSetBackslash_w: |
| 1638 | { |
| 1639 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1640 | set->addAll(RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]); |
| 1641 | break; |
| 1642 | } |
| 1643 | |
| 1644 | case doSetBackslash_W: |
| 1645 | { |
| 1646 | UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); |
| 1647 | UnicodeSet SSet; |
| 1648 | SSet.addAll(RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]).complement(); |
| 1649 | set->addAll(SSet); |
| 1650 | break; |
| 1651 | } |
| 1652 | |
| 1653 | case doSetBegin: |
| 1654 | { |
| 1655 | fixLiterals(FALSE0); |
| 1656 | LocalPointer<UnicodeSet> lpSet(new UnicodeSet(), *fStatus); |
| 1657 | fSetStack.push(lpSet.orphan(), *fStatus); |
| 1658 | fSetOpStack.push(setStart, *fStatus); |
| 1659 | if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { |
| 1660 | fSetOpStack.push(setCaseClose, *fStatus); |
| 1661 | } |
| 1662 | break; |
| 1663 | } |
| 1664 | |
| 1665 | case doSetBeginDifference1: |
| 1666 | // We have scanned something like [[abc]-[ |
| 1667 | // Set up a new UnicodeSet for the set beginning with the just-scanned '[' |
| 1668 | // Push a Difference operator, which will cause the new set to be subtracted from what |
| 1669 | // went before once it is created. |
| 1670 | setPushOp(setDifference1); |
| 1671 | fSetOpStack.push(setStart, *fStatus); |
| 1672 | if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { |
| 1673 | fSetOpStack.push(setCaseClose, *fStatus); |
| 1674 | } |
| 1675 | break; |
| 1676 | |
| 1677 | case doSetBeginIntersection1: |
| 1678 | // We have scanned something like [[abc]&[ |
| 1679 | // Need both the '&' operator and the open '[' operator. |
| 1680 | setPushOp(setIntersection1); |
| 1681 | fSetOpStack.push(setStart, *fStatus); |
| 1682 | if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { |
| 1683 | fSetOpStack.push(setCaseClose, *fStatus); |
| 1684 | } |
| 1685 | break; |
| 1686 | |
| 1687 | case doSetBeginUnion: |
| 1688 | // We have scanned something like [[abc][ |
| 1689 | // Need to handle the union operation explicitly [[abc] | [ |
| 1690 | setPushOp(setUnion); |
| 1691 | fSetOpStack.push(setStart, *fStatus); |
| 1692 | if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { |
| 1693 | fSetOpStack.push(setCaseClose, *fStatus); |
| 1694 | } |
| 1695 | break; |
| 1696 | |
| 1697 | case doSetDifference2: |
| 1698 | // We have scanned something like [abc-- |
| 1699 | // Consider this to unambiguously be a set difference operator. |
| 1700 | setPushOp(setDifference2); |
| 1701 | break; |
| 1702 | |
| 1703 | case doSetEnd: |
| 1704 | // Have encountered the ']' that closes a set. |
| 1705 | // Force the evaluation of any pending operations within this set, |
| 1706 | // leave the completed set on the top of the set stack. |
| 1707 | setEval(setEnd); |
| 1708 | U_ASSERT(fSetOpStack.peeki()==setStart)(void)0; |
| 1709 | fSetOpStack.popi(); |
| 1710 | break; |
| 1711 | |
| 1712 | case doSetFinish: |
| 1713 | { |
| 1714 | // Finished a complete set expression, including all nested sets. |
| 1715 | // The close bracket has already triggered clearing out pending set operators, |
| 1716 | // the operator stack should be empty and the operand stack should have just |
| 1717 | // one entry, the result set. |
| 1718 | U_ASSERT(fSetOpStack.empty())(void)0; |
| 1719 | UnicodeSet *theSet = (UnicodeSet *)fSetStack.pop(); |
| 1720 | U_ASSERT(fSetStack.empty())(void)0; |
| 1721 | compileSet(theSet); |
| 1722 | break; |
| 1723 | } |
| 1724 | |
| 1725 | case doSetIntersection2: |
| 1726 | // Have scanned something like [abc&& |
| 1727 | setPushOp(setIntersection2); |
| 1728 | break; |
| 1729 | |
| 1730 | case doSetLiteral: |
| 1731 | // Union the just-scanned literal character into the set being built. |
| 1732 | // This operation is the highest precedence set operation, so we can always do |
| 1733 | // it immediately, without waiting to see what follows. It is necessary to perform |
| 1734 | // any pending '-' or '&' operation first, because these have the same precedence |
| 1735 | // as union-ing in a literal' |
| 1736 | { |
| 1737 | setEval(setUnion); |
| 1738 | UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); |
| 1739 | s->add(fC.fChar); |
| 1740 | fLastSetLiteral = fC.fChar; |
| 1741 | break; |
| 1742 | } |
| 1743 | |
| 1744 | case doSetLiteralEscaped: |
| 1745 | // A back-slash escaped literal character was encountered. |
| 1746 | // Processing is the same as with setLiteral, above, with the addition of |
| 1747 | // the optional check for errors on escaped ASCII letters. |
| 1748 | { |
| 1749 | if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 && |
| 1750 | ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) || // in [A-Z] |
| 1751 | (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) { // in [a-z] |
| 1752 | error(U_REGEX_BAD_ESCAPE_SEQUENCE); |
| 1753 | } |
| 1754 | setEval(setUnion); |
| 1755 | UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); |
| 1756 | s->add(fC.fChar); |
| 1757 | fLastSetLiteral = fC.fChar; |
| 1758 | break; |
| 1759 | } |
| 1760 | |
| 1761 | case doSetNamedChar: |
| 1762 | // Scanning a \N{UNICODE CHARACTER NAME} |
| 1763 | // Aside from the source of the character, the processing is identical to doSetLiteral, |
| 1764 | // above. |
| 1765 | { |
| 1766 | UChar32 c = scanNamedChar(); |
| 1767 | setEval(setUnion); |
| 1768 | UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); |
| 1769 | s->add(c); |
| 1770 | fLastSetLiteral = c; |
| 1771 | break; |
| 1772 | } |
| 1773 | |
| 1774 | case doSetNamedRange: |
| 1775 | // We have scanned literal-\N{CHAR NAME}. Add the range to the set. |
| 1776 | // The left character is already in the set, and is saved in fLastSetLiteral. |
| 1777 | // The right side needs to be picked up, the scan is at the 'N'. |
| 1778 | // Lower Limit > Upper limit being an error matches both Java |
| 1779 | // and ICU UnicodeSet behavior. |
| 1780 | { |
| 1781 | UChar32 c = scanNamedChar(); |
| 1782 | if (U_SUCCESS(*fStatus) && (fLastSetLiteral == U_SENTINEL(-1) || fLastSetLiteral > c)) { |
| 1783 | error(U_REGEX_INVALID_RANGE); |
| 1784 | } |
| 1785 | UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); |
| 1786 | s->add(fLastSetLiteral, c); |
| 1787 | fLastSetLiteral = c; |
| 1788 | break; |
| 1789 | } |
| 1790 | |
| 1791 | |
| 1792 | case doSetNegate: |
| 1793 | // Scanned a '^' at the start of a set. |
| 1794 | // Push the negation operator onto the set op stack. |
| 1795 | // A twist for case-insensitive matching: |
| 1796 | // the case closure operation must happen _before_ negation. |
| 1797 | // But the case closure operation will already be on the stack if it's required. |
| 1798 | // This requires checking for case closure, and swapping the stack order |
| 1799 | // if it is present. |
| 1800 | { |
| 1801 | int32_t tosOp = fSetOpStack.peeki(); |
| 1802 | if (tosOp == setCaseClose) { |
| 1803 | fSetOpStack.popi(); |
| 1804 | fSetOpStack.push(setNegation, *fStatus); |
| 1805 | fSetOpStack.push(setCaseClose, *fStatus); |
| 1806 | } else { |
| 1807 | fSetOpStack.push(setNegation, *fStatus); |
| 1808 | } |
| 1809 | } |
| 1810 | break; |
| 1811 | |
| 1812 | case doSetNoCloseError: |
| 1813 | error(U_REGEX_MISSING_CLOSE_BRACKET); |
| 1814 | break; |
| 1815 | |
| 1816 | case doSetOpError: |
| 1817 | error(U_REGEX_RULE_SYNTAX); // -- or && at the end of a set. Illegal. |
| 1818 | break; |
| 1819 | |
| 1820 | case doSetPosixProp: |
| 1821 | { |
| 1822 | UnicodeSet *s = scanPosixProp(); |
| 1823 | if (s != NULL__null) { |
| 1824 | UnicodeSet *tos = (UnicodeSet *)fSetStack.peek(); |
| 1825 | tos->addAll(*s); |
| 1826 | delete s; |
| 1827 | } // else error. scanProp() reported the error status already. |
| 1828 | } |
| 1829 | break; |
| 1830 | |
| 1831 | case doSetProp: |
| 1832 | // Scanned a \p \P within [brackets]. |
| 1833 | { |
| 1834 | UnicodeSet *s = scanProp(); |
| 1835 | if (s != NULL__null) { |
| 1836 | UnicodeSet *tos = (UnicodeSet *)fSetStack.peek(); |
| 1837 | tos->addAll(*s); |
| 1838 | delete s; |
| 1839 | } // else error. scanProp() reported the error status already. |
| 1840 | } |
| 1841 | break; |
| 1842 | |
| 1843 | |
| 1844 | case doSetRange: |
| 1845 | // We have scanned literal-literal. Add the range to the set. |
| 1846 | // The left character is already in the set, and is saved in fLastSetLiteral. |
| 1847 | // The right side is the current character. |
| 1848 | // Lower Limit > Upper limit being an error matches both Java |
| 1849 | // and ICU UnicodeSet behavior. |
| 1850 | { |
| 1851 | |
| 1852 | if (fLastSetLiteral == U_SENTINEL(-1) || fLastSetLiteral > fC.fChar) { |
| 1853 | error(U_REGEX_INVALID_RANGE); |
| 1854 | } |
| 1855 | UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); |
| 1856 | s->add(fLastSetLiteral, fC.fChar); |
| 1857 | break; |
| 1858 | } |
| 1859 | |
| 1860 | default: |
| 1861 | UPRV_UNREACHABLE_EXITabort(); |
| 1862 | } |
| 1863 | |
| 1864 | if (U_FAILURE(*fStatus)) { |
| 1865 | returnVal = FALSE0; |
| 1866 | } |
| 1867 | |
| 1868 | return returnVal; |
| 1869 | } |
| 1870 | |
| 1871 | |
| 1872 | |
| 1873 | //------------------------------------------------------------------------------ |
| 1874 | // |
| 1875 | // literalChar We've encountered a literal character from the pattern, |
| 1876 | // or an escape sequence that reduces to a character. |
| 1877 | // Add it to the string containing all literal chars/strings from |
| 1878 | // the pattern. |
| 1879 | // |
| 1880 | //------------------------------------------------------------------------------ |
| 1881 | void RegexCompile::literalChar(UChar32 c) { |
| 1882 | fLiteralChars.append(c); |
| 1883 | } |
| 1884 | |
| 1885 | |
| 1886 | //------------------------------------------------------------------------------ |
| 1887 | // |
| 1888 | // fixLiterals When compiling something that can follow a literal |
| 1889 | // string in a pattern, emit the code to match the |
| 1890 | // accumulated literal string. |
| 1891 | // |
| 1892 | // Optionally, split the last char of the string off into |
| 1893 | // a single "ONE_CHAR" operation, so that quantifiers can |
| 1894 | // apply to that char alone. Example: abc* |
| 1895 | // The * must apply to the 'c' only. |
| 1896 | // |
| 1897 | //------------------------------------------------------------------------------ |
| 1898 | void RegexCompile::fixLiterals(UBool split) { |
| 1899 | |
| 1900 | // If no literal characters have been scanned but not yet had code generated |
| 1901 | // for them, nothing needs to be done. |
| 1902 | if (fLiteralChars.length() == 0) { |
| 1903 | return; |
| 1904 | } |
| 1905 | |
| 1906 | int32_t indexOfLastCodePoint = fLiteralChars.moveIndex32(fLiteralChars.length(), -1); |
| 1907 | UChar32 lastCodePoint = fLiteralChars.char32At(indexOfLastCodePoint); |
| 1908 | |
| 1909 | // Split: We need to ensure that the last item in the compiled pattern |
| 1910 | // refers only to the last literal scanned in the pattern, so that |
| 1911 | // quantifiers (*, +, etc.) affect only it, and not a longer string. |
| 1912 | // Split before case folding for case insensitive matches. |
| 1913 | |
| 1914 | if (split) { |
| 1915 | fLiteralChars.truncate(indexOfLastCodePoint); |
| 1916 | fixLiterals(FALSE0); // Recursive call, emit code to match the first part of the string. |
| 1917 | // Note that the truncated literal string may be empty, in which case |
| 1918 | // nothing will be emitted. |
| 1919 | |
| 1920 | literalChar(lastCodePoint); // Re-add the last code point as if it were a new literal. |
| 1921 | fixLiterals(FALSE0); // Second recursive call, code for the final code point. |
| 1922 | return; |
| 1923 | } |
| 1924 | |
| 1925 | // If we are doing case-insensitive matching, case fold the string. This may expand |
| 1926 | // the string, e.g. the German sharp-s turns into "ss" |
| 1927 | if (fModeFlags & UREGEX_CASE_INSENSITIVE) { |
| 1928 | fLiteralChars.foldCase(); |
| 1929 | indexOfLastCodePoint = fLiteralChars.moveIndex32(fLiteralChars.length(), -1); |
| 1930 | lastCodePoint = fLiteralChars.char32At(indexOfLastCodePoint); |
| 1931 | } |
| 1932 | |
| 1933 | if (indexOfLastCodePoint == 0) { |
| 1934 | // Single character, emit a URX_ONECHAR op to match it. |
| 1935 | if ((fModeFlags & UREGEX_CASE_INSENSITIVE) && |
| 1936 | u_hasBinaryPropertyu_hasBinaryProperty_71(lastCodePoint, UCHAR_CASE_SENSITIVE)) { |
| 1937 | appendOp(URX_ONECHAR_I, lastCodePoint); |
| 1938 | } else { |
| 1939 | appendOp(URX_ONECHAR, lastCodePoint); |
| 1940 | } |
| 1941 | } else { |
| 1942 | // Two or more chars, emit a URX_STRING to match them. |
| 1943 | if (fLiteralChars.length() > 0x00ffffff || fRXPat->fLiteralText.length() > 0x00ffffff) { |
| 1944 | error(U_REGEX_PATTERN_TOO_BIG); |
| 1945 | } |
| 1946 | if (fModeFlags & UREGEX_CASE_INSENSITIVE) { |
| 1947 | appendOp(URX_STRING_I, fRXPat->fLiteralText.length()); |
| 1948 | } else { |
| 1949 | // TODO here: add optimization to split case sensitive strings of length two |
| 1950 | // into two single char ops, for efficiency. |
| 1951 | appendOp(URX_STRING, fRXPat->fLiteralText.length()); |
| 1952 | } |
| 1953 | appendOp(URX_STRING_LEN, fLiteralChars.length()); |
| 1954 | |
| 1955 | // Add this string into the accumulated strings of the compiled pattern. |
| 1956 | fRXPat->fLiteralText.append(fLiteralChars); |
| 1957 | } |
| 1958 | |
| 1959 | fLiteralChars.remove(); |
| 1960 | } |
| 1961 | |
| 1962 | |
| 1963 | int32_t RegexCompile::buildOp(int32_t type, int32_t val) { |
| 1964 | if (U_FAILURE(*fStatus)) { |
| 1965 | return 0; |
| 1966 | } |
| 1967 | if (type < 0 || type > 255) { |
| 1968 | UPRV_UNREACHABLE_EXITabort(); |
| 1969 | } |
| 1970 | if (val > 0x00ffffff) { |
| 1971 | UPRV_UNREACHABLE_EXITabort(); |
| 1972 | } |
| 1973 | if (val < 0) { |
| 1974 | if (!(type == URX_RESERVED_OP_N || type == URX_RESERVED_OP)) { |
| 1975 | UPRV_UNREACHABLE_EXITabort(); |
| 1976 | } |
| 1977 | if (URX_TYPE(val)((uint32_t)(val) >> 24) != 0xff) { |
| 1978 | UPRV_UNREACHABLE_EXITabort(); |
| 1979 | } |
| 1980 | type = URX_RESERVED_OP_N; |
| 1981 | } |
| 1982 | return (type << 24) | val; |
| 1983 | } |
| 1984 | |
| 1985 | |
| 1986 | //------------------------------------------------------------------------------ |
| 1987 | // |
| 1988 | // appendOp() Append a new instruction onto the compiled pattern |
| 1989 | // Includes error checking, limiting the size of the |
| 1990 | // pattern to lengths that can be represented in the |
| 1991 | // 24 bit operand field of an instruction. |
| 1992 | // |
| 1993 | //------------------------------------------------------------------------------ |
| 1994 | void RegexCompile::appendOp(int32_t op) { |
| 1995 | if (U_FAILURE(*fStatus)) { |
| 1996 | return; |
| 1997 | } |
| 1998 | fRXPat->fCompiledPat->addElement(op, *fStatus); |
| 1999 | if ((fRXPat->fCompiledPat->size() > 0x00fffff0) && U_SUCCESS(*fStatus)) { |
| 2000 | error(U_REGEX_PATTERN_TOO_BIG); |
| 2001 | } |
| 2002 | } |
| 2003 | |
| 2004 | void RegexCompile::appendOp(int32_t type, int32_t val) { |
| 2005 | appendOp(buildOp(type, val)); |
| 2006 | } |
| 2007 | |
| 2008 | |
| 2009 | //------------------------------------------------------------------------------ |
| 2010 | // |
| 2011 | // insertOp() Insert a slot for a new opcode into the already |
| 2012 | // compiled pattern code. |
| 2013 | // |
| 2014 | // Fill the slot with a NOP. Our caller will replace it |
| 2015 | // with what they really wanted. |
| 2016 | // |
| 2017 | //------------------------------------------------------------------------------ |
| 2018 | void RegexCompile::insertOp(int32_t where) { |
| 2019 | UVector64 *code = fRXPat->fCompiledPat; |
| 2020 | U_ASSERT(where>0 && where < code->size())(void)0; |
| 2021 | |
| 2022 | int32_t nop = buildOp(URX_NOP, 0); |
| 2023 | code->insertElementAt(nop, where, *fStatus); |
| 2024 | |
| 2025 | // Walk through the pattern, looking for any ops with targets that |
| 2026 | // were moved down by the insert. Fix them. |
| 2027 | int32_t loc; |
| 2028 | for (loc=0; loc<code->size(); loc++) { |
| 2029 | int32_t op = (int32_t)code->elementAti(loc); |
| 2030 | int32_t opType = URX_TYPE(op)((uint32_t)(op) >> 24); |
| 2031 | int32_t opValue = URX_VAL(op)((op) & 0xffffff); |
| 2032 | if ((opType == URX_JMP || |
| 2033 | opType == URX_JMPX || |
| 2034 | opType == URX_STATE_SAVE || |
| 2035 | opType == URX_CTR_LOOP || |
| 2036 | opType == URX_CTR_LOOP_NG || |
| 2037 | opType == URX_JMP_SAV || |
| 2038 | opType == URX_JMP_SAV_X || |
| 2039 | opType == URX_RELOC_OPRND) && opValue > where) { |
| 2040 | // Target location for this opcode is after the insertion point and |
| 2041 | // needs to be incremented to adjust for the insertion. |
| 2042 | opValue++; |
| 2043 | op = buildOp(opType, opValue); |
| 2044 | code->setElementAt(op, loc); |
| 2045 | } |
| 2046 | } |
| 2047 | |
| 2048 | // Now fix up the parentheses stack. All positive values in it are locations in |
| 2049 | // the compiled pattern. (Negative values are frame boundaries, and don't need fixing.) |
| 2050 | for (loc=0; loc<fParenStack.size(); loc++) { |
| 2051 | int32_t x = fParenStack.elementAti(loc); |
| 2052 | U_ASSERT(x < code->size())(void)0; |
| 2053 | if (x>where) { |
| 2054 | x++; |
| 2055 | fParenStack.setElementAt(x, loc); |
| 2056 | } |
| 2057 | } |
| 2058 | |
| 2059 | if (fMatchCloseParen > where) { |
| 2060 | fMatchCloseParen++; |
| 2061 | } |
| 2062 | if (fMatchOpenParen > where) { |
| 2063 | fMatchOpenParen++; |
| 2064 | } |
| 2065 | } |
| 2066 | |
| 2067 | |
| 2068 | //------------------------------------------------------------------------------ |
| 2069 | // |
| 2070 | // allocateData() Allocate storage in the matcher's static data area. |
| 2071 | // Return the index for the newly allocated data. |
| 2072 | // The storage won't actually exist until we are running a match |
| 2073 | // operation, but the storage indexes are inserted into various |
| 2074 | // opcodes while compiling the pattern. |
| 2075 | // |
| 2076 | //------------------------------------------------------------------------------ |
| 2077 | int32_t RegexCompile::allocateData(int32_t size) { |
| 2078 | if (U_FAILURE(*fStatus)) { |
| 2079 | return 0; |
| 2080 | } |
| 2081 | if (size <= 0 || size > 0x100 || fRXPat->fDataSize < 0) { |
| 2082 | error(U_REGEX_INTERNAL_ERROR); |
| 2083 | return 0; |
| 2084 | } |
| 2085 | int32_t dataIndex = fRXPat->fDataSize; |
| 2086 | fRXPat->fDataSize += size; |
| 2087 | if (fRXPat->fDataSize >= 0x00fffff0) { |
| 2088 | error(U_REGEX_INTERNAL_ERROR); |
| 2089 | } |
| 2090 | return dataIndex; |
| 2091 | } |
| 2092 | |
| 2093 | |
| 2094 | //------------------------------------------------------------------------------ |
| 2095 | // |
| 2096 | // allocateStackData() Allocate space in the back-tracking stack frame. |
| 2097 | // Return the index for the newly allocated data. |
| 2098 | // The frame indexes are inserted into various |
| 2099 | // opcodes while compiling the pattern, meaning that frame |
| 2100 | // size must be restricted to the size that will fit |
| 2101 | // as an operand (24 bits). |
| 2102 | // |
| 2103 | //------------------------------------------------------------------------------ |
| 2104 | int32_t RegexCompile::allocateStackData(int32_t size) { |
| 2105 | if (U_FAILURE(*fStatus)) { |
| 2106 | return 0; |
| 2107 | } |
| 2108 | if (size <= 0 || size > 0x100 || fRXPat->fFrameSize < 0) { |
| 2109 | error(U_REGEX_INTERNAL_ERROR); |
| 2110 | return 0; |
| 2111 | } |
| 2112 | int32_t dataIndex = fRXPat->fFrameSize; |
| 2113 | fRXPat->fFrameSize += size; |
| 2114 | if (fRXPat->fFrameSize >= 0x00fffff0) { |
| 2115 | error(U_REGEX_PATTERN_TOO_BIG); |
| 2116 | } |
| 2117 | return dataIndex; |
| 2118 | } |
| 2119 | |
| 2120 | |
| 2121 | //------------------------------------------------------------------------------ |
| 2122 | // |
| 2123 | // blockTopLoc() Find or create a location in the compiled pattern |
| 2124 | // at the start of the operation or block that has |
| 2125 | // just been compiled. Needed when a quantifier (* or |
| 2126 | // whatever) appears, and we need to add an operation |
| 2127 | // at the start of the thing being quantified. |
| 2128 | // |
| 2129 | // (Parenthesized Blocks) have a slot with a NOP that |
| 2130 | // is reserved for this purpose. .* or similar don't |
| 2131 | // and a slot needs to be added. |
| 2132 | // |
| 2133 | // parameter reserveLoc : TRUE - ensure that there is space to add an opcode |
| 2134 | // at the returned location. |
| 2135 | // FALSE - just return the address, |
| 2136 | // do not reserve a location there. |
| 2137 | // |
| 2138 | //------------------------------------------------------------------------------ |
| 2139 | int32_t RegexCompile::blockTopLoc(UBool reserveLoc) { |
| 2140 | int32_t theLoc; |
| 2141 | fixLiterals(TRUE1); // Emit code for any pending literals. |
| 2142 | // If last item was a string, emit separate op for the its last char. |
| 2143 | if (fRXPat->fCompiledPat->size() == fMatchCloseParen) |
| 2144 | { |
| 2145 | // The item just processed is a parenthesized block. |
| 2146 | theLoc = fMatchOpenParen; // A slot is already reserved for us. |
| 2147 | U_ASSERT(theLoc > 0)(void)0; |
| 2148 | U_ASSERT(URX_TYPE(((uint32_t)fRXPat->fCompiledPat->elementAti(theLoc))) == URX_NOP)(void)0; |
| 2149 | } |
| 2150 | else { |
| 2151 | // Item just compiled is a single thing, a ".", or a single char, a string or a set reference. |
| 2152 | // No slot for STATE_SAVE was pre-reserved in the compiled code. |
| 2153 | // We need to make space now. |
| 2154 | theLoc = fRXPat->fCompiledPat->size()-1; |
| 2155 | int32_t opAtTheLoc = (int32_t)fRXPat->fCompiledPat->elementAti(theLoc); |
| 2156 | if (URX_TYPE(opAtTheLoc)((uint32_t)(opAtTheLoc) >> 24) == URX_STRING_LEN) { |
| 2157 | // Strings take two opcode, we want the position of the first one. |
| 2158 | // We can have a string at this point if a single character case-folded to two. |
| 2159 | theLoc--; |
| 2160 | } |
| 2161 | if (reserveLoc) { |
| 2162 | int32_t nop = buildOp(URX_NOP, 0); |
| 2163 | fRXPat->fCompiledPat->insertElementAt(nop, theLoc, *fStatus); |
| 2164 | } |
| 2165 | } |
| 2166 | return theLoc; |
| 2167 | } |
| 2168 | |
| 2169 | |
| 2170 | |
| 2171 | //------------------------------------------------------------------------------ |
| 2172 | // |
| 2173 | // handleCloseParen When compiling a close paren, we need to go back |
| 2174 | // and fix up any JMP or SAVE operations within the |
| 2175 | // parenthesized block that need to target the end |
| 2176 | // of the block. The locations of these are kept on |
| 2177 | // the paretheses stack. |
| 2178 | // |
| 2179 | // This function is called both when encountering a |
| 2180 | // real ) and at the end of the pattern. |
| 2181 | // |
| 2182 | //------------------------------------------------------------------------------ |
| 2183 | void RegexCompile::handleCloseParen() { |
| 2184 | int32_t patIdx; |
| 2185 | int32_t patOp; |
| 2186 | if (fParenStack.size() <= 0) { |
| 2187 | error(U_REGEX_MISMATCHED_PAREN); |
| 2188 | return; |
| 2189 | } |
| 2190 | |
| 2191 | // Emit code for any pending literals. |
| 2192 | fixLiterals(FALSE0); |
| 2193 | |
| 2194 | // Fixup any operations within the just-closed parenthesized group |
| 2195 | // that need to reference the end of the (block). |
| 2196 | // (The first one popped from the stack is an unused slot for |
| 2197 | // alternation (OR) state save, but applying the fixup to it does no harm.) |
| 2198 | for (;;) { |
| 2199 | patIdx = fParenStack.popi(); |
| 2200 | if (patIdx < 0) { |
| 2201 | // value < 0 flags the start of the frame on the paren stack. |
| 2202 | break; |
| 2203 | } |
| 2204 | U_ASSERT(patIdx>0 && patIdx <= fRXPat->fCompiledPat->size())(void)0; |
| 2205 | patOp = (int32_t)fRXPat->fCompiledPat->elementAti(patIdx); |
| 2206 | U_ASSERT(URX_VAL(patOp) == 0)(void)0; // Branch target for JMP should not be set. |
| 2207 | patOp |= fRXPat->fCompiledPat->size(); // Set it now. |
| 2208 | fRXPat->fCompiledPat->setElementAt(patOp, patIdx); |
| 2209 | fMatchOpenParen = patIdx; |
| 2210 | } |
| 2211 | |
| 2212 | // At the close of any parenthesized block, restore the match mode flags to |
| 2213 | // the value they had at the open paren. Saved value is |
| 2214 | // at the top of the paren stack. |
| 2215 | fModeFlags = fParenStack.popi(); |
| 2216 | U_ASSERT(fModeFlags < 0)(void)0; |
| 2217 | |
| 2218 | // DO any additional fixups, depending on the specific kind of |
| 2219 | // parentesized grouping this is |
| 2220 | |
| 2221 | switch (patIdx) { |
| 2222 | case plain: |
| 2223 | case flags: |
| 2224 | // No additional fixups required. |
| 2225 | // (Grouping-only parentheses) |
| 2226 | break; |
| 2227 | case capturing: |
| 2228 | // Capturing Parentheses. |
| 2229 | // Insert a End Capture op into the pattern. |
| 2230 | // The frame offset of the variables for this cg is obtained from the |
| 2231 | // start capture op and put it into the end-capture op. |
| 2232 | { |
| 2233 | int32_t captureOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1); |
| 2234 | U_ASSERT(URX_TYPE(captureOp) == URX_START_CAPTURE)(void)0; |
| 2235 | |
| 2236 | int32_t frameVarLocation = URX_VAL(captureOp)((captureOp) & 0xffffff); |
| 2237 | appendOp(URX_END_CAPTURE, frameVarLocation); |
| 2238 | } |
| 2239 | break; |
| 2240 | case atomic: |
| 2241 | // Atomic Parenthesis. |
| 2242 | // Insert a LD_SP operation to restore the state stack to the position |
| 2243 | // it was when the atomic parens were entered. |
| 2244 | { |
| 2245 | int32_t stoOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1); |
| 2246 | U_ASSERT(URX_TYPE(stoOp) == URX_STO_SP)(void)0; |
| 2247 | int32_t stoLoc = URX_VAL(stoOp)((stoOp) & 0xffffff); |
| 2248 | appendOp(URX_LD_SP, stoLoc); |
| 2249 | } |
| 2250 | break; |
| 2251 | |
| 2252 | case lookAhead: |
| 2253 | { |
| 2254 | int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5); |
| 2255 | U_ASSERT(URX_TYPE(startOp) == URX_LA_START)(void)0; |
| 2256 | int32_t dataLoc = URX_VAL(startOp)((startOp) & 0xffffff); |
| 2257 | appendOp(URX_LA_END, dataLoc); |
| 2258 | } |
| 2259 | break; |
| 2260 | |
| 2261 | case negLookAhead: |
| 2262 | { |
| 2263 | // See comment at doOpenLookAheadNeg |
| 2264 | int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-1); |
| 2265 | U_ASSERT(URX_TYPE(startOp) == URX_LA_START)(void)0; |
| 2266 | int32_t dataLoc = URX_VAL(startOp)((startOp) & 0xffffff); |
| 2267 | appendOp(URX_LA_END, dataLoc); |
| 2268 | appendOp(URX_BACKTRACK, 0); |
| 2269 | appendOp(URX_LA_END, dataLoc); |
| 2270 | |
| 2271 | // Patch the URX_SAVE near the top of the block. |
| 2272 | // The destination of the SAVE is the final LA_END that was just added. |
| 2273 | int32_t saveOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen); |
| 2274 | U_ASSERT(URX_TYPE(saveOp) == URX_STATE_SAVE)(void)0; |
| 2275 | int32_t dest = fRXPat->fCompiledPat->size()-1; |
| 2276 | saveOp = buildOp(URX_STATE_SAVE, dest); |
| 2277 | fRXPat->fCompiledPat->setElementAt(saveOp, fMatchOpenParen); |
| 2278 | } |
| 2279 | break; |
| 2280 | |
| 2281 | case lookBehind: |
| 2282 | { |
| 2283 | // See comment at doOpenLookBehind. |
| 2284 | |
| 2285 | // Append the URX_LB_END and URX_LA_END to the compiled pattern. |
| 2286 | int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-4); |
| 2287 | U_ASSERT(URX_TYPE(startOp) == URX_LB_START)(void)0; |
| 2288 | int32_t dataLoc = URX_VAL(startOp)((startOp) & 0xffffff); |
| 2289 | appendOp(URX_LB_END, dataLoc); |
| 2290 | appendOp(URX_LA_END, dataLoc); |
| 2291 | |
| 2292 | // Determine the min and max bounds for the length of the |
| 2293 | // string that the pattern can match. |
| 2294 | // An unbounded upper limit is an error. |
| 2295 | int32_t patEnd = fRXPat->fCompiledPat->size() - 1; |
| 2296 | int32_t minML = minMatchLength(fMatchOpenParen, patEnd); |
| 2297 | int32_t maxML = maxMatchLength(fMatchOpenParen, patEnd); |
| 2298 | if (URX_TYPE(maxML)((uint32_t)(maxML) >> 24) != 0) { |
| 2299 | error(U_REGEX_LOOK_BEHIND_LIMIT); |
| 2300 | break; |
| 2301 | } |
| 2302 | if (maxML == INT32_MAX(2147483647)) { |
| 2303 | error(U_REGEX_LOOK_BEHIND_LIMIT); |
| 2304 | break; |
| 2305 | } |
| 2306 | if (minML == INT32_MAX(2147483647)) { |
| 2307 | // This condition happens when no match is possible, such as with a |
| 2308 | // [set] expression containing no elements. |
| 2309 | // In principle, the generated code to evaluate the expression could be deleted, |
| 2310 | // but it's probably not worth the complication. |
| 2311 | minML = 0; |
| 2312 | } |
| 2313 | U_ASSERT(minML <= maxML)(void)0; |
| 2314 | |
| 2315 | // Insert the min and max match len bounds into the URX_LB_CONT op that |
| 2316 | // appears at the top of the look-behind block, at location fMatchOpenParen+1 |
| 2317 | fRXPat->fCompiledPat->setElementAt(minML, fMatchOpenParen-2); |
| 2318 | fRXPat->fCompiledPat->setElementAt(maxML, fMatchOpenParen-1); |
| 2319 | |
| 2320 | } |
| 2321 | break; |
| 2322 | |
| 2323 | |
| 2324 | |
| 2325 | case lookBehindN: |
| 2326 | { |
| 2327 | // See comment at doOpenLookBehindNeg. |
| 2328 | |
| 2329 | // Append the URX_LBN_END to the compiled pattern. |
| 2330 | int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5); |
| 2331 | U_ASSERT(URX_TYPE(startOp) == URX_LB_START)(void)0; |
| 2332 | int32_t dataLoc = URX_VAL(startOp)((startOp) & 0xffffff); |
| 2333 | appendOp(URX_LBN_END, dataLoc); |
| 2334 | |
| 2335 | // Determine the min and max bounds for the length of the |
| 2336 | // string that the pattern can match. |
| 2337 | // An unbounded upper limit is an error. |
| 2338 | int32_t patEnd = fRXPat->fCompiledPat->size() - 1; |
| 2339 | int32_t minML = minMatchLength(fMatchOpenParen, patEnd); |
| 2340 | int32_t maxML = maxMatchLength(fMatchOpenParen, patEnd); |
| 2341 | if (URX_TYPE(maxML)((uint32_t)(maxML) >> 24) != 0) { |
| 2342 | error(U_REGEX_LOOK_BEHIND_LIMIT); |
| 2343 | break; |
| 2344 | } |
| 2345 | if (maxML == INT32_MAX(2147483647)) { |
| 2346 | error(U_REGEX_LOOK_BEHIND_LIMIT); |
| 2347 | break; |
| 2348 | } |
| 2349 | if (minML == INT32_MAX(2147483647)) { |
| 2350 | // This condition happens when no match is possible, such as with a |
| 2351 | // [set] expression containing no elements. |
| 2352 | // In principle, the generated code to evaluate the expression could be deleted, |
| 2353 | // but it's probably not worth the complication. |
| 2354 | minML = 0; |
| 2355 | } |
| 2356 | |
| 2357 | U_ASSERT(minML <= maxML)(void)0; |
| 2358 | |
| 2359 | // Insert the min and max match len bounds into the URX_LB_CONT op that |
| 2360 | // appears at the top of the look-behind block, at location fMatchOpenParen+1 |
| 2361 | fRXPat->fCompiledPat->setElementAt(minML, fMatchOpenParen-3); |
| 2362 | fRXPat->fCompiledPat->setElementAt(maxML, fMatchOpenParen-2); |
| 2363 | |
| 2364 | // Insert the pattern location to continue at after a successful match |
| 2365 | // as the last operand of the URX_LBN_CONT |
| 2366 | int32_t op = buildOp(URX_RELOC_OPRND, fRXPat->fCompiledPat->size()); |
| 2367 | fRXPat->fCompiledPat->setElementAt(op, fMatchOpenParen-1); |
| 2368 | } |
| 2369 | break; |
| 2370 | |
| 2371 | |
| 2372 | |
| 2373 | default: |
| 2374 | UPRV_UNREACHABLE_EXITabort(); |
| 2375 | } |
| 2376 | |
| 2377 | // remember the next location in the compiled pattern. |
| 2378 | // The compilation of Quantifiers will look at this to see whether its looping |
| 2379 | // over a parenthesized block or a single item |
| 2380 | fMatchCloseParen = fRXPat->fCompiledPat->size(); |
| 2381 | } |
| 2382 | |
| 2383 | |
| 2384 | |
| 2385 | //------------------------------------------------------------------------------ |
| 2386 | // |
| 2387 | // compileSet Compile the pattern operations for a reference to a |
| 2388 | // UnicodeSet. |
| 2389 | // |
| 2390 | //------------------------------------------------------------------------------ |
| 2391 | void RegexCompile::compileSet(UnicodeSet *theSet) |
| 2392 | { |
| 2393 | if (theSet == NULL__null) { |
| 2394 | return; |
| 2395 | } |
| 2396 | // Remove any strings from the set. |
| 2397 | // There shouldn't be any, but just in case. |
| 2398 | // (Case Closure can add them; if we had a simple case closure available that |
| 2399 | // ignored strings, that would be better.) |
| 2400 | theSet->removeAllStrings(); |
| 2401 | int32_t setSize = theSet->size(); |
| 2402 | |
| 2403 | switch (setSize) { |
| 2404 | case 0: |
| 2405 | { |
| 2406 | // Set of no elements. Always fails to match. |
| 2407 | appendOp(URX_BACKTRACK, 0); |
| 2408 | delete theSet; |
| 2409 | } |
| 2410 | break; |
| 2411 | |
| 2412 | case 1: |
| 2413 | { |
| 2414 | // The set contains only a single code point. Put it into |
| 2415 | // the compiled pattern as a single char operation rather |
| 2416 | // than a set, and discard the set itself. |
| 2417 | literalChar(theSet->charAt(0)); |
| 2418 | delete theSet; |
| 2419 | } |
| 2420 | break; |
| 2421 | |
| 2422 | default: |
| 2423 | { |
| 2424 | // The set contains two or more chars. (the normal case) |
| 2425 | // Put it into the compiled pattern as a set. |
| 2426 | theSet->freeze(); |
| 2427 | int32_t setNumber = fRXPat->fSets->size(); |
| 2428 | fRXPat->fSets->addElement(theSet, *fStatus); |
| 2429 | if (U_SUCCESS(*fStatus)) { |
| 2430 | appendOp(URX_SETREF, setNumber); |
| 2431 | } else { |
| 2432 | delete theSet; |
| 2433 | } |
| 2434 | } |
| 2435 | } |
| 2436 | } |
| 2437 | |
| 2438 | |
| 2439 | //------------------------------------------------------------------------------ |
| 2440 | // |
| 2441 | // compileInterval Generate the code for a {min, max} style interval quantifier. |
| 2442 | // Except for the specific opcodes used, the code is the same |
| 2443 | // for all three types (greedy, non-greedy, possessive) of |
| 2444 | // intervals. The opcodes are supplied as parameters. |
| 2445 | // (There are two sets of opcodes - greedy & possessive use the |
| 2446 | // same ones, while non-greedy has it's own.) |
| 2447 | // |
| 2448 | // The code for interval loops has this form: |
| 2449 | // 0 CTR_INIT counter loc (in stack frame) |
| 2450 | // 1 5 patt address of CTR_LOOP at bottom of block |
| 2451 | // 2 min count |
| 2452 | // 3 max count (-1 for unbounded) |
| 2453 | // 4 ... block to be iterated over |
| 2454 | // 5 CTR_LOOP |
| 2455 | // |
| 2456 | // In |
| 2457 | //------------------------------------------------------------------------------ |
| 2458 | void RegexCompile::compileInterval(int32_t InitOp, int32_t LoopOp) |
| 2459 | { |
| 2460 | // The CTR_INIT op at the top of the block with the {n,m} quantifier takes |
| 2461 | // four slots in the compiled code. Reserve them. |
| 2462 | int32_t topOfBlock = blockTopLoc(TRUE1); |
| 2463 | insertOp(topOfBlock); |
| 2464 | insertOp(topOfBlock); |
| 2465 | insertOp(topOfBlock); |
| 2466 | |
| 2467 | // The operands for the CTR_INIT opcode include the index in the matcher data |
| 2468 | // of the counter. Allocate it now. There are two data items |
| 2469 | // counterLoc --> Loop counter |
| 2470 | // +1 --> Input index (for breaking non-progressing loops) |
| 2471 | // (Only present if unbounded upper limit on loop) |
| 2472 | int32_t dataSize = fIntervalUpper < 0 ? 2 : 1; |
| 2473 | int32_t counterLoc = allocateStackData(dataSize); |
| 2474 | |
| 2475 | int32_t op = buildOp(InitOp, counterLoc); |
| 2476 | fRXPat->fCompiledPat->setElementAt(op, topOfBlock); |
| 2477 | |
| 2478 | // The second operand of CTR_INIT is the location following the end of the loop. |
| 2479 | // Must put in as a URX_RELOC_OPRND so that the value will be adjusted if the |
| 2480 | // compilation of something later on causes the code to grow and the target |
| 2481 | // position to move. |
| 2482 | int32_t loopEnd = fRXPat->fCompiledPat->size(); |
| 2483 | op = buildOp(URX_RELOC_OPRND, loopEnd); |
| 2484 | fRXPat->fCompiledPat->setElementAt(op, topOfBlock+1); |
| 2485 | |
| 2486 | // Followed by the min and max counts. |
| 2487 | fRXPat->fCompiledPat->setElementAt(fIntervalLow, topOfBlock+2); |
| 2488 | fRXPat->fCompiledPat->setElementAt(fIntervalUpper, topOfBlock+3); |
| 2489 | |
| 2490 | // Append the CTR_LOOP op. The operand is the location of the CTR_INIT op. |
| 2491 | // Goes at end of the block being looped over, so just append to the code so far. |
| 2492 | appendOp(LoopOp, topOfBlock); |
| 2493 | |
| 2494 | if ((fIntervalLow & 0xff000000) != 0 || |
| 2495 | (fIntervalUpper > 0 && (fIntervalUpper & 0xff000000) != 0)) { |
| 2496 | error(U_REGEX_NUMBER_TOO_BIG); |
| 2497 | } |
| 2498 | |
| 2499 | if (fIntervalLow > fIntervalUpper && fIntervalUpper != -1) { |
| 2500 | error(U_REGEX_MAX_LT_MIN); |
| 2501 | } |
| 2502 | } |
| 2503 | |
| 2504 | |
| 2505 | |
| 2506 | UBool RegexCompile::compileInlineInterval() { |
| 2507 | if (fIntervalUpper > 10 || fIntervalUpper < fIntervalLow) { |
| 2508 | // Too big to inline. Fail, which will cause looping code to be generated. |
| 2509 | // (Upper < Lower picks up unbounded upper and errors, both.) |
| 2510 | return FALSE0; |
| 2511 | } |
| 2512 | |
| 2513 | int32_t topOfBlock = blockTopLoc(FALSE0); |
| 2514 | if (fIntervalUpper == 0) { |
| 2515 | // Pathological case. Attempt no matches, as if the block doesn't exist. |
| 2516 | // Discard the generated code for the block. |
| 2517 | // If the block included parens, discard the info pertaining to them as well. |
| 2518 | fRXPat->fCompiledPat->setSize(topOfBlock); |
| 2519 | if (fMatchOpenParen >= topOfBlock) { |
| 2520 | fMatchOpenParen = -1; |
| 2521 | } |
| 2522 | if (fMatchCloseParen >= topOfBlock) { |
| 2523 | fMatchCloseParen = -1; |
| 2524 | } |
| 2525 | return TRUE1; |
| 2526 | } |
| 2527 | |
| 2528 | if (topOfBlock != fRXPat->fCompiledPat->size()-1 && fIntervalUpper != 1) { |
| 2529 | // The thing being repeated is not a single op, but some |
| 2530 | // more complex block. Do it as a loop, not inlines. |
| 2531 | // Note that things "repeated" a max of once are handled as inline, because |
| 2532 | // the one copy of the code already generated is just fine. |
| 2533 | return FALSE0; |
| 2534 | } |
| 2535 | |
| 2536 | // Pick up the opcode that is to be repeated |
| 2537 | // |
| 2538 | int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(topOfBlock); |
| 2539 | |
| 2540 | // Compute the pattern location where the inline sequence |
| 2541 | // will end, and set up the state save op that will be needed. |
| 2542 | // |
| 2543 | int32_t endOfSequenceLoc = fRXPat->fCompiledPat->size()-1 |
| 2544 | + fIntervalUpper + (fIntervalUpper-fIntervalLow); |
| 2545 | int32_t saveOp = buildOp(URX_STATE_SAVE, endOfSequenceLoc); |
| 2546 | if (fIntervalLow == 0) { |
| 2547 | insertOp(topOfBlock); |
| 2548 | fRXPat->fCompiledPat->setElementAt(saveOp, topOfBlock); |
| 2549 | } |
| 2550 | |
| 2551 | |
| 2552 | |
| 2553 | // Loop, emitting the op for the thing being repeated each time. |
| 2554 | // Loop starts at 1 because one instance of the op already exists in the pattern, |
| 2555 | // it was put there when it was originally encountered. |
| 2556 | int32_t i; |
| 2557 | for (i=1; i<fIntervalUpper; i++ ) { |
| 2558 | if (i >= fIntervalLow) { |
| 2559 | appendOp(saveOp); |
| 2560 | } |
| 2561 | appendOp(op); |
| 2562 | } |
| 2563 | return TRUE1; |
| 2564 | } |
| 2565 | |
| 2566 | |
| 2567 | |
| 2568 | //------------------------------------------------------------------------------ |
| 2569 | // |
| 2570 | // caseInsensitiveStart given a single code point from a pattern string, determine the |
| 2571 | // set of characters that could potentially begin a case-insensitive |
| 2572 | // match of a string beginning with that character, using full Unicode |
| 2573 | // case insensitive matching. |
| 2574 | // |
| 2575 | // This is used in optimizing find(). |
| 2576 | // |
| 2577 | // closeOver(USET_CASE_INSENSITIVE) does most of what is needed, but |
| 2578 | // misses cases like this: |
| 2579 | // A string from the pattern begins with 'ss' (although all we know |
| 2580 | // in this context is that it begins with 's') |
| 2581 | // The pattern could match a string beginning with a German sharp-s |
| 2582 | // |
| 2583 | // To the ordinary case closure for a character c, we add all other |
| 2584 | // characters cx where the case closure of cx includes a string form that begins |
| 2585 | // with the original character c. |
| 2586 | // |
| 2587 | // This function could be made smarter. The full pattern string is available |
| 2588 | // and it would be possible to verify that the extra characters being added |
| 2589 | // to the starting set fully match, rather than having just a first-char of the |
| 2590 | // folded form match. |
| 2591 | // |
| 2592 | //------------------------------------------------------------------------------ |
| 2593 | void RegexCompile::findCaseInsensitiveStarters(UChar32 c, UnicodeSet *starterChars) { |
| 2594 | |
| 2595 | // Machine Generated below. |
| 2596 | // It may need updating with new versions of Unicode. |
| 2597 | // Intltest test RegexTest::TestCaseInsensitiveStarters will fail if an update is needed. |
| 2598 | // The update tool is here: |
| 2599 | // https://github.com/unicode-org/icu/tree/main/tools/unicode/c/genregexcasing |
| 2600 | |
| 2601 | // Machine Generated Data. Do not hand edit. |
| 2602 | static const UChar32 RECaseFixCodePoints[] = { |
| 2603 | 0x61, 0x66, 0x68, 0x69, 0x6a, 0x73, 0x74, 0x77, 0x79, 0x2bc, |
| 2604 | 0x3ac, 0x3ae, 0x3b1, 0x3b7, 0x3b9, 0x3c1, 0x3c5, 0x3c9, 0x3ce, 0x565, |
| 2605 | 0x574, 0x57e, 0x1f00, 0x1f01, 0x1f02, 0x1f03, 0x1f04, 0x1f05, 0x1f06, 0x1f07, |
| 2606 | 0x1f20, 0x1f21, 0x1f22, 0x1f23, 0x1f24, 0x1f25, 0x1f26, 0x1f27, 0x1f60, 0x1f61, |
| 2607 | 0x1f62, 0x1f63, 0x1f64, 0x1f65, 0x1f66, 0x1f67, 0x1f70, 0x1f74, 0x1f7c, 0x110000}; |
| 2608 | |
| 2609 | static const int16_t RECaseFixStringOffsets[] = { |
| 2610 | 0x0, 0x1, 0x6, 0x7, 0x8, 0x9, 0xd, 0xe, 0xf, 0x10, |
| 2611 | 0x11, 0x12, 0x13, 0x17, 0x1b, 0x20, 0x21, 0x2a, 0x2e, 0x2f, |
| 2612 | 0x30, 0x34, 0x35, 0x37, 0x39, 0x3b, 0x3d, 0x3f, 0x41, 0x43, |
| 2613 | 0x45, 0x47, 0x49, 0x4b, 0x4d, 0x4f, 0x51, 0x53, 0x55, 0x57, |
| 2614 | 0x59, 0x5b, 0x5d, 0x5f, 0x61, 0x63, 0x65, 0x66, 0x67, 0}; |
| 2615 | |
| 2616 | static const int16_t RECaseFixCounts[] = { |
| 2617 | 0x1, 0x5, 0x1, 0x1, 0x1, 0x4, 0x1, 0x1, 0x1, 0x1, |
| 2618 | 0x1, 0x1, 0x4, 0x4, 0x5, 0x1, 0x9, 0x4, 0x1, 0x1, |
| 2619 | 0x4, 0x1, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, |
| 2620 | 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, |
| 2621 | 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x1, 0x1, 0x1, 0}; |
| 2622 | |
| 2623 | static const UChar RECaseFixData[] = { |
| 2624 | 0x1e9a, 0xfb00, 0xfb01, 0xfb02, 0xfb03, 0xfb04, 0x1e96, 0x130, 0x1f0, 0xdf, |
| 2625 | 0x1e9e, 0xfb05, 0xfb06, 0x1e97, 0x1e98, 0x1e99, 0x149, 0x1fb4, 0x1fc4, 0x1fb3, |
| 2626 | 0x1fb6, 0x1fb7, 0x1fbc, 0x1fc3, 0x1fc6, 0x1fc7, 0x1fcc, 0x390, 0x1fd2, 0x1fd3, |
| 2627 | 0x1fd6, 0x1fd7, 0x1fe4, 0x3b0, 0x1f50, 0x1f52, 0x1f54, 0x1f56, 0x1fe2, 0x1fe3, |
| 2628 | 0x1fe6, 0x1fe7, 0x1ff3, 0x1ff6, 0x1ff7, 0x1ffc, 0x1ff4, 0x587, 0xfb13, 0xfb14, |
| 2629 | 0xfb15, 0xfb17, 0xfb16, 0x1f80, 0x1f88, 0x1f81, 0x1f89, 0x1f82, 0x1f8a, 0x1f83, |
| 2630 | 0x1f8b, 0x1f84, 0x1f8c, 0x1f85, 0x1f8d, 0x1f86, 0x1f8e, 0x1f87, 0x1f8f, 0x1f90, |
| 2631 | 0x1f98, 0x1f91, 0x1f99, 0x1f92, 0x1f9a, 0x1f93, 0x1f9b, 0x1f94, 0x1f9c, 0x1f95, |
| 2632 | 0x1f9d, 0x1f96, 0x1f9e, 0x1f97, 0x1f9f, 0x1fa0, 0x1fa8, 0x1fa1, 0x1fa9, 0x1fa2, |
| 2633 | 0x1faa, 0x1fa3, 0x1fab, 0x1fa4, 0x1fac, 0x1fa5, 0x1fad, 0x1fa6, 0x1fae, 0x1fa7, |
| 2634 | 0x1faf, 0x1fb2, 0x1fc2, 0x1ff2, 0}; |
| 2635 | |
| 2636 | // End of machine generated data. |
| 2637 | |
| 2638 | if (c < UCHAR_MIN_VALUE0 || c > UCHAR_MAX_VALUE0x10ffff) { |
| 2639 | // This function should never be called with an invalid input character. |
| 2640 | UPRV_UNREACHABLE_EXITabort(); |
| 2641 | } else if (u_hasBinaryPropertyu_hasBinaryProperty_71(c, UCHAR_CASE_SENSITIVE)) { |
| 2642 | UChar32 caseFoldedC = u_foldCaseu_foldCase_71(c, U_FOLD_CASE_DEFAULT0); |
| 2643 | starterChars->set(caseFoldedC, caseFoldedC); |
| 2644 | |
| 2645 | int32_t i; |
| 2646 | for (i=0; RECaseFixCodePoints[i]<c ; i++) { |
| 2647 | // Simple linear search through the sorted list of interesting code points. |
| 2648 | } |
| 2649 | |
| 2650 | if (RECaseFixCodePoints[i] == c) { |
| 2651 | int32_t dataIndex = RECaseFixStringOffsets[i]; |
| 2652 | int32_t numCharsToAdd = RECaseFixCounts[i]; |
| 2653 | UChar32 cpToAdd = 0; |
| 2654 | for (int32_t j=0; j<numCharsToAdd; j++) { |
| 2655 | U16_NEXT_UNSAFE(RECaseFixData, dataIndex, cpToAdd)do { (cpToAdd)=(RECaseFixData)[(dataIndex)++]; if((((cpToAdd) &0xfffffc00)==0xd800)) { (cpToAdd)=(((UChar32)((cpToAdd)) <<10UL)+(UChar32)((RECaseFixData)[(dataIndex)++])-((0xd800 <<10UL)+0xdc00-0x10000)); } } while (false); |
| 2656 | starterChars->add(cpToAdd); |
| 2657 | } |
| 2658 | } |
| 2659 | |
| 2660 | starterChars->closeOver(USET_CASE_INSENSITIVE); |
| 2661 | starterChars->removeAllStrings(); |
| 2662 | } else { |
| 2663 | // Not a cased character. Just return it alone. |
| 2664 | starterChars->set(c, c); |
| 2665 | } |
| 2666 | } |
| 2667 | |
| 2668 | |
| 2669 | // Increment with overflow check. |
| 2670 | // val and delta will both be positive. |
| 2671 | |
| 2672 | static int32_t safeIncrement(int32_t val, int32_t delta) { |
| 2673 | if (INT32_MAX(2147483647) - val > delta) { |
| 2674 | return val + delta; |
| 2675 | } else { |
| 2676 | return INT32_MAX(2147483647); |
| 2677 | } |
| 2678 | } |
| 2679 | |
| 2680 | |
| 2681 | //------------------------------------------------------------------------------ |
| 2682 | // |
| 2683 | // matchStartType Determine how a match can start. |
| 2684 | // Used to optimize find() operations. |
| 2685 | // |
| 2686 | // Operation is very similar to minMatchLength(). Walk the compiled |
| 2687 | // pattern, keeping an on-going minimum-match-length. For any |
| 2688 | // op where the min match coming in is zero, add that ops possible |
| 2689 | // starting matches to the possible starts for the overall pattern. |
| 2690 | // |
| 2691 | //------------------------------------------------------------------------------ |
| 2692 | void RegexCompile::matchStartType() { |
| 2693 | if (U_FAILURE(*fStatus)) { |
| 2694 | return; |
| 2695 | } |
| 2696 | |
| 2697 | |
| 2698 | int32_t loc; // Location in the pattern of the current op being processed. |
| 2699 | int32_t op; // The op being processed |
| 2700 | int32_t opType; // The opcode type of the op |
| 2701 | int32_t currentLen = 0; // Minimum length of a match to this point (loc) in the pattern |
| 2702 | int32_t numInitialStrings = 0; // Number of strings encountered that could match at start. |
| 2703 | |
| 2704 | UBool atStart = TRUE1; // True if no part of the pattern yet encountered |
| 2705 | // could have advanced the position in a match. |
| 2706 | // (Maximum match length so far == 0) |
| 2707 | |
| 2708 | // forwardedLength is a vector holding minimum-match-length values that |
| 2709 | // are propagated forward in the pattern by JMP or STATE_SAVE operations. |
| 2710 | // It must be one longer than the pattern being checked because some ops |
| 2711 | // will jmp to a end-of-block+1 location from within a block, and we must |
| 2712 | // count those when checking the block. |
| 2713 | int32_t end = fRXPat->fCompiledPat->size(); |
| 2714 | UVector32 forwardedLength(end+1, *fStatus); |
| 2715 | forwardedLength.setSize(end+1); |
| 2716 | for (loc=3; loc<end; loc++) { |
| 2717 | forwardedLength.setElementAt(INT32_MAX(2147483647), loc); |
| 2718 | } |
| 2719 | |
| 2720 | for (loc = 3; loc<end; loc++) { |
| 2721 | op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 2722 | opType = URX_TYPE(op)((uint32_t)(op) >> 24); |
| 2723 | |
| 2724 | // The loop is advancing linearly through the pattern. |
| 2725 | // If the op we are now at was the destination of a branch in the pattern, |
| 2726 | // and that path has a shorter minimum length than the current accumulated value, |
| 2727 | // replace the current accumulated value. |
| 2728 | if (forwardedLength.elementAti(loc) < currentLen) { |
| 2729 | currentLen = forwardedLength.elementAti(loc); |
| 2730 | U_ASSERT(currentLen>=0 && currentLen < INT32_MAX)(void)0; |
| 2731 | } |
| 2732 | |
| 2733 | switch (opType) { |
| 2734 | // Ops that don't change the total length matched |
| 2735 | case URX_RESERVED_OP: |
| 2736 | case URX_END: |
| 2737 | case URX_FAIL: |
| 2738 | case URX_STRING_LEN: |
| 2739 | case URX_NOP: |
| 2740 | case URX_START_CAPTURE: |
| 2741 | case URX_END_CAPTURE: |
| 2742 | case URX_BACKSLASH_B: |
| 2743 | case URX_BACKSLASH_BU: |
| 2744 | case URX_BACKSLASH_G: |
| 2745 | case URX_BACKSLASH_Z: |
| 2746 | case URX_DOLLAR: |
| 2747 | case URX_DOLLAR_M: |
| 2748 | case URX_DOLLAR_D: |
| 2749 | case URX_DOLLAR_MD: |
| 2750 | case URX_RELOC_OPRND: |
| 2751 | case URX_STO_INP_LOC: |
| 2752 | case URX_BACKREF: // BackRef. Must assume that it might be a zero length match |
| 2753 | case URX_BACKREF_I: |
| 2754 | |
| 2755 | case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. |
| 2756 | case URX_LD_SP: |
| 2757 | break; |
| 2758 | |
| 2759 | case URX_CARET: |
| 2760 | if (atStart) { |
| 2761 | fRXPat->fStartType = START_START; |
| 2762 | } |
| 2763 | break; |
| 2764 | |
| 2765 | case URX_CARET_M: |
| 2766 | case URX_CARET_M_UNIX: |
| 2767 | if (atStart) { |
| 2768 | fRXPat->fStartType = START_LINE; |
| 2769 | } |
| 2770 | break; |
| 2771 | |
| 2772 | case URX_ONECHAR: |
| 2773 | if (currentLen == 0) { |
| 2774 | // This character could appear at the start of a match. |
| 2775 | // Add it to the set of possible starting characters. |
| 2776 | fRXPat->fInitialChars->add(URX_VAL(op)((op) & 0xffffff)); |
| 2777 | numInitialStrings += 2; |
| 2778 | } |
| 2779 | currentLen = safeIncrement(currentLen, 1); |
| 2780 | atStart = FALSE0; |
| 2781 | break; |
| 2782 | |
| 2783 | |
| 2784 | case URX_SETREF: |
| 2785 | if (currentLen == 0) { |
| 2786 | int32_t sn = URX_VAL(op)((op) & 0xffffff); |
| 2787 | U_ASSERT(sn > 0 && sn < fRXPat->fSets->size())(void)0; |
| 2788 | const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn); |
| 2789 | fRXPat->fInitialChars->addAll(*s); |
| 2790 | numInitialStrings += 2; |
| 2791 | } |
| 2792 | currentLen = safeIncrement(currentLen, 1); |
| 2793 | atStart = FALSE0; |
| 2794 | break; |
| 2795 | |
| 2796 | case URX_LOOP_SR_I: |
| 2797 | // [Set]*, like a SETREF, above, in what it can match, |
| 2798 | // but may not match at all, so currentLen is not incremented. |
| 2799 | if (currentLen == 0) { |
| 2800 | int32_t sn = URX_VAL(op)((op) & 0xffffff); |
| 2801 | U_ASSERT(sn > 0 && sn < fRXPat->fSets->size())(void)0; |
| 2802 | const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn); |
| 2803 | fRXPat->fInitialChars->addAll(*s); |
| 2804 | numInitialStrings += 2; |
| 2805 | } |
| 2806 | atStart = FALSE0; |
| 2807 | break; |
| 2808 | |
| 2809 | case URX_LOOP_DOT_I: |
| 2810 | if (currentLen == 0) { |
| 2811 | // .* at the start of a pattern. |
| 2812 | // Any character can begin the match. |
| 2813 | fRXPat->fInitialChars->clear(); |
| 2814 | fRXPat->fInitialChars->complement(); |
| 2815 | numInitialStrings += 2; |
| 2816 | } |
| 2817 | atStart = FALSE0; |
| 2818 | break; |
| 2819 | |
| 2820 | |
| 2821 | case URX_STATIC_SETREF: |
| 2822 | if (currentLen == 0) { |
| 2823 | int32_t sn = URX_VAL(op)((op) & 0xffffff); |
| 2824 | U_ASSERT(sn>0 && sn<URX_LAST_SET)(void)0; |
| 2825 | const UnicodeSet &s = RegexStaticSets::gStaticSets->fPropSets[sn]; |
| 2826 | fRXPat->fInitialChars->addAll(s); |
| 2827 | numInitialStrings += 2; |
| 2828 | } |
| 2829 | currentLen = safeIncrement(currentLen, 1); |
| 2830 | atStart = FALSE0; |
| 2831 | break; |
| 2832 | |
| 2833 | |
| 2834 | |
| 2835 | case URX_STAT_SETREF_N: |
| 2836 | if (currentLen == 0) { |
| 2837 | int32_t sn = URX_VAL(op)((op) & 0xffffff); |
| 2838 | UnicodeSet sc; |
| 2839 | sc.addAll(RegexStaticSets::gStaticSets->fPropSets[sn]).complement(); |
| 2840 | fRXPat->fInitialChars->addAll(sc); |
| 2841 | numInitialStrings += 2; |
| 2842 | } |
| 2843 | currentLen = safeIncrement(currentLen, 1); |
| 2844 | atStart = FALSE0; |
| 2845 | break; |
| 2846 | |
| 2847 | |
| 2848 | |
| 2849 | case URX_BACKSLASH_D: |
| 2850 | // Digit Char |
| 2851 | if (currentLen == 0) { |
| 2852 | UnicodeSet s; |
| 2853 | s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), *fStatus); |
| 2854 | if (URX_VAL(op)((op) & 0xffffff) != 0) { |
| 2855 | s.complement(); |
| 2856 | } |
| 2857 | fRXPat->fInitialChars->addAll(s); |
| 2858 | numInitialStrings += 2; |
| 2859 | } |
| 2860 | currentLen = safeIncrement(currentLen, 1); |
| 2861 | atStart = FALSE0; |
| 2862 | break; |
| 2863 | |
| 2864 | |
| 2865 | case URX_BACKSLASH_H: |
| 2866 | // Horiz white space |
| 2867 | if (currentLen == 0) { |
| 2868 | UnicodeSet s; |
| 2869 | s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ZS_MASK((uint32_t)1<<(U_SPACE_SEPARATOR)), *fStatus); |
| 2870 | s.add((UChar32)9); // Tab |
| 2871 | if (URX_VAL(op)((op) & 0xffffff) != 0) { |
| 2872 | s.complement(); |
| 2873 | } |
| 2874 | fRXPat->fInitialChars->addAll(s); |
| 2875 | numInitialStrings += 2; |
| 2876 | } |
| 2877 | currentLen = safeIncrement(currentLen, 1); |
| 2878 | atStart = FALSE0; |
| 2879 | break; |
| 2880 | |
| 2881 | |
| 2882 | case URX_BACKSLASH_R: // Any line ending sequence |
| 2883 | case URX_BACKSLASH_V: // Any line ending code point, with optional negation |
| 2884 | if (currentLen == 0) { |
| 2885 | UnicodeSet s; |
| 2886 | s.add((UChar32)0x0a, (UChar32)0x0d); // add range |
| 2887 | s.add((UChar32)0x85); |
| 2888 | s.add((UChar32)0x2028, (UChar32)0x2029); |
| 2889 | if (URX_VAL(op)((op) & 0xffffff) != 0) { |
| 2890 | // Complement option applies to URX_BACKSLASH_V only. |
| 2891 | s.complement(); |
| 2892 | } |
| 2893 | fRXPat->fInitialChars->addAll(s); |
| 2894 | numInitialStrings += 2; |
| 2895 | } |
| 2896 | currentLen = safeIncrement(currentLen, 1); |
| 2897 | atStart = FALSE0; |
| 2898 | break; |
| 2899 | |
| 2900 | |
| 2901 | |
| 2902 | case URX_ONECHAR_I: |
| 2903 | // Case Insensitive Single Character. |
| 2904 | if (currentLen == 0) { |
| 2905 | UChar32 c = URX_VAL(op)((op) & 0xffffff); |
| 2906 | if (u_hasBinaryPropertyu_hasBinaryProperty_71(c, UCHAR_CASE_SENSITIVE)) { |
| 2907 | UnicodeSet starters(c, c); |
| 2908 | starters.closeOver(USET_CASE_INSENSITIVE); |
| 2909 | // findCaseInsensitiveStarters(c, &starters); |
| 2910 | // For ONECHAR_I, no need to worry about text chars that expand on folding into strings. |
| 2911 | // The expanded folding can't match the pattern. |
| 2912 | fRXPat->fInitialChars->addAll(starters); |
| 2913 | } else { |
| 2914 | // Char has no case variants. Just add it as-is to the |
| 2915 | // set of possible starting chars. |
| 2916 | fRXPat->fInitialChars->add(c); |
| 2917 | } |
| 2918 | numInitialStrings += 2; |
| 2919 | } |
| 2920 | currentLen = safeIncrement(currentLen, 1); |
| 2921 | atStart = FALSE0; |
| 2922 | break; |
| 2923 | |
| 2924 | |
| 2925 | case URX_BACKSLASH_X: // Grapheme Cluster. Minimum is 1, max unbounded. |
| 2926 | case URX_DOTANY_ALL: // . matches one or two. |
| 2927 | case URX_DOTANY: |
| 2928 | case URX_DOTANY_UNIX: |
| 2929 | if (currentLen == 0) { |
| 2930 | // These constructs are all bad news when they appear at the start |
| 2931 | // of a match. Any character can begin the match. |
| 2932 | fRXPat->fInitialChars->clear(); |
| 2933 | fRXPat->fInitialChars->complement(); |
| 2934 | numInitialStrings += 2; |
| 2935 | } |
| 2936 | currentLen = safeIncrement(currentLen, 1); |
| 2937 | atStart = FALSE0; |
| 2938 | break; |
| 2939 | |
| 2940 | |
| 2941 | case URX_JMPX: |
| 2942 | loc++; // Except for extra operand on URX_JMPX, same as URX_JMP. |
| 2943 | U_FALLTHROUGH[[clang::fallthrough]]; |
| 2944 | case URX_JMP: |
| 2945 | { |
| 2946 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 2947 | if (jmpDest < loc) { |
| 2948 | // Loop of some kind. Can safely ignore, the worst that will happen |
| 2949 | // is that we understate the true minimum length |
| 2950 | currentLen = forwardedLength.elementAti(loc+1); |
| 2951 | |
| 2952 | } else { |
| 2953 | // Forward jump. Propagate the current min length to the target loc of the jump. |
| 2954 | U_ASSERT(jmpDest <= end+1)(void)0; |
| 2955 | if (forwardedLength.elementAti(jmpDest) > currentLen) { |
| 2956 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 2957 | } |
| 2958 | } |
| 2959 | } |
| 2960 | atStart = FALSE0; |
| 2961 | break; |
| 2962 | |
| 2963 | case URX_JMP_SAV: |
| 2964 | case URX_JMP_SAV_X: |
| 2965 | // Combo of state save to the next loc, + jmp backwards. |
| 2966 | // Net effect on min. length computation is nothing. |
| 2967 | atStart = FALSE0; |
| 2968 | break; |
| 2969 | |
| 2970 | case URX_BACKTRACK: |
| 2971 | // Fails are kind of like a branch, except that the min length was |
| 2972 | // propagated already, by the state save. |
| 2973 | currentLen = forwardedLength.elementAti(loc+1); |
| 2974 | atStart = FALSE0; |
| 2975 | break; |
| 2976 | |
| 2977 | |
| 2978 | case URX_STATE_SAVE: |
| 2979 | { |
| 2980 | // State Save, for forward jumps, propagate the current minimum. |
| 2981 | // of the state save. |
| 2982 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 2983 | if (jmpDest > loc) { |
| 2984 | if (currentLen < forwardedLength.elementAti(jmpDest)) { |
| 2985 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 2986 | } |
| 2987 | } |
| 2988 | } |
| 2989 | atStart = FALSE0; |
| 2990 | break; |
| 2991 | |
| 2992 | |
| 2993 | |
| 2994 | |
| 2995 | case URX_STRING: |
| 2996 | { |
| 2997 | loc++; |
| 2998 | int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 2999 | int32_t stringLen = URX_VAL(stringLenOp)((stringLenOp) & 0xffffff); |
| 3000 | U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN)(void)0; |
| 3001 | U_ASSERT(stringLenOp >= 2)(void)0; |
| 3002 | if (currentLen == 0) { |
| 3003 | // Add the starting character of this string to the set of possible starting |
| 3004 | // characters for this pattern. |
| 3005 | int32_t stringStartIdx = URX_VAL(op)((op) & 0xffffff); |
| 3006 | UChar32 c = fRXPat->fLiteralText.char32At(stringStartIdx); |
| 3007 | fRXPat->fInitialChars->add(c); |
| 3008 | |
| 3009 | // Remember this string. After the entire pattern has been checked, |
| 3010 | // if nothing else is identified that can start a match, we'll use it. |
| 3011 | numInitialStrings++; |
| 3012 | fRXPat->fInitialStringIdx = stringStartIdx; |
| 3013 | fRXPat->fInitialStringLen = stringLen; |
| 3014 | } |
| 3015 | |
| 3016 | currentLen = safeIncrement(currentLen, stringLen); |
| 3017 | atStart = FALSE0; |
| 3018 | } |
| 3019 | break; |
| 3020 | |
| 3021 | case URX_STRING_I: |
| 3022 | { |
| 3023 | // Case-insensitive string. Unlike exact-match strings, we won't |
| 3024 | // attempt a string search for possible match positions. But we |
| 3025 | // do update the set of possible starting characters. |
| 3026 | loc++; |
| 3027 | int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3028 | int32_t stringLen = URX_VAL(stringLenOp)((stringLenOp) & 0xffffff); |
| 3029 | U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN)(void)0; |
| 3030 | U_ASSERT(stringLenOp >= 2)(void)0; |
| 3031 | if (currentLen == 0) { |
| 3032 | // Add the starting character of this string to the set of possible starting |
| 3033 | // characters for this pattern. |
| 3034 | int32_t stringStartIdx = URX_VAL(op)((op) & 0xffffff); |
| 3035 | UChar32 c = fRXPat->fLiteralText.char32At(stringStartIdx); |
| 3036 | UnicodeSet s; |
| 3037 | findCaseInsensitiveStarters(c, &s); |
| 3038 | fRXPat->fInitialChars->addAll(s); |
| 3039 | numInitialStrings += 2; // Matching on an initial string not possible. |
| 3040 | } |
| 3041 | currentLen = safeIncrement(currentLen, stringLen); |
| 3042 | atStart = FALSE0; |
| 3043 | } |
| 3044 | break; |
| 3045 | |
| 3046 | case URX_CTR_INIT: |
| 3047 | case URX_CTR_INIT_NG: |
| 3048 | { |
| 3049 | // Loop Init Ops. These don't change the min length, but they are 4 word ops |
| 3050 | // so location must be updated accordingly. |
| 3051 | // Loop Init Ops. |
| 3052 | // If the min loop count == 0 |
| 3053 | // move loc forwards to the end of the loop, skipping over the body. |
| 3054 | // If the min count is > 0, |
| 3055 | // continue normal processing of the body of the loop. |
| 3056 | int32_t loopEndLoc = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1); |
| 3057 | loopEndLoc = URX_VAL(loopEndLoc)((loopEndLoc) & 0xffffff); |
| 3058 | int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2); |
| 3059 | if (minLoopCount == 0) { |
| 3060 | // Min Loop Count of 0, treat like a forward branch and |
| 3061 | // move the current minimum length up to the target |
| 3062 | // (end of loop) location. |
| 3063 | U_ASSERT(loopEndLoc <= end+1)(void)0; |
| 3064 | if (forwardedLength.elementAti(loopEndLoc) > currentLen) { |
| 3065 | forwardedLength.setElementAt(currentLen, loopEndLoc); |
| 3066 | } |
| 3067 | } |
| 3068 | loc+=3; // Skips over operands of CTR_INIT |
| 3069 | } |
| 3070 | atStart = FALSE0; |
| 3071 | break; |
| 3072 | |
| 3073 | |
| 3074 | case URX_CTR_LOOP: |
| 3075 | case URX_CTR_LOOP_NG: |
| 3076 | // Loop ops. |
| 3077 | // The jump is conditional, backwards only. |
| 3078 | atStart = FALSE0; |
| 3079 | break; |
| 3080 | |
| 3081 | case URX_LOOP_C: |
| 3082 | // More loop ops. These state-save to themselves. |
| 3083 | // don't change the minimum match |
| 3084 | atStart = FALSE0; |
| 3085 | break; |
| 3086 | |
| 3087 | |
| 3088 | case URX_LA_START: |
| 3089 | case URX_LB_START: |
| 3090 | { |
| 3091 | // Look-around. Scan forward until the matching look-ahead end, |
| 3092 | // without processing the look-around block. This is overly pessimistic. |
| 3093 | |
| 3094 | // Keep track of the nesting depth of look-around blocks. Boilerplate code for |
| 3095 | // lookahead contains two LA_END instructions, so count goes up by two |
| 3096 | // for each LA_START. |
| 3097 | int32_t depth = (opType == URX_LA_START? 2: 1); |
| 3098 | for (;;) { |
| 3099 | loc++; |
| 3100 | op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3101 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_LA_START) { |
| 3102 | depth+=2; |
| 3103 | } |
| 3104 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_LB_START) { |
| 3105 | depth++; |
| 3106 | } |
| 3107 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_LA_END || URX_TYPE(op)((uint32_t)(op) >> 24)==URX_LBN_END) { |
| 3108 | depth--; |
| 3109 | if (depth == 0) { |
| 3110 | break; |
| 3111 | } |
| 3112 | } |
| 3113 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_STATE_SAVE) { |
| 3114 | // Need this because neg lookahead blocks will FAIL to outside |
| 3115 | // of the block. |
| 3116 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 3117 | if (jmpDest > loc) { |
| 3118 | if (currentLen < forwardedLength.elementAti(jmpDest)) { |
| 3119 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 3120 | } |
| 3121 | } |
| 3122 | } |
| 3123 | U_ASSERT(loc <= end)(void)0; |
| 3124 | } |
| 3125 | } |
| 3126 | break; |
| 3127 | |
| 3128 | case URX_LA_END: |
| 3129 | case URX_LB_CONT: |
| 3130 | case URX_LB_END: |
| 3131 | case URX_LBN_CONT: |
| 3132 | case URX_LBN_END: |
| 3133 | UPRV_UNREACHABLE_EXITabort(); // Shouldn't get here. These ops should be |
| 3134 | // consumed by the scan in URX_LA_START and LB_START |
| 3135 | default: |
| 3136 | UPRV_UNREACHABLE_EXITabort(); |
| 3137 | } |
| 3138 | |
| 3139 | } |
| 3140 | |
| 3141 | |
| 3142 | // We have finished walking through the ops. Check whether some forward jump |
| 3143 | // propagated a shorter length to location end+1. |
| 3144 | if (forwardedLength.elementAti(end+1) < currentLen) { |
| 3145 | currentLen = forwardedLength.elementAti(end+1); |
| 3146 | } |
| 3147 | |
| 3148 | |
| 3149 | fRXPat->fInitialChars8->init(fRXPat->fInitialChars); |
| 3150 | |
| 3151 | |
| 3152 | // Sort out what we should check for when looking for candidate match start positions. |
| 3153 | // In order of preference, |
| 3154 | // 1. Start of input text buffer. |
| 3155 | // 2. A literal string. |
| 3156 | // 3. Start of line in multi-line mode. |
| 3157 | // 4. A single literal character. |
| 3158 | // 5. A character from a set of characters. |
| 3159 | // |
| 3160 | if (fRXPat->fStartType == START_START) { |
| 3161 | // Match only at the start of an input text string. |
| 3162 | // start type is already set. We're done. |
| 3163 | } else if (numInitialStrings == 1 && fRXPat->fMinMatchLen > 0) { |
| 3164 | // Match beginning only with a literal string. |
| 3165 | UChar32 c = fRXPat->fLiteralText.char32At(fRXPat->fInitialStringIdx); |
| 3166 | U_ASSERT(fRXPat->fInitialChars->contains(c))(void)0; |
| 3167 | fRXPat->fStartType = START_STRING; |
| 3168 | fRXPat->fInitialChar = c; |
| 3169 | } else if (fRXPat->fStartType == START_LINE) { |
| 3170 | // Match at start of line in Multi-Line mode. |
| 3171 | // Nothing to do here; everything is already set. |
| 3172 | } else if (fRXPat->fMinMatchLen == 0) { |
| 3173 | // Zero length match possible. We could start anywhere. |
| 3174 | fRXPat->fStartType = START_NO_INFO; |
| 3175 | } else if (fRXPat->fInitialChars->size() == 1) { |
| 3176 | // All matches begin with the same char. |
| 3177 | fRXPat->fStartType = START_CHAR; |
| 3178 | fRXPat->fInitialChar = fRXPat->fInitialChars->charAt(0); |
| 3179 | U_ASSERT(fRXPat->fInitialChar != (UChar32)-1)(void)0; |
| 3180 | } else if (fRXPat->fInitialChars->contains((UChar32)0, (UChar32)0x10ffff) == FALSE0 && |
| 3181 | fRXPat->fMinMatchLen > 0) { |
| 3182 | // Matches start with a set of character smaller than the set of all chars. |
| 3183 | fRXPat->fStartType = START_SET; |
| 3184 | } else { |
| 3185 | // Matches can start with anything |
| 3186 | fRXPat->fStartType = START_NO_INFO; |
| 3187 | } |
| 3188 | |
| 3189 | return; |
| 3190 | } |
| 3191 | |
| 3192 | |
| 3193 | |
| 3194 | //------------------------------------------------------------------------------ |
| 3195 | // |
| 3196 | // minMatchLength Calculate the length of the shortest string that could |
| 3197 | // match the specified pattern. |
| 3198 | // Length is in 16 bit code units, not code points. |
| 3199 | // |
| 3200 | // The calculated length may not be exact. The returned |
| 3201 | // value may be shorter than the actual minimum; it must |
| 3202 | // never be longer. |
| 3203 | // |
| 3204 | // start and end are the range of p-code operations to be |
| 3205 | // examined. The endpoints are included in the range. |
| 3206 | // |
| 3207 | //------------------------------------------------------------------------------ |
| 3208 | int32_t RegexCompile::minMatchLength(int32_t start, int32_t end) { |
| 3209 | if (U_FAILURE(*fStatus)) { |
| 3210 | return 0; |
| 3211 | } |
| 3212 | |
| 3213 | U_ASSERT(start <= end)(void)0; |
| 3214 | U_ASSERT(end < fRXPat->fCompiledPat->size())(void)0; |
| 3215 | |
| 3216 | |
| 3217 | int32_t loc; |
| 3218 | int32_t op; |
| 3219 | int32_t opType; |
| 3220 | int32_t currentLen = 0; |
| 3221 | |
| 3222 | |
| 3223 | // forwardedLength is a vector holding minimum-match-length values that |
| 3224 | // are propagated forward in the pattern by JMP or STATE_SAVE operations. |
| 3225 | // It must be one longer than the pattern being checked because some ops |
| 3226 | // will jmp to a end-of-block+1 location from within a block, and we must |
| 3227 | // count those when checking the block. |
| 3228 | UVector32 forwardedLength(end+2, *fStatus); |
| 3229 | forwardedLength.setSize(end+2); |
| 3230 | for (loc=start; loc<=end+1; loc++) { |
| 3231 | forwardedLength.setElementAt(INT32_MAX(2147483647), loc); |
| 3232 | } |
| 3233 | |
| 3234 | for (loc = start; loc<=end; loc++) { |
| 3235 | op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3236 | opType = URX_TYPE(op)((uint32_t)(op) >> 24); |
| 3237 | |
| 3238 | // The loop is advancing linearly through the pattern. |
| 3239 | // If the op we are now at was the destination of a branch in the pattern, |
| 3240 | // and that path has a shorter minimum length than the current accumulated value, |
| 3241 | // replace the current accumulated value. |
| 3242 | // U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); // MinLength == INT32_MAX for some |
| 3243 | // no-match-possible cases. |
| 3244 | if (forwardedLength.elementAti(loc) < currentLen) { |
| 3245 | currentLen = forwardedLength.elementAti(loc); |
| 3246 | U_ASSERT(currentLen>=0 && currentLen < INT32_MAX)(void)0; |
| 3247 | } |
| 3248 | |
| 3249 | switch (opType) { |
| 3250 | // Ops that don't change the total length matched |
| 3251 | case URX_RESERVED_OP: |
| 3252 | case URX_END: |
| 3253 | case URX_STRING_LEN: |
| 3254 | case URX_NOP: |
| 3255 | case URX_START_CAPTURE: |
| 3256 | case URX_END_CAPTURE: |
| 3257 | case URX_BACKSLASH_B: |
| 3258 | case URX_BACKSLASH_BU: |
| 3259 | case URX_BACKSLASH_G: |
| 3260 | case URX_BACKSLASH_Z: |
| 3261 | case URX_CARET: |
| 3262 | case URX_DOLLAR: |
| 3263 | case URX_DOLLAR_M: |
| 3264 | case URX_DOLLAR_D: |
| 3265 | case URX_DOLLAR_MD: |
| 3266 | case URX_RELOC_OPRND: |
| 3267 | case URX_STO_INP_LOC: |
| 3268 | case URX_CARET_M: |
| 3269 | case URX_CARET_M_UNIX: |
| 3270 | case URX_BACKREF: // BackRef. Must assume that it might be a zero length match |
| 3271 | case URX_BACKREF_I: |
| 3272 | |
| 3273 | case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. |
| 3274 | case URX_LD_SP: |
| 3275 | |
| 3276 | case URX_JMP_SAV: |
| 3277 | case URX_JMP_SAV_X: |
| 3278 | break; |
| 3279 | |
| 3280 | |
| 3281 | // Ops that match a minimum of one character (one or two 16 bit code units.) |
| 3282 | // |
| 3283 | case URX_ONECHAR: |
| 3284 | case URX_STATIC_SETREF: |
| 3285 | case URX_STAT_SETREF_N: |
| 3286 | case URX_SETREF: |
| 3287 | case URX_BACKSLASH_D: |
| 3288 | case URX_BACKSLASH_H: |
| 3289 | case URX_BACKSLASH_R: |
| 3290 | case URX_BACKSLASH_V: |
| 3291 | case URX_ONECHAR_I: |
| 3292 | case URX_BACKSLASH_X: // Grapheme Cluster. Minimum is 1, max unbounded. |
| 3293 | case URX_DOTANY_ALL: // . matches one or two. |
| 3294 | case URX_DOTANY: |
| 3295 | case URX_DOTANY_UNIX: |
| 3296 | currentLen = safeIncrement(currentLen, 1); |
| 3297 | break; |
| 3298 | |
| 3299 | |
| 3300 | case URX_JMPX: |
| 3301 | loc++; // URX_JMPX has an extra operand, ignored here, |
| 3302 | // otherwise processed identically to URX_JMP. |
| 3303 | U_FALLTHROUGH[[clang::fallthrough]]; |
| 3304 | case URX_JMP: |
| 3305 | { |
| 3306 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 3307 | if (jmpDest < loc) { |
| 3308 | // Loop of some kind. Can safely ignore, the worst that will happen |
| 3309 | // is that we understate the true minimum length |
| 3310 | currentLen = forwardedLength.elementAti(loc+1); |
| 3311 | } else { |
| 3312 | // Forward jump. Propagate the current min length to the target loc of the jump. |
| 3313 | U_ASSERT(jmpDest <= end+1)(void)0; |
| 3314 | if (forwardedLength.elementAti(jmpDest) > currentLen) { |
| 3315 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 3316 | } |
| 3317 | } |
| 3318 | } |
| 3319 | break; |
| 3320 | |
| 3321 | case URX_BACKTRACK: |
| 3322 | { |
| 3323 | // Back-tracks are kind of like a branch, except that the min length was |
| 3324 | // propagated already, by the state save. |
| 3325 | currentLen = forwardedLength.elementAti(loc+1); |
| 3326 | } |
| 3327 | break; |
| 3328 | |
| 3329 | |
| 3330 | case URX_STATE_SAVE: |
| 3331 | { |
| 3332 | // State Save, for forward jumps, propagate the current minimum. |
| 3333 | // of the state save. |
| 3334 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 3335 | if (jmpDest > loc) { |
| 3336 | if (currentLen < forwardedLength.elementAti(jmpDest)) { |
| 3337 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 3338 | } |
| 3339 | } |
| 3340 | } |
| 3341 | break; |
| 3342 | |
| 3343 | |
| 3344 | case URX_STRING: |
| 3345 | { |
| 3346 | loc++; |
| 3347 | int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3348 | currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp)((stringLenOp) & 0xffffff)); |
| 3349 | } |
| 3350 | break; |
| 3351 | |
| 3352 | |
| 3353 | case URX_STRING_I: |
| 3354 | { |
| 3355 | loc++; |
| 3356 | // TODO: with full case folding, matching input text may be shorter than |
| 3357 | // the string we have here. More smarts could put some bounds on it. |
| 3358 | // Assume a min length of one for now. A min length of zero causes |
| 3359 | // optimization failures for a pattern like "string"+ |
| 3360 | // currentLen += URX_VAL(stringLenOp); |
| 3361 | currentLen = safeIncrement(currentLen, 1); |
| 3362 | } |
| 3363 | break; |
| 3364 | |
| 3365 | case URX_CTR_INIT: |
| 3366 | case URX_CTR_INIT_NG: |
| 3367 | { |
| 3368 | // Loop Init Ops. |
| 3369 | // If the min loop count == 0 |
| 3370 | // move loc forwards to the end of the loop, skipping over the body. |
| 3371 | // If the min count is > 0, |
| 3372 | // continue normal processing of the body of the loop. |
| 3373 | int32_t loopEndLoc = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1); |
| 3374 | loopEndLoc = URX_VAL(loopEndLoc)((loopEndLoc) & 0xffffff); |
| 3375 | int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2); |
| 3376 | if (minLoopCount == 0) { |
| 3377 | loc = loopEndLoc; |
| 3378 | } else { |
| 3379 | loc+=3; // Skips over operands of CTR_INIT |
| 3380 | } |
| 3381 | } |
| 3382 | break; |
| 3383 | |
| 3384 | |
| 3385 | case URX_CTR_LOOP: |
| 3386 | case URX_CTR_LOOP_NG: |
| 3387 | // Loop ops. |
| 3388 | // The jump is conditional, backwards only. |
| 3389 | break; |
| 3390 | |
| 3391 | case URX_LOOP_SR_I: |
| 3392 | case URX_LOOP_DOT_I: |
| 3393 | case URX_LOOP_C: |
| 3394 | // More loop ops. These state-save to themselves. |
| 3395 | // don't change the minimum match - could match nothing at all. |
| 3396 | break; |
| 3397 | |
| 3398 | |
| 3399 | case URX_LA_START: |
| 3400 | case URX_LB_START: |
| 3401 | { |
| 3402 | // Look-around. Scan forward until the matching look-ahead end, |
| 3403 | // without processing the look-around block. This is overly pessimistic for look-ahead, |
| 3404 | // it assumes that the look-ahead match might be zero-length. |
| 3405 | // TODO: Positive lookahead could recursively do the block, then continue |
| 3406 | // with the longer of the block or the value coming in. Ticket 6060 |
| 3407 | int32_t depth = (opType == URX_LA_START? 2: 1); |
| 3408 | for (;;) { |
| 3409 | loc++; |
| 3410 | op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3411 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_LA_START) { |
| 3412 | // The boilerplate for look-ahead includes two LA_END instructions, |
| 3413 | // Depth will be decremented by each one when it is seen. |
| 3414 | depth += 2; |
| 3415 | } |
| 3416 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_LB_START) { |
| 3417 | depth++; |
| 3418 | } |
| 3419 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_LA_END) { |
| 3420 | depth--; |
| 3421 | if (depth == 0) { |
| 3422 | break; |
| 3423 | } |
| 3424 | } |
| 3425 | if (URX_TYPE(op)((uint32_t)(op) >> 24)==URX_LBN_END) { |
| 3426 | depth--; |
| 3427 | if (depth == 0) { |
| 3428 | break; |
| 3429 | } |
| 3430 | } |
| 3431 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_STATE_SAVE) { |
| 3432 | // Need this because neg lookahead blocks will FAIL to outside |
| 3433 | // of the block. |
| 3434 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 3435 | if (jmpDest > loc) { |
| 3436 | if (currentLen < forwardedLength.elementAti(jmpDest)) { |
| 3437 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 3438 | } |
| 3439 | } |
| 3440 | } |
| 3441 | U_ASSERT(loc <= end)(void)0; |
| 3442 | } |
| 3443 | } |
| 3444 | break; |
| 3445 | |
| 3446 | case URX_LA_END: |
| 3447 | case URX_LB_CONT: |
| 3448 | case URX_LB_END: |
| 3449 | case URX_LBN_CONT: |
| 3450 | case URX_LBN_END: |
| 3451 | // Only come here if the matching URX_LA_START or URX_LB_START was not in the |
| 3452 | // range being sized, which happens when measuring size of look-behind blocks. |
| 3453 | break; |
| 3454 | |
| 3455 | default: |
| 3456 | UPRV_UNREACHABLE_EXITabort(); |
| 3457 | } |
| 3458 | |
| 3459 | } |
| 3460 | |
| 3461 | // We have finished walking through the ops. Check whether some forward jump |
| 3462 | // propagated a shorter length to location end+1. |
| 3463 | if (forwardedLength.elementAti(end+1) < currentLen) { |
| 3464 | currentLen = forwardedLength.elementAti(end+1); |
| 3465 | U_ASSERT(currentLen>=0 && currentLen < INT32_MAX)(void)0; |
| 3466 | } |
| 3467 | |
| 3468 | return currentLen; |
| 3469 | } |
| 3470 | |
| 3471 | //------------------------------------------------------------------------------ |
| 3472 | // |
| 3473 | // maxMatchLength Calculate the length of the longest string that could |
| 3474 | // match the specified pattern. |
| 3475 | // Length is in 16 bit code units, not code points. |
| 3476 | // |
| 3477 | // The calculated length may not be exact. The returned |
| 3478 | // value may be longer than the actual maximum; it must |
| 3479 | // never be shorter. |
| 3480 | // |
| 3481 | // start, end: the range of the pattern to check. |
| 3482 | // end is inclusive. |
| 3483 | // |
| 3484 | //------------------------------------------------------------------------------ |
| 3485 | int32_t RegexCompile::maxMatchLength(int32_t start, int32_t end) { |
| 3486 | if (U_FAILURE(*fStatus)) { |
| 3487 | return 0; |
| 3488 | } |
| 3489 | U_ASSERT(start <= end)(void)0; |
| 3490 | U_ASSERT(end < fRXPat->fCompiledPat->size())(void)0; |
| 3491 | |
| 3492 | int32_t loc; |
| 3493 | int32_t op; |
| 3494 | int32_t opType; |
| 3495 | int32_t currentLen = 0; |
| 3496 | UVector32 forwardedLength(end+1, *fStatus); |
| 3497 | forwardedLength.setSize(end+1); |
| 3498 | |
| 3499 | for (loc=start; loc<=end; loc++) { |
| 3500 | forwardedLength.setElementAt(0, loc); |
| 3501 | } |
| 3502 | |
| 3503 | for (loc = start; loc<=end; loc++) { |
| 3504 | op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3505 | opType = URX_TYPE(op)((uint32_t)(op) >> 24); |
| 3506 | |
| 3507 | // The loop is advancing linearly through the pattern. |
| 3508 | // If the op we are now at was the destination of a branch in the pattern, |
| 3509 | // and that path has a longer maximum length than the current accumulated value, |
| 3510 | // replace the current accumulated value. |
| 3511 | if (forwardedLength.elementAti(loc) > currentLen) { |
| 3512 | currentLen = forwardedLength.elementAti(loc); |
| 3513 | } |
| 3514 | |
| 3515 | switch (opType) { |
| 3516 | // Ops that don't change the total length matched |
| 3517 | case URX_RESERVED_OP: |
| 3518 | case URX_END: |
| 3519 | case URX_STRING_LEN: |
| 3520 | case URX_NOP: |
| 3521 | case URX_START_CAPTURE: |
| 3522 | case URX_END_CAPTURE: |
| 3523 | case URX_BACKSLASH_B: |
| 3524 | case URX_BACKSLASH_BU: |
| 3525 | case URX_BACKSLASH_G: |
| 3526 | case URX_BACKSLASH_Z: |
| 3527 | case URX_CARET: |
| 3528 | case URX_DOLLAR: |
| 3529 | case URX_DOLLAR_M: |
| 3530 | case URX_DOLLAR_D: |
| 3531 | case URX_DOLLAR_MD: |
| 3532 | case URX_RELOC_OPRND: |
| 3533 | case URX_STO_INP_LOC: |
| 3534 | case URX_CARET_M: |
| 3535 | case URX_CARET_M_UNIX: |
| 3536 | |
| 3537 | case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. |
| 3538 | case URX_LD_SP: |
| 3539 | |
| 3540 | case URX_LB_END: |
| 3541 | case URX_LB_CONT: |
| 3542 | case URX_LBN_CONT: |
| 3543 | case URX_LBN_END: |
| 3544 | break; |
| 3545 | |
| 3546 | |
| 3547 | // Ops that increase that cause an unbounded increase in the length |
| 3548 | // of a matched string, or that increase it a hard to characterize way. |
| 3549 | // Call the max length unbounded, and stop further checking. |
| 3550 | case URX_BACKREF: // BackRef. Must assume that it might be a zero length match |
| 3551 | case URX_BACKREF_I: |
| 3552 | case URX_BACKSLASH_X: // Grapheme Cluster. Minimum is 1, max unbounded. |
| 3553 | currentLen = INT32_MAX(2147483647); |
| 3554 | break; |
| 3555 | |
| 3556 | |
| 3557 | // Ops that match a max of one character (possibly two 16 bit code units.) |
| 3558 | // |
| 3559 | case URX_STATIC_SETREF: |
| 3560 | case URX_STAT_SETREF_N: |
| 3561 | case URX_SETREF: |
| 3562 | case URX_BACKSLASH_D: |
| 3563 | case URX_BACKSLASH_H: |
| 3564 | case URX_BACKSLASH_R: |
| 3565 | case URX_BACKSLASH_V: |
| 3566 | case URX_ONECHAR_I: |
| 3567 | case URX_DOTANY_ALL: |
| 3568 | case URX_DOTANY: |
| 3569 | case URX_DOTANY_UNIX: |
| 3570 | currentLen = safeIncrement(currentLen, 2); |
| 3571 | break; |
| 3572 | |
| 3573 | // Single literal character. Increase current max length by one or two, |
| 3574 | // depending on whether the char is in the supplementary range. |
| 3575 | case URX_ONECHAR: |
| 3576 | currentLen = safeIncrement(currentLen, 1); |
| 3577 | if (URX_VAL(op)((op) & 0xffffff) > 0x10000) { |
| 3578 | currentLen = safeIncrement(currentLen, 1); |
| 3579 | } |
| 3580 | break; |
| 3581 | |
| 3582 | // Jumps. |
| 3583 | // |
| 3584 | case URX_JMP: |
| 3585 | case URX_JMPX: |
| 3586 | case URX_JMP_SAV: |
| 3587 | case URX_JMP_SAV_X: |
| 3588 | { |
| 3589 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 3590 | if (jmpDest < loc) { |
| 3591 | // Loop of some kind. Max match length is unbounded. |
| 3592 | currentLen = INT32_MAX(2147483647); |
| 3593 | } else { |
| 3594 | // Forward jump. Propagate the current min length to the target loc of the jump. |
| 3595 | if (forwardedLength.elementAti(jmpDest) < currentLen) { |
| 3596 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 3597 | } |
| 3598 | currentLen = 0; |
| 3599 | } |
| 3600 | } |
| 3601 | break; |
| 3602 | |
| 3603 | case URX_BACKTRACK: |
| 3604 | // back-tracks are kind of like a branch, except that the max length was |
| 3605 | // propagated already, by the state save. |
| 3606 | currentLen = forwardedLength.elementAti(loc+1); |
| 3607 | break; |
| 3608 | |
| 3609 | |
| 3610 | case URX_STATE_SAVE: |
| 3611 | { |
| 3612 | // State Save, for forward jumps, propagate the current minimum. |
| 3613 | // of the state save. |
| 3614 | // For backwards jumps, they create a loop, maximum |
| 3615 | // match length is unbounded. |
| 3616 | int32_t jmpDest = URX_VAL(op)((op) & 0xffffff); |
| 3617 | if (jmpDest > loc) { |
| 3618 | if (currentLen > forwardedLength.elementAti(jmpDest)) { |
| 3619 | forwardedLength.setElementAt(currentLen, jmpDest); |
| 3620 | } |
| 3621 | } else { |
| 3622 | currentLen = INT32_MAX(2147483647); |
| 3623 | } |
| 3624 | } |
| 3625 | break; |
| 3626 | |
| 3627 | |
| 3628 | |
| 3629 | |
| 3630 | case URX_STRING: |
| 3631 | { |
| 3632 | loc++; |
| 3633 | int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3634 | currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp)((stringLenOp) & 0xffffff)); |
| 3635 | break; |
| 3636 | } |
| 3637 | |
| 3638 | case URX_STRING_I: |
| 3639 | // TODO: This code assumes that any user string that matches will be no longer |
| 3640 | // than our compiled string, with case insensitive matching. |
| 3641 | // Our compiled string has been case-folded already. |
| 3642 | // |
| 3643 | // Any matching user string will have no more code points than our |
| 3644 | // compiled (folded) string. Folding may add code points, but |
| 3645 | // not remove them. |
| 3646 | // |
| 3647 | // There is a potential problem if a supplemental code point |
| 3648 | // case-folds to a BMP code point. In this case our compiled string |
| 3649 | // could be shorter (in code units) than a matching user string. |
| 3650 | // |
| 3651 | // At this time (Unicode 6.1) there are no such characters, and this case |
| 3652 | // is not being handled. A test, intltest regex/Bug9283, will fail if |
| 3653 | // any problematic characters are added to Unicode. |
| 3654 | // |
| 3655 | // If this happens, we can make a set of the BMP chars that the |
| 3656 | // troublesome supplementals fold to, scan our string, and bump the |
| 3657 | // currentLen one extra for each that is found. |
| 3658 | // |
| 3659 | { |
| 3660 | loc++; |
| 3661 | int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3662 | currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp)((stringLenOp) & 0xffffff)); |
| 3663 | } |
| 3664 | break; |
| 3665 | |
| 3666 | case URX_CTR_INIT: |
| 3667 | case URX_CTR_INIT_NG: |
| 3668 | // For Loops, recursively call this function on the pattern for the loop body, |
| 3669 | // then multiply the result by the maximum loop count. |
| 3670 | { |
| 3671 | int32_t loopEndLoc = URX_VAL(fRXPat->fCompiledPat->elementAti(loc+1))((fRXPat->fCompiledPat->elementAti(loc+1)) & 0xffffff ); |
| 3672 | if (loopEndLoc == loc+4) { |
| 3673 | // Loop has an empty body. No affect on max match length. |
| 3674 | // Continue processing with code after the loop end. |
| 3675 | loc = loopEndLoc; |
| 3676 | break; |
| 3677 | } |
| 3678 | |
| 3679 | int32_t maxLoopCount = static_cast<int32_t>(fRXPat->fCompiledPat->elementAti(loc+3)); |
| 3680 | if (maxLoopCount == -1) { |
| 3681 | // Unbounded Loop. No upper bound on match length. |
| 3682 | currentLen = INT32_MAX(2147483647); |
| 3683 | break; |
| 3684 | } |
| 3685 | |
| 3686 | U_ASSERT(loopEndLoc >= loc+4)(void)0; |
| 3687 | int64_t blockLen = maxMatchLength(loc+4, loopEndLoc-1); // Recursive call. |
| 3688 | int64_t updatedLen = (int64_t)currentLen + blockLen * maxLoopCount; |
| 3689 | if (updatedLen >= INT32_MAX(2147483647)) { |
| 3690 | currentLen = INT32_MAX(2147483647); |
| 3691 | break; |
| 3692 | } |
| 3693 | currentLen = (int32_t)updatedLen; |
| 3694 | loc = loopEndLoc; |
| 3695 | break; |
| 3696 | } |
| 3697 | |
| 3698 | case URX_CTR_LOOP: |
| 3699 | case URX_CTR_LOOP_NG: |
| 3700 | // These opcodes will be skipped over by code for URX_CTR_INIT. |
| 3701 | // We shouldn't encounter them here. |
| 3702 | UPRV_UNREACHABLE_EXITabort(); |
| 3703 | |
| 3704 | case URX_LOOP_SR_I: |
| 3705 | case URX_LOOP_DOT_I: |
| 3706 | case URX_LOOP_C: |
| 3707 | // For anything to do with loops, make the match length unbounded. |
| 3708 | currentLen = INT32_MAX(2147483647); |
| 3709 | break; |
| 3710 | |
| 3711 | |
| 3712 | |
| 3713 | case URX_LA_START: |
| 3714 | case URX_LA_END: |
| 3715 | // Look-ahead. Just ignore, treat the look-ahead block as if |
| 3716 | // it were normal pattern. Gives a too-long match length, |
| 3717 | // but good enough for now. |
| 3718 | break; |
| 3719 | |
| 3720 | // End of look-ahead ops should always be consumed by the processing at |
| 3721 | // the URX_LA_START op. |
| 3722 | // UPRV_UNREACHABLE_EXIT; |
| 3723 | |
| 3724 | case URX_LB_START: |
| 3725 | { |
| 3726 | // Look-behind. Scan forward until the matching look-around end, |
| 3727 | // without processing the look-behind block. |
| 3728 | int32_t dataLoc = URX_VAL(op)((op) & 0xffffff); |
| 3729 | for (loc = loc + 1; loc <= end; ++loc) { |
| 3730 | op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3731 | int32_t opType = URX_TYPE(op)((uint32_t)(op) >> 24); |
| 3732 | if ((opType == URX_LA_END || opType == URX_LBN_END) && (URX_VAL(op)((op) & 0xffffff) == dataLoc)) { |
| 3733 | break; |
| 3734 | } |
| 3735 | } |
| 3736 | U_ASSERT(loc <= end)(void)0; |
| 3737 | } |
| 3738 | break; |
| 3739 | |
| 3740 | default: |
| 3741 | UPRV_UNREACHABLE_EXITabort(); |
| 3742 | } |
| 3743 | |
| 3744 | |
| 3745 | if (currentLen == INT32_MAX(2147483647)) { |
| 3746 | // The maximum length is unbounded. |
| 3747 | // Stop further processing of the pattern. |
| 3748 | break; |
| 3749 | } |
| 3750 | |
| 3751 | } |
| 3752 | return currentLen; |
| 3753 | |
| 3754 | } |
| 3755 | |
| 3756 | |
| 3757 | //------------------------------------------------------------------------------ |
| 3758 | // |
| 3759 | // stripNOPs Remove any NOP operations from the compiled pattern code. |
| 3760 | // Extra NOPs are inserted for some constructs during the initial |
| 3761 | // code generation to provide locations that may be patched later. |
| 3762 | // Many end up unneeded, and are removed by this function. |
| 3763 | // |
| 3764 | // In order to minimize the number of passes through the pattern, |
| 3765 | // back-reference fixup is also performed here (adjusting |
| 3766 | // back-reference operands to point to the correct frame offsets). |
| 3767 | // |
| 3768 | //------------------------------------------------------------------------------ |
| 3769 | void RegexCompile::stripNOPs() { |
| 3770 | |
| 3771 | if (U_FAILURE(*fStatus)) { |
| 3772 | return; |
| 3773 | } |
| 3774 | |
| 3775 | int32_t end = fRXPat->fCompiledPat->size(); |
| 3776 | UVector32 deltas(end, *fStatus); |
| 3777 | |
| 3778 | // Make a first pass over the code, computing the amount that things |
| 3779 | // will be offset at each location in the original code. |
| 3780 | int32_t loc; |
| 3781 | int32_t d = 0; |
| 3782 | for (loc=0; loc<end; loc++) { |
| 3783 | deltas.addElement(d, *fStatus); |
| 3784 | int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); |
| 3785 | if (URX_TYPE(op)((uint32_t)(op) >> 24) == URX_NOP) { |
| 3786 | d++; |
| 3787 | } |
| 3788 | } |
| 3789 | |
| 3790 | UnicodeString caseStringBuffer; |
| 3791 | |
| 3792 | // Make a second pass over the code, removing the NOPs by moving following |
| 3793 | // code up, and patching operands that refer to code locations that |
| 3794 | // are being moved. The array of offsets from the first step is used |
| 3795 | // to compute the new operand values. |
| 3796 | int32_t src; |
| 3797 | int32_t dst = 0; |
| 3798 | for (src=0; src<end; src++) { |
| 3799 | int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(src); |
| 3800 | int32_t opType = URX_TYPE(op)((uint32_t)(op) >> 24); |
| 3801 | switch (opType) { |
| 3802 | case URX_NOP: |
| 3803 | break; |
| 3804 | |
| 3805 | case URX_STATE_SAVE: |
| 3806 | case URX_JMP: |
| 3807 | case URX_CTR_LOOP: |
| 3808 | case URX_CTR_LOOP_NG: |
| 3809 | case URX_RELOC_OPRND: |
| 3810 | case URX_JMPX: |
| 3811 | case URX_JMP_SAV: |
| 3812 | case URX_JMP_SAV_X: |
| 3813 | // These are instructions with operands that refer to code locations. |
| 3814 | { |
| 3815 | int32_t operandAddress = URX_VAL(op)((op) & 0xffffff); |
| 3816 | U_ASSERT(operandAddress>=0 && operandAddress<deltas.size())(void)0; |
| 3817 | int32_t fixedOperandAddress = operandAddress - deltas.elementAti(operandAddress); |
| 3818 | op = buildOp(opType, fixedOperandAddress); |
| 3819 | fRXPat->fCompiledPat->setElementAt(op, dst); |
| 3820 | dst++; |
| 3821 | break; |
| 3822 | } |
| 3823 | |
| 3824 | case URX_BACKREF: |
| 3825 | case URX_BACKREF_I: |
| 3826 | { |
| 3827 | int32_t where = URX_VAL(op)((op) & 0xffffff); |
| 3828 | if (where > fRXPat->fGroupMap->size()) { |
| 3829 | error(U_REGEX_INVALID_BACK_REF); |
| 3830 | break; |
| 3831 | } |
| 3832 | where = fRXPat->fGroupMap->elementAti(where-1); |
| 3833 | op = buildOp(opType, where); |
| 3834 | fRXPat->fCompiledPat->setElementAt(op, dst); |
| 3835 | dst++; |
| 3836 | |
| 3837 | fRXPat->fNeedsAltInput = TRUE1; |
| 3838 | break; |
| 3839 | } |
| 3840 | case URX_RESERVED_OP: |
| 3841 | case URX_RESERVED_OP_N: |
| 3842 | case URX_BACKTRACK: |
| 3843 | case URX_END: |
| 3844 | case URX_ONECHAR: |
| 3845 | case URX_STRING: |
| 3846 | case URX_STRING_LEN: |
| 3847 | case URX_START_CAPTURE: |
| 3848 | case URX_END_CAPTURE: |
| 3849 | case URX_STATIC_SETREF: |
| 3850 | case URX_STAT_SETREF_N: |
| 3851 | case URX_SETREF: |
| 3852 | case URX_DOTANY: |
| 3853 | case URX_FAIL: |
| 3854 | case URX_BACKSLASH_B: |
| 3855 | case URX_BACKSLASH_BU: |
| 3856 | case URX_BACKSLASH_G: |
| 3857 | case URX_BACKSLASH_X: |
| 3858 | case URX_BACKSLASH_Z: |
| 3859 | case URX_DOTANY_ALL: |
| 3860 | case URX_BACKSLASH_D: |
| 3861 | case URX_CARET: |
| 3862 | case URX_DOLLAR: |
| 3863 | case URX_CTR_INIT: |
| 3864 | case URX_CTR_INIT_NG: |
| 3865 | case URX_DOTANY_UNIX: |
| 3866 | case URX_STO_SP: |
| 3867 | case URX_LD_SP: |
| 3868 | case URX_STO_INP_LOC: |
| 3869 | case URX_LA_START: |
| 3870 | case URX_LA_END: |
| 3871 | case URX_ONECHAR_I: |
| 3872 | case URX_STRING_I: |
| 3873 | case URX_DOLLAR_M: |
| 3874 | case URX_CARET_M: |
| 3875 | case URX_CARET_M_UNIX: |
| 3876 | case URX_LB_START: |
| 3877 | case URX_LB_CONT: |
| 3878 | case URX_LB_END: |
| 3879 | case URX_LBN_CONT: |
| 3880 | case URX_LBN_END: |
| 3881 | case URX_LOOP_SR_I: |
| 3882 | case URX_LOOP_DOT_I: |
| 3883 | case URX_LOOP_C: |
| 3884 | case URX_DOLLAR_D: |
| 3885 | case URX_DOLLAR_MD: |
| 3886 | case URX_BACKSLASH_H: |
| 3887 | case URX_BACKSLASH_R: |
| 3888 | case URX_BACKSLASH_V: |
| 3889 | // These instructions are unaltered by the relocation. |
| 3890 | fRXPat->fCompiledPat->setElementAt(op, dst); |
| 3891 | dst++; |
| 3892 | break; |
| 3893 | |
| 3894 | default: |
| 3895 | // Some op is unaccounted for. |
| 3896 | UPRV_UNREACHABLE_EXITabort(); |
| 3897 | } |
| 3898 | } |
| 3899 | |
| 3900 | fRXPat->fCompiledPat->setSize(dst); |
| 3901 | } |
| 3902 | |
| 3903 | |
| 3904 | |
| 3905 | |
| 3906 | //------------------------------------------------------------------------------ |
| 3907 | // |
| 3908 | // Error Report a rule parse error. |
| 3909 | // Only report it if no previous error has been recorded. |
| 3910 | // |
| 3911 | //------------------------------------------------------------------------------ |
| 3912 | void RegexCompile::error(UErrorCode e) { |
| 3913 | if (U_SUCCESS(*fStatus) || e == U_MEMORY_ALLOCATION_ERROR) { |
| 3914 | *fStatus = e; |
| 3915 | // Hmm. fParseErr (UParseError) line & offset fields are int32_t in public |
| 3916 | // API (see common/unicode/parseerr.h), while fLineNum and fCharNum are |
| 3917 | // int64_t. If the values of the latter are out of range for the former, |
| 3918 | // set them to the appropriate "field not supported" values. |
| 3919 | if (fLineNum > 0x7FFFFFFF) { |
| 3920 | fParseErr->line = 0; |
| 3921 | fParseErr->offset = -1; |
| 3922 | } else if (fCharNum > 0x7FFFFFFF) { |
| 3923 | fParseErr->line = (int32_t)fLineNum; |
| 3924 | fParseErr->offset = -1; |
| 3925 | } else { |
| 3926 | fParseErr->line = (int32_t)fLineNum; |
| 3927 | fParseErr->offset = (int32_t)fCharNum; |
| 3928 | } |
| 3929 | |
| 3930 | UErrorCode status = U_ZERO_ERROR; // throwaway status for extracting context |
| 3931 | |
| 3932 | // Fill in the context. |
| 3933 | // Note: extractBetween() pins supplied indices to the string bounds. |
| 3934 | uprv_memset(fParseErr->preContext, 0, sizeof(fParseErr->preContext)):: memset(fParseErr->preContext, 0, sizeof(fParseErr->preContext )); |
| 3935 | uprv_memset(fParseErr->postContext, 0, sizeof(fParseErr->postContext)):: memset(fParseErr->postContext, 0, sizeof(fParseErr-> postContext)); |
| 3936 | utext_extractutext_extract_71(fRXPat->fPattern, fScanIndex-U_PARSE_CONTEXT_LEN+1, fScanIndex, fParseErr->preContext, U_PARSE_CONTEXT_LEN, &status); |
| 3937 | utext_extractutext_extract_71(fRXPat->fPattern, fScanIndex, fScanIndex+U_PARSE_CONTEXT_LEN-1, fParseErr->postContext, U_PARSE_CONTEXT_LEN, &status); |
| 3938 | } |
| 3939 | } |
| 3940 | |
| 3941 | |
| 3942 | // |
| 3943 | // Assorted Unicode character constants. |
| 3944 | // Numeric because there is no portable way to enter them as literals. |
| 3945 | // (Think EBCDIC). |
| 3946 | // |
| 3947 | static const UChar chCR = 0x0d; // New lines, for terminating comments. |
| 3948 | static const UChar chLF = 0x0a; // Line Feed |
| 3949 | static const UChar chPound = 0x23; // '#', introduces a comment. |
| 3950 | static const UChar chDigit0 = 0x30; // '0' |
| 3951 | static const UChar chDigit7 = 0x37; // '9' |
| 3952 | static const UChar chColon = 0x3A; // ':' |
| 3953 | static const UChar chE = 0x45; // 'E' |
| 3954 | static const UChar chQ = 0x51; // 'Q' |
| 3955 | //static const UChar chN = 0x4E; // 'N' |
| 3956 | static const UChar chP = 0x50; // 'P' |
| 3957 | static const UChar chBackSlash = 0x5c; // '\' introduces a char escape |
| 3958 | //static const UChar chLBracket = 0x5b; // '[' |
| 3959 | static const UChar chRBracket = 0x5d; // ']' |
| 3960 | static const UChar chUp = 0x5e; // '^' |
| 3961 | static const UChar chLowerP = 0x70; |
| 3962 | static const UChar chLBrace = 0x7b; // '{' |
| 3963 | static const UChar chRBrace = 0x7d; // '}' |
| 3964 | static const UChar chNEL = 0x85; // NEL newline variant |
| 3965 | static const UChar chLS = 0x2028; // Unicode Line Separator |
| 3966 | |
| 3967 | |
| 3968 | //------------------------------------------------------------------------------ |
| 3969 | // |
| 3970 | // nextCharLL Low Level Next Char from the regex pattern. |
| 3971 | // Get a char from the string, keep track of input position |
| 3972 | // for error reporting. |
| 3973 | // |
| 3974 | //------------------------------------------------------------------------------ |
| 3975 | UChar32 RegexCompile::nextCharLL() { |
| 3976 | UChar32 ch; |
| 3977 | |
| 3978 | if (fPeekChar != -1) { |
| 3979 | ch = fPeekChar; |
| 3980 | fPeekChar = -1; |
| 3981 | return ch; |
| 3982 | } |
| 3983 | |
| 3984 | // assume we're already in the right place |
| 3985 | ch = UTEXT_NEXT32(fRXPat->fPattern)((fRXPat->fPattern)->chunkOffset < (fRXPat->fPattern )->chunkLength && ((fRXPat->fPattern)->chunkContents )[(fRXPat->fPattern)->chunkOffset]<0xd800 ? ((fRXPat ->fPattern)->chunkContents)[((fRXPat->fPattern)-> chunkOffset)++] : utext_next32_71(fRXPat->fPattern)); |
| 3986 | if (ch == U_SENTINEL(-1)) { |
| 3987 | return ch; |
| 3988 | } |
| 3989 | |
| 3990 | if (ch == chCR || |
| 3991 | ch == chNEL || |
| 3992 | ch == chLS || |
| 3993 | (ch == chLF && fLastChar != chCR)) { |
| 3994 | // Character is starting a new line. Bump up the line number, and |
| 3995 | // reset the column to 0. |
| 3996 | fLineNum++; |
| 3997 | fCharNum=0; |
| 3998 | } |
| 3999 | else { |
| 4000 | // Character is not starting a new line. Except in the case of a |
| 4001 | // LF following a CR, increment the column position. |
| 4002 | if (ch != chLF) { |
| 4003 | fCharNum++; |
| 4004 | } |
| 4005 | } |
| 4006 | fLastChar = ch; |
| 4007 | return ch; |
| 4008 | } |
| 4009 | |
| 4010 | //------------------------------------------------------------------------------ |
| 4011 | // |
| 4012 | // peekCharLL Low Level Character Scanning, sneak a peek at the next |
| 4013 | // character without actually getting it. |
| 4014 | // |
| 4015 | //------------------------------------------------------------------------------ |
| 4016 | UChar32 RegexCompile::peekCharLL() { |
| 4017 | if (fPeekChar == -1) { |
| 4018 | fPeekChar = nextCharLL(); |
| 4019 | } |
| 4020 | return fPeekChar; |
| 4021 | } |
| 4022 | |
| 4023 | |
| 4024 | //------------------------------------------------------------------------------ |
| 4025 | // |
| 4026 | // nextChar for pattern scanning. At this level, we handle stripping |
| 4027 | // out comments and processing some backslash character escapes. |
| 4028 | // The rest of the pattern grammar is handled at the next level up. |
| 4029 | // |
| 4030 | //------------------------------------------------------------------------------ |
| 4031 | void RegexCompile::nextChar(RegexPatternChar &c) { |
| 4032 | tailRecursion: |
| 4033 | fScanIndex = UTEXT_GETNATIVEINDEX(fRXPat->fPattern)((fRXPat->fPattern)->chunkOffset <= (fRXPat->fPattern )->nativeIndexingLimit? (fRXPat->fPattern)->chunkNativeStart +(fRXPat->fPattern)->chunkOffset : (fRXPat->fPattern )->pFuncs->mapOffsetToNative(fRXPat->fPattern)); |
| 4034 | c.fChar = nextCharLL(); |
| 4035 | c.fQuoted = FALSE0; |
| 4036 | |
| 4037 | if (fQuoteMode) { |
| 4038 | c.fQuoted = TRUE1; |
| 4039 | if ((c.fChar==chBackSlash && peekCharLL()==chE && ((fModeFlags & UREGEX_LITERAL) == 0)) || |
| 4040 | c.fChar == (UChar32)-1) { |
| 4041 | fQuoteMode = FALSE0; // Exit quote mode, |
| 4042 | nextCharLL(); // discard the E |
| 4043 | // nextChar(c); // recurse to get the real next char |
| 4044 | goto tailRecursion; // Note: fuzz testing produced testcases that |
| 4045 | // resulted in stack overflow here. |
| 4046 | } |
| 4047 | } |
| 4048 | else if (fInBackslashQuote) { |
| 4049 | // The current character immediately follows a '\' |
| 4050 | // Don't check for any further escapes, just return it as-is. |
| 4051 | // Don't set c.fQuoted, because that would prevent the state machine from |
| 4052 | // dispatching on the character. |
| 4053 | fInBackslashQuote = FALSE0; |
| 4054 | } |
| 4055 | else |
| 4056 | { |
| 4057 | // We are not in a \Q quoted region \E of the source. |
| 4058 | // |
| 4059 | if (fModeFlags & UREGEX_COMMENTS) { |
| 4060 | // |
| 4061 | // We are in free-spacing and comments mode. |
| 4062 | // Scan through any white space and comments, until we |
| 4063 | // reach a significant character or the end of input. |
| 4064 | for (;;) { |
| 4065 | if (c.fChar == (UChar32)-1) { |
| 4066 | break; // End of Input |
| 4067 | } |
| 4068 | if (c.fChar == chPound && fEOLComments == TRUE1) { |
| 4069 | // Start of a comment. Consume the rest of it, until EOF or a new line |
| 4070 | for (;;) { |
| 4071 | c.fChar = nextCharLL(); |
| 4072 | if (c.fChar == (UChar32)-1 || // EOF |
| 4073 | c.fChar == chCR || |
| 4074 | c.fChar == chLF || |
| 4075 | c.fChar == chNEL || |
| 4076 | c.fChar == chLS) { |
| 4077 | break; |
| 4078 | } |
| 4079 | } |
| 4080 | } |
| 4081 | // TODO: check what Java & Perl do with non-ASCII white spaces. Ticket 6061. |
| 4082 | if (PatternProps::isWhiteSpace(c.fChar) == FALSE0) { |
| 4083 | break; |
| 4084 | } |
| 4085 | c.fChar = nextCharLL(); |
| 4086 | } |
| 4087 | } |
| 4088 | |
| 4089 | // |
| 4090 | // check for backslash escaped characters. |
| 4091 | // |
| 4092 | if (c.fChar == chBackSlash) { |
| 4093 | int64_t pos = UTEXT_GETNATIVEINDEX(fRXPat->fPattern)((fRXPat->fPattern)->chunkOffset <= (fRXPat->fPattern )->nativeIndexingLimit? (fRXPat->fPattern)->chunkNativeStart +(fRXPat->fPattern)->chunkOffset : (fRXPat->fPattern )->pFuncs->mapOffsetToNative(fRXPat->fPattern)); |
| 4094 | if (RegexStaticSets::gStaticSets->fUnescapeCharSet.contains(peekCharLL())) { |
| 4095 | // |
| 4096 | // A '\' sequence that is handled by ICU's standard unescapeAt function. |
| 4097 | // Includes \uxxxx, \n, \r, many others. |
| 4098 | // Return the single equivalent character. |
| 4099 | // |
| 4100 | nextCharLL(); // get & discard the peeked char. |
| 4101 | c.fQuoted = TRUE1; |
| 4102 | |
| 4103 | if (UTEXT_FULL_TEXT_IN_CHUNK(fRXPat->fPattern, fPatternLength)((0==((fRXPat->fPattern)->chunkNativeStart))&&( (fPatternLength)==((fRXPat->fPattern)->chunkNativeLimit ))&&((fPatternLength)==((fRXPat->fPattern)->nativeIndexingLimit )))) { |
| 4104 | int32_t endIndex = (int32_t)pos; |
| 4105 | c.fChar = u_unescapeAtu_unescapeAt_71(uregex_ucstr_unescape_charAturegex_ucstr_unescape_charAt_71, &endIndex, (int32_t)fPatternLength, (void *)fRXPat->fPattern->chunkContents); |
| 4106 | |
| 4107 | if (endIndex == pos) { |
| 4108 | error(U_REGEX_BAD_ESCAPE_SEQUENCE); |
| 4109 | } |
| 4110 | fCharNum += endIndex - pos; |
| 4111 | UTEXT_SETNATIVEINDEX(fRXPat->fPattern, endIndex)do { int64_t __offset = (endIndex) - (fRXPat->fPattern)-> chunkNativeStart; if (__offset>=0 && __offset<( int64_t)(fRXPat->fPattern)->nativeIndexingLimit && (fRXPat->fPattern)->chunkContents[__offset]<0xdc00) { (fRXPat->fPattern)->chunkOffset=(int32_t)__offset; } else { utext_setNativeIndex_71((fRXPat->fPattern), (endIndex )); } } while (false); |
| 4112 | } else { |
| 4113 | int32_t offset = 0; |
| 4114 | struct URegexUTextUnescapeCharContext context = U_REGEX_UTEXT_UNESCAPE_CONTEXT(fRXPat->fPattern){ (fRXPat->fPattern), -1 }; |
| 4115 | |
| 4116 | UTEXT_SETNATIVEINDEX(fRXPat->fPattern, pos)do { int64_t __offset = (pos) - (fRXPat->fPattern)->chunkNativeStart ; if (__offset>=0 && __offset<(int64_t)(fRXPat-> fPattern)->nativeIndexingLimit && (fRXPat->fPattern )->chunkContents[__offset]<0xdc00) { (fRXPat->fPattern )->chunkOffset=(int32_t)__offset; } else { utext_setNativeIndex_71 ((fRXPat->fPattern), (pos)); } } while (false); |
| 4117 | c.fChar = u_unescapeAtu_unescapeAt_71(uregex_utext_unescape_charAturegex_utext_unescape_charAt_71, &offset, INT32_MAX(2147483647), &context); |
| 4118 | |
| 4119 | if (offset == 0) { |
| 4120 | error(U_REGEX_BAD_ESCAPE_SEQUENCE); |
| 4121 | } else if (context.lastOffset == offset) { |
| 4122 | UTEXT_PREVIOUS32(fRXPat->fPattern)((fRXPat->fPattern)->chunkOffset > 0 && (fRXPat ->fPattern)->chunkContents[(fRXPat->fPattern)->chunkOffset -1] < 0xd800 ? (fRXPat->fPattern)->chunkContents[--( (fRXPat->fPattern)->chunkOffset)] : utext_previous32_71 (fRXPat->fPattern)); |
| 4123 | } else if (context.lastOffset != offset-1) { |
| 4124 | utext_moveIndex32utext_moveIndex32_71(fRXPat->fPattern, offset - context.lastOffset - 1); |
| 4125 | } |
| 4126 | fCharNum += offset; |
| 4127 | } |
| 4128 | } |
| 4129 | else if (peekCharLL() == chDigit0) { |
| 4130 | // Octal Escape, using Java Regexp Conventions |
| 4131 | // which are \0 followed by 1-3 octal digits. |
| 4132 | // Different from ICU Unescape handling of Octal, which does not |
| 4133 | // require the leading 0. |
| 4134 | // Java also has the convention of only consuming 2 octal digits if |
| 4135 | // the three digit number would be > 0xff |
| 4136 | // |
| 4137 | c.fChar = 0; |
| 4138 | nextCharLL(); // Consume the initial 0. |
| 4139 | int index; |
| 4140 | for (index=0; index<3; index++) { |
| 4141 | int32_t ch = peekCharLL(); |
| 4142 | if (ch<chDigit0 || ch>chDigit7) { |
| 4143 | if (index==0) { |
| 4144 | // \0 is not followed by any octal digits. |
| 4145 | error(U_REGEX_BAD_ESCAPE_SEQUENCE); |
| 4146 | } |
| 4147 | break; |
| 4148 | } |
| 4149 | c.fChar <<= 3; |
| 4150 | c.fChar += ch&7; |
| 4151 | if (c.fChar <= 255) { |
| 4152 | nextCharLL(); |
| 4153 | } else { |
| 4154 | // The last digit made the number too big. Forget we saw it. |
| 4155 | c.fChar >>= 3; |
| 4156 | } |
| 4157 | } |
| 4158 | c.fQuoted = TRUE1; |
| 4159 | } |
| 4160 | else if (peekCharLL() == chQ) { |
| 4161 | // "\Q" enter quote mode, which will continue until "\E" |
| 4162 | fQuoteMode = TRUE1; |
| 4163 | nextCharLL(); // discard the 'Q'. |
| 4164 | // nextChar(c); // recurse to get the real next char. |
| 4165 | goto tailRecursion; // Note: fuzz testing produced test cases that |
| 4166 | // resulted in stack overflow here. |
| 4167 | } |
| 4168 | else |
| 4169 | { |
| 4170 | // We are in a '\' escape that will be handled by the state table scanner. |
| 4171 | // Just return the backslash, but remember that the following char is to |
| 4172 | // be taken literally. |
| 4173 | fInBackslashQuote = TRUE1; |
| 4174 | } |
| 4175 | } |
| 4176 | } |
| 4177 | |
| 4178 | // re-enable # to end-of-line comments, in case they were disabled. |
| 4179 | // They are disabled by the parser upon seeing '(?', but this lasts for |
| 4180 | // the fetching of the next character only. |
| 4181 | fEOLComments = TRUE1; |
| 4182 | |
| 4183 | // putc(c.fChar, stdout); |
| 4184 | } |
| 4185 | |
| 4186 | |
| 4187 | |
| 4188 | //------------------------------------------------------------------------------ |
| 4189 | // |
| 4190 | // scanNamedChar |
| 4191 | // Get a UChar32 from a \N{UNICODE CHARACTER NAME} in the pattern. |
| 4192 | // |
| 4193 | // The scan position will be at the 'N'. On return |
| 4194 | // the scan position should be just after the '}' |
| 4195 | // |
| 4196 | // Return the UChar32 |
| 4197 | // |
| 4198 | //------------------------------------------------------------------------------ |
| 4199 | UChar32 RegexCompile::scanNamedChar() { |
| 4200 | if (U_FAILURE(*fStatus)) { |
| 4201 | return 0; |
| 4202 | } |
| 4203 | |
| 4204 | nextChar(fC); |
| 4205 | if (fC.fChar != chLBrace) { |
| 4206 | error(U_REGEX_PROPERTY_SYNTAX); |
| 4207 | return 0; |
| 4208 | } |
| 4209 | |
| 4210 | UnicodeString charName; |
| 4211 | for (;;) { |
| 4212 | nextChar(fC); |
| 4213 | if (fC.fChar == chRBrace) { |
| 4214 | break; |
| 4215 | } |
| 4216 | if (fC.fChar == -1) { |
| 4217 | error(U_REGEX_PROPERTY_SYNTAX); |
| 4218 | return 0; |
| 4219 | } |
| 4220 | charName.append(fC.fChar); |
| 4221 | } |
| 4222 | |
| 4223 | char name[100]; |
| 4224 | if (!uprv_isInvariantUStringuprv_isInvariantUString_71(charName.getBuffer(), charName.length()) || |
| 4225 | (uint32_t)charName.length()>=sizeof(name)) { |
| 4226 | // All Unicode character names have only invariant characters. |
| 4227 | // The API to get a character, given a name, accepts only char *, forcing us to convert, |
| 4228 | // which requires this error check |
| 4229 | error(U_REGEX_PROPERTY_SYNTAX); |
| 4230 | return 0; |
| 4231 | } |
| 4232 | charName.extract(0, charName.length(), name, sizeof(name), US_INVicu::UnicodeString::kInvariant); |
| 4233 | |
| 4234 | UChar32 theChar = u_charFromNameu_charFromName_71(U_UNICODE_CHAR_NAME, name, fStatus); |
| 4235 | if (U_FAILURE(*fStatus)) { |
| 4236 | error(U_REGEX_PROPERTY_SYNTAX); |
| 4237 | } |
| 4238 | |
| 4239 | nextChar(fC); // Continue overall regex pattern processing with char after the '}' |
| 4240 | return theChar; |
| 4241 | } |
| 4242 | |
| 4243 | //------------------------------------------------------------------------------ |
| 4244 | // |
| 4245 | // scanProp Construct a UnicodeSet from the text at the current scan |
| 4246 | // position, which will be of the form \p{whaterver} |
| 4247 | // |
| 4248 | // The scan position will be at the 'p' or 'P'. On return |
| 4249 | // the scan position should be just after the '}' |
| 4250 | // |
| 4251 | // Return a UnicodeSet, constructed from the \P pattern, |
| 4252 | // or NULL if the pattern is invalid. |
| 4253 | // |
| 4254 | //------------------------------------------------------------------------------ |
| 4255 | UnicodeSet *RegexCompile::scanProp() { |
| 4256 | UnicodeSet *uset = NULL__null; |
| 4257 | |
| 4258 | if (U_FAILURE(*fStatus)) { |
| 4259 | return NULL__null; |
| 4260 | } |
| 4261 | (void)chLowerP; // Suppress compiler unused variable warning. |
| 4262 | U_ASSERT(fC.fChar == chLowerP || fC.fChar == chP)(void)0; |
| 4263 | UBool negated = (fC.fChar == chP); |
| 4264 | |
| 4265 | UnicodeString propertyName; |
| 4266 | nextChar(fC); |
| 4267 | if (fC.fChar != chLBrace) { |
| 4268 | error(U_REGEX_PROPERTY_SYNTAX); |
| 4269 | return NULL__null; |
| 4270 | } |
| 4271 | for (;;) { |
| 4272 | nextChar(fC); |
| 4273 | if (fC.fChar == chRBrace) { |
| 4274 | break; |
| 4275 | } |
| 4276 | if (fC.fChar == -1) { |
| 4277 | // Hit the end of the input string without finding the closing '}' |
| 4278 | error(U_REGEX_PROPERTY_SYNTAX); |
| 4279 | return NULL__null; |
| 4280 | } |
| 4281 | propertyName.append(fC.fChar); |
| 4282 | } |
| 4283 | uset = createSetForProperty(propertyName, negated); |
| 4284 | nextChar(fC); // Move input scan to position following the closing '}' |
| 4285 | return uset; |
| 4286 | } |
| 4287 | |
| 4288 | //------------------------------------------------------------------------------ |
| 4289 | // |
| 4290 | // scanPosixProp Construct a UnicodeSet from the text at the current scan |
| 4291 | // position, which is expected be of the form [:property expression:] |
| 4292 | // |
| 4293 | // The scan position will be at the opening ':'. On return |
| 4294 | // the scan position must be on the closing ']' |
| 4295 | // |
| 4296 | // Return a UnicodeSet constructed from the pattern, |
| 4297 | // or NULL if this is not a valid POSIX-style set expression. |
| 4298 | // If not a property expression, restore the initial scan position |
| 4299 | // (to the opening ':') |
| 4300 | // |
| 4301 | // Note: the opening '[:' is not sufficient to guarantee that |
| 4302 | // this is a [:property:] expression. |
| 4303 | // [:'+=,] is a perfectly good ordinary set expression that |
| 4304 | // happens to include ':' as one of its characters. |
| 4305 | // |
| 4306 | //------------------------------------------------------------------------------ |
| 4307 | UnicodeSet *RegexCompile::scanPosixProp() { |
| 4308 | UnicodeSet *uset = NULL__null; |
| 4309 | |
| 4310 | if (U_FAILURE(*fStatus)) { |
| 4311 | return NULL__null; |
| 4312 | } |
| 4313 | |
| 4314 | U_ASSERT(fC.fChar == chColon)(void)0; |
| 4315 | |
| 4316 | // Save the scanner state. |
| 4317 | // TODO: move this into the scanner, with the state encapsulated in some way. Ticket 6062 |
| 4318 | int64_t savedScanIndex = fScanIndex; |
| 4319 | int64_t savedNextIndex = UTEXT_GETNATIVEINDEX(fRXPat->fPattern)((fRXPat->fPattern)->chunkOffset <= (fRXPat->fPattern )->nativeIndexingLimit? (fRXPat->fPattern)->chunkNativeStart +(fRXPat->fPattern)->chunkOffset : (fRXPat->fPattern )->pFuncs->mapOffsetToNative(fRXPat->fPattern)); |
| 4320 | UBool savedQuoteMode = fQuoteMode; |
| 4321 | UBool savedInBackslashQuote = fInBackslashQuote; |
| 4322 | UBool savedEOLComments = fEOLComments; |
| 4323 | int64_t savedLineNum = fLineNum; |
| 4324 | int64_t savedCharNum = fCharNum; |
| 4325 | UChar32 savedLastChar = fLastChar; |
| 4326 | UChar32 savedPeekChar = fPeekChar; |
| 4327 | RegexPatternChar savedfC = fC; |
| 4328 | |
| 4329 | // Scan for a closing ]. A little tricky because there are some perverse |
| 4330 | // edge cases possible. "[:abc\Qdef:] \E]" is a valid non-property expression, |
| 4331 | // ending on the second closing ]. |
| 4332 | |
| 4333 | UnicodeString propName; |
| 4334 | UBool negated = FALSE0; |
| 4335 | |
| 4336 | // Check for and consume the '^' in a negated POSIX property, e.g. [:^Letter:] |
| 4337 | nextChar(fC); |
| 4338 | if (fC.fChar == chUp) { |
| 4339 | negated = TRUE1; |
| 4340 | nextChar(fC); |
| 4341 | } |
| 4342 | |
| 4343 | // Scan for the closing ":]", collecting the property name along the way. |
| 4344 | UBool sawPropSetTerminator = FALSE0; |
| 4345 | for (;;) { |
| 4346 | propName.append(fC.fChar); |
| 4347 | nextChar(fC); |
| 4348 | if (fC.fQuoted || fC.fChar == -1) { |
| 4349 | // Escaped characters or end of input - either says this isn't a [:Property:] |
| 4350 | break; |
| 4351 | } |
| 4352 | if (fC.fChar == chColon) { |
| 4353 | nextChar(fC); |
| 4354 | if (fC.fChar == chRBracket) { |
| 4355 | sawPropSetTerminator = TRUE1; |
| 4356 | } |
| 4357 | break; |
| 4358 | } |
| 4359 | } |
| 4360 | |
| 4361 | if (sawPropSetTerminator) { |
| 4362 | uset = createSetForProperty(propName, negated); |
| 4363 | } |
| 4364 | else |
| 4365 | { |
| 4366 | // No closing ":]". |
| 4367 | // Restore the original scan position. |
| 4368 | // The main scanner will retry the input as a normal set expression, |
| 4369 | // not a [:Property:] expression. |
| 4370 | fScanIndex = savedScanIndex; |
| 4371 | fQuoteMode = savedQuoteMode; |
| 4372 | fInBackslashQuote = savedInBackslashQuote; |
| 4373 | fEOLComments = savedEOLComments; |
| 4374 | fLineNum = savedLineNum; |
| 4375 | fCharNum = savedCharNum; |
| 4376 | fLastChar = savedLastChar; |
| 4377 | fPeekChar = savedPeekChar; |
| 4378 | fC = savedfC; |
| 4379 | UTEXT_SETNATIVEINDEX(fRXPat->fPattern, savedNextIndex)do { int64_t __offset = (savedNextIndex) - (fRXPat->fPattern )->chunkNativeStart; if (__offset>=0 && __offset <(int64_t)(fRXPat->fPattern)->nativeIndexingLimit && (fRXPat->fPattern)->chunkContents[__offset]<0xdc00) { (fRXPat->fPattern)->chunkOffset=(int32_t)__offset; } else { utext_setNativeIndex_71((fRXPat->fPattern), (savedNextIndex )); } } while (false); |
| 4380 | } |
| 4381 | return uset; |
| 4382 | } |
| 4383 | |
| 4384 | static inline void addIdentifierIgnorable(UnicodeSet *set, UErrorCode& ec) { |
| 4385 | set->add(0, 8).add(0x0e, 0x1b).add(0x7f, 0x9f); |
| 4386 | addCategory(set, U_GC_CF_MASK((uint32_t)1<<(U_FORMAT_CHAR)), ec); |
| 4387 | } |
| 4388 | |
| 4389 | // |
| 4390 | // Create a Unicode Set from a Unicode Property expression. |
| 4391 | // This is common code underlying both \p{...} and [:...:] expressions. |
| 4392 | // Includes trying the Java "properties" that aren't supported as |
| 4393 | // normal ICU UnicodeSet properties |
| 4394 | // |
| 4395 | UnicodeSet *RegexCompile::createSetForProperty(const UnicodeString &propName, UBool negated) { |
| 4396 | |
| 4397 | if (U_FAILURE(*fStatus)) { |
| 4398 | return nullptr; |
| 4399 | } |
| 4400 | LocalPointer<UnicodeSet> set; |
| 4401 | UErrorCode status = U_ZERO_ERROR; |
| 4402 | |
| 4403 | do { // non-loop, exists to allow breaks from the block. |
| 4404 | // |
| 4405 | // First try the property as we received it |
| 4406 | // |
| 4407 | UnicodeString setExpr; |
| 4408 | uint32_t usetFlags = 0; |
| 4409 | setExpr.append(u"[\\p{", -1); |
| 4410 | setExpr.append(propName); |
| 4411 | setExpr.append(u"}]", -1); |
| 4412 | if (fModeFlags & UREGEX_CASE_INSENSITIVE) { |
| 4413 | usetFlags |= USET_CASE_INSENSITIVE; |
| 4414 | } |
| 4415 | set.adoptInsteadAndCheckErrorCode(new UnicodeSet(setExpr, usetFlags, NULL__null, status), status); |
| 4416 | if (U_SUCCESS(status) || status == U_MEMORY_ALLOCATION_ERROR) { |
| 4417 | break; |
| 4418 | } |
| 4419 | |
| 4420 | // |
| 4421 | // The incoming property wasn't directly recognized by ICU. |
| 4422 | |
| 4423 | // Check [:word:] and [:all:]. These are not recognized as a properties by ICU UnicodeSet. |
| 4424 | // Java accepts 'word' with mixed case. |
| 4425 | // Java accepts 'all' only in all lower case. |
| 4426 | |
| 4427 | status = U_ZERO_ERROR; |
| 4428 | if (propName.caseCompare(u"word", -1, 0) == 0) { |
| 4429 | set.adoptInsteadAndCheckErrorCode( |
| 4430 | RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET].cloneAsThawed(), status); |
| 4431 | break; |
| 4432 | } |
| 4433 | if (propName.compare(u"all", -1) == 0) { |
| 4434 | set.adoptInsteadAndCheckErrorCode(new UnicodeSet(0, 0x10ffff), status); |
| 4435 | break; |
| 4436 | } |
| 4437 | |
| 4438 | |
| 4439 | // Do Java InBlock expressions |
| 4440 | // |
| 4441 | UnicodeString mPropName = propName; |
| 4442 | if (mPropName.startsWith(u"In", 2) && mPropName.length() >= 3) { |
| 4443 | status = U_ZERO_ERROR; |
| 4444 | set.adoptInsteadAndCheckErrorCode(new UnicodeSet(), status); |
| 4445 | if (U_FAILURE(status)) { |
| 4446 | break; |
| 4447 | } |
| 4448 | UnicodeString blockName(mPropName, 2); // Property with the leading "In" removed. |
| 4449 | set->applyPropertyAlias(UnicodeString(u"Block"), blockName, status); |
| 4450 | break; |
| 4451 | } |
| 4452 | |
| 4453 | // Check for the Java form "IsBooleanPropertyValue", which we will recast |
| 4454 | // as "BooleanPropertyValue". The property value can be either a |
| 4455 | // a General Category or a Script Name. |
| 4456 | |
| 4457 | if (propName.startsWith(u"Is", 2) && propName.length()>=3) { |
| 4458 | mPropName.remove(0, 2); // Strip the "Is" |
| 4459 | if (mPropName.indexOf(u'=') >= 0) { |
| 4460 | // Reject any "Is..." property expression containing an '=', that is, |
| 4461 | // any non-binary property expression. |
| 4462 | status = U_REGEX_PROPERTY_SYNTAX; |
| 4463 | break; |
| 4464 | } |
| 4465 | |
| 4466 | if (mPropName.caseCompare(u"assigned", -1, 0) == 0) { |
| 4467 | mPropName.setTo(u"unassigned", -1); |
| 4468 | negated = !negated; |
| 4469 | } else if (mPropName.caseCompare(u"TitleCase", -1, 0) == 0) { |
| 4470 | mPropName.setTo(u"Titlecase_Letter", -1); |
| 4471 | } |
| 4472 | |
| 4473 | mPropName.insert(0, u"[\\p{", -1); |
| 4474 | mPropName.append(u"}]", -1); |
| 4475 | set.adoptInsteadAndCheckErrorCode(new UnicodeSet(mPropName, *fStatus), status); |
| 4476 | |
| 4477 | if (U_SUCCESS(status) && !set->isEmpty() && (usetFlags & USET_CASE_INSENSITIVE)) { |
| 4478 | set->closeOver(USET_CASE_INSENSITIVE); |
| 4479 | } |
| 4480 | break; |
| 4481 | |
| 4482 | } |
| 4483 | |
| 4484 | if (propName.startsWith(u"java", -1)) { |
| 4485 | status = U_ZERO_ERROR; |
| 4486 | set.adoptInsteadAndCheckErrorCode(new UnicodeSet(), status); |
| 4487 | if (U_FAILURE(status)) { |
| 4488 | break; |
| 4489 | } |
| 4490 | // |
| 4491 | // Try the various Java specific properties. |
| 4492 | // These all begin with "java" |
| 4493 | // |
| 4494 | if (propName.compare(u"javaDefined", -1) == 0) { |
| 4495 | addCategory(set.getAlias(), U_GC_CN_MASK((uint32_t)1<<(U_GENERAL_OTHER_TYPES)), status); |
| 4496 | set->complement(); |
| 4497 | } |
| 4498 | else if (propName.compare(u"javaDigit", -1) == 0) { |
| 4499 | addCategory(set.getAlias(), U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), status); |
| 4500 | } |
| 4501 | else if (propName.compare(u"javaIdentifierIgnorable", -1) == 0) { |
| 4502 | addIdentifierIgnorable(set.getAlias(), status); |
| 4503 | } |
| 4504 | else if (propName.compare(u"javaISOControl", -1) == 0) { |
| 4505 | set->add(0, 0x1F).add(0x7F, 0x9F); |
| 4506 | } |
| 4507 | else if (propName.compare(u"javaJavaIdentifierPart", -1) == 0) { |
| 4508 | addCategory(set.getAlias(), U_GC_L_MASK(((uint32_t)1<<(U_UPPERCASE_LETTER))|((uint32_t)1<< (U_LOWERCASE_LETTER))|((uint32_t)1<<(U_TITLECASE_LETTER ))|((uint32_t)1<<(U_MODIFIER_LETTER))|((uint32_t)1<< (U_OTHER_LETTER))), status); |
| 4509 | addCategory(set.getAlias(), U_GC_SC_MASK((uint32_t)1<<(U_CURRENCY_SYMBOL)), status); |
| 4510 | addCategory(set.getAlias(), U_GC_PC_MASK((uint32_t)1<<(U_CONNECTOR_PUNCTUATION)), status); |
| 4511 | addCategory(set.getAlias(), U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), status); |
| 4512 | addCategory(set.getAlias(), U_GC_NL_MASK((uint32_t)1<<(U_LETTER_NUMBER)), status); |
| 4513 | addCategory(set.getAlias(), U_GC_MC_MASK((uint32_t)1<<(U_COMBINING_SPACING_MARK)), status); |
| 4514 | addCategory(set.getAlias(), U_GC_MN_MASK((uint32_t)1<<(U_NON_SPACING_MARK)), status); |
| 4515 | addIdentifierIgnorable(set.getAlias(), status); |
| 4516 | } |
| 4517 | else if (propName.compare(u"javaJavaIdentifierStart", -1) == 0) { |
| 4518 | addCategory(set.getAlias(), U_GC_L_MASK(((uint32_t)1<<(U_UPPERCASE_LETTER))|((uint32_t)1<< (U_LOWERCASE_LETTER))|((uint32_t)1<<(U_TITLECASE_LETTER ))|((uint32_t)1<<(U_MODIFIER_LETTER))|((uint32_t)1<< (U_OTHER_LETTER))), status); |
| 4519 | addCategory(set.getAlias(), U_GC_NL_MASK((uint32_t)1<<(U_LETTER_NUMBER)), status); |
| 4520 | addCategory(set.getAlias(), U_GC_SC_MASK((uint32_t)1<<(U_CURRENCY_SYMBOL)), status); |
| 4521 | addCategory(set.getAlias(), U_GC_PC_MASK((uint32_t)1<<(U_CONNECTOR_PUNCTUATION)), status); |
| 4522 | } |
| 4523 | else if (propName.compare(u"javaLetter", -1) == 0) { |
| 4524 | addCategory(set.getAlias(), U_GC_L_MASK(((uint32_t)1<<(U_UPPERCASE_LETTER))|((uint32_t)1<< (U_LOWERCASE_LETTER))|((uint32_t)1<<(U_TITLECASE_LETTER ))|((uint32_t)1<<(U_MODIFIER_LETTER))|((uint32_t)1<< (U_OTHER_LETTER))), status); |
| 4525 | } |
| 4526 | else if (propName.compare(u"javaLetterOrDigit", -1) == 0) { |
| 4527 | addCategory(set.getAlias(), U_GC_L_MASK(((uint32_t)1<<(U_UPPERCASE_LETTER))|((uint32_t)1<< (U_LOWERCASE_LETTER))|((uint32_t)1<<(U_TITLECASE_LETTER ))|((uint32_t)1<<(U_MODIFIER_LETTER))|((uint32_t)1<< (U_OTHER_LETTER))), status); |
| 4528 | addCategory(set.getAlias(), U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), status); |
| 4529 | } |
| 4530 | else if (propName.compare(u"javaLowerCase", -1) == 0) { |
| 4531 | addCategory(set.getAlias(), U_GC_LL_MASK((uint32_t)1<<(U_LOWERCASE_LETTER)), status); |
| 4532 | } |
| 4533 | else if (propName.compare(u"javaMirrored", -1) == 0) { |
| 4534 | set->applyIntPropertyValue(UCHAR_BIDI_MIRRORED, 1, status); |
| 4535 | } |
| 4536 | else if (propName.compare(u"javaSpaceChar", -1) == 0) { |
| 4537 | addCategory(set.getAlias(), U_GC_Z_MASK(((uint32_t)1<<(U_SPACE_SEPARATOR))|((uint32_t)1<< (U_LINE_SEPARATOR))|((uint32_t)1<<(U_PARAGRAPH_SEPARATOR ))), status); |
| 4538 | } |
| 4539 | else if (propName.compare(u"javaSupplementaryCodePoint", -1) == 0) { |
| 4540 | set->add(0x10000, UnicodeSet::MAX_VALUE); |
| 4541 | } |
| 4542 | else if (propName.compare(u"javaTitleCase", -1) == 0) { |
| 4543 | addCategory(set.getAlias(), U_GC_LT_MASK((uint32_t)1<<(U_TITLECASE_LETTER)), status); |
| 4544 | } |
| 4545 | else if (propName.compare(u"javaUnicodeIdentifierStart", -1) == 0) { |
| 4546 | addCategory(set.getAlias(), U_GC_L_MASK(((uint32_t)1<<(U_UPPERCASE_LETTER))|((uint32_t)1<< (U_LOWERCASE_LETTER))|((uint32_t)1<<(U_TITLECASE_LETTER ))|((uint32_t)1<<(U_MODIFIER_LETTER))|((uint32_t)1<< (U_OTHER_LETTER))), status); |
| 4547 | addCategory(set.getAlias(), U_GC_NL_MASK((uint32_t)1<<(U_LETTER_NUMBER)), status); |
| 4548 | } |
| 4549 | else if (propName.compare(u"javaUnicodeIdentifierPart", -1) == 0) { |
| 4550 | addCategory(set.getAlias(), U_GC_L_MASK(((uint32_t)1<<(U_UPPERCASE_LETTER))|((uint32_t)1<< (U_LOWERCASE_LETTER))|((uint32_t)1<<(U_TITLECASE_LETTER ))|((uint32_t)1<<(U_MODIFIER_LETTER))|((uint32_t)1<< (U_OTHER_LETTER))), status); |
| 4551 | addCategory(set.getAlias(), U_GC_PC_MASK((uint32_t)1<<(U_CONNECTOR_PUNCTUATION)), status); |
| 4552 | addCategory(set.getAlias(), U_GC_ND_MASK((uint32_t)1<<(U_DECIMAL_DIGIT_NUMBER)), status); |
| 4553 | addCategory(set.getAlias(), U_GC_NL_MASK((uint32_t)1<<(U_LETTER_NUMBER)), status); |
| 4554 | addCategory(set.getAlias(), U_GC_MC_MASK((uint32_t)1<<(U_COMBINING_SPACING_MARK)), status); |
| 4555 | addCategory(set.getAlias(), U_GC_MN_MASK((uint32_t)1<<(U_NON_SPACING_MARK)), status); |
| 4556 | addIdentifierIgnorable(set.getAlias(), status); |
| 4557 | } |
| 4558 | else if (propName.compare(u"javaUpperCase", -1) == 0) { |
| 4559 | addCategory(set.getAlias(), U_GC_LU_MASK((uint32_t)1<<(U_UPPERCASE_LETTER)), status); |
| 4560 | } |
| 4561 | else if (propName.compare(u"javaValidCodePoint", -1) == 0) { |
| 4562 | set->add(0, UnicodeSet::MAX_VALUE); |
| 4563 | } |
| 4564 | else if (propName.compare(u"javaWhitespace", -1) == 0) { |
| 4565 | addCategory(set.getAlias(), U_GC_Z_MASK(((uint32_t)1<<(U_SPACE_SEPARATOR))|((uint32_t)1<< (U_LINE_SEPARATOR))|((uint32_t)1<<(U_PARAGRAPH_SEPARATOR ))), status); |
| 4566 | set->removeAll(UnicodeSet().add(0xa0).add(0x2007).add(0x202f)); |
| 4567 | set->add(9, 0x0d).add(0x1c, 0x1f); |
| 4568 | } else { |
| 4569 | status = U_REGEX_PROPERTY_SYNTAX; |
| 4570 | } |
| 4571 | |
| 4572 | if (U_SUCCESS(status) && !set->isEmpty() && (usetFlags & USET_CASE_INSENSITIVE)) { |
| 4573 | set->closeOver(USET_CASE_INSENSITIVE); |
| 4574 | } |
| 4575 | break; |
| 4576 | } |
| 4577 | |
| 4578 | // Unrecognized property. ICU didn't like it as it was, and none of the Java compatibility |
| 4579 | // extensions matched it. |
| 4580 | status = U_REGEX_PROPERTY_SYNTAX; |
| 4581 | } while (false); // End of do loop block. Code above breaks out of the block on success or hard failure. |
| 4582 | |
| 4583 | if (U_SUCCESS(status)) { |
| 4584 | // ICU 70 adds emoji properties of strings, but as long as Java does not say how to |
| 4585 | // deal with properties of strings and character classes with strings, we ignore them. |
| 4586 | // Just in case something downstream might stumble over the strings, |
| 4587 | // we remove them from the set. |
| 4588 | // Note that when we support strings, the complement of a property (as with \P) |
| 4589 | // should be implemented as .complement().removeAllStrings() (code point complement). |
| 4590 | set->removeAllStrings(); |
| 4591 | U_ASSERT(set.isValid())(void)0; |
| 4592 | if (negated) { |
| 4593 | set->complement(); |
| 4594 | } |
| 4595 | return set.orphan(); |
| 4596 | } else { |
| 4597 | if (status == U_ILLEGAL_ARGUMENT_ERROR) { |
| 4598 | status = U_REGEX_PROPERTY_SYNTAX; |
| 4599 | } |
| 4600 | error(status); |
| 4601 | return nullptr; |
| 4602 | } |
| 4603 | } |
| 4604 | |
| 4605 | |
| 4606 | // |
| 4607 | // SetEval Part of the evaluation of [set expressions]. |
| 4608 | // Perform any pending (stacked) operations with precedence |
| 4609 | // equal or greater to that of the next operator encountered |
| 4610 | // in the expression. |
| 4611 | // |
| 4612 | void RegexCompile::setEval(int32_t nextOp) { |
| 4613 | UnicodeSet *rightOperand = NULL__null; |
| 4614 | UnicodeSet *leftOperand = NULL__null; |
| 4615 | for (;;) { |
| 4616 | U_ASSERT(fSetOpStack.empty()==FALSE)(void)0; |
| 4617 | int32_t pendingSetOperation = fSetOpStack.peeki(); |
| 4618 | if ((pendingSetOperation&0xffff0000) < (nextOp&0xffff0000)) { |
| 4619 | break; |
| 4620 | } |
| 4621 | fSetOpStack.popi(); |
| 4622 | U_ASSERT(fSetStack.empty() == FALSE)(void)0; |
| 4623 | rightOperand = (UnicodeSet *)fSetStack.peek(); |
| 4624 | // ICU 70 adds emoji properties of strings, but createSetForProperty() removes all strings |
| 4625 | // (see comments there). |
| 4626 | // We also do not yet support string literals in character classes, |
| 4627 | // so there should not be any strings. |
| 4628 | // Note that when we support strings, the complement of a set (as with ^ or \P) |
| 4629 | // should be implemented as .complement().removeAllStrings() (code point complement). |
| 4630 | U_ASSERT(!rightOperand->hasStrings())(void)0; |
| 4631 | switch (pendingSetOperation) { |
| 4632 | case setNegation: |
| 4633 | rightOperand->complement(); |
| 4634 | break; |
| 4635 | case setCaseClose: |
| 4636 | // TODO: need a simple close function. Ticket 6065 |
| 4637 | rightOperand->closeOver(USET_CASE_INSENSITIVE); |
| 4638 | rightOperand->removeAllStrings(); |
| 4639 | break; |
| 4640 | case setDifference1: |
| 4641 | case setDifference2: |
| 4642 | fSetStack.pop(); |
| 4643 | leftOperand = (UnicodeSet *)fSetStack.peek(); |
| 4644 | leftOperand->removeAll(*rightOperand); |
| 4645 | delete rightOperand; |
| 4646 | break; |
| 4647 | case setIntersection1: |
| 4648 | case setIntersection2: |
| 4649 | fSetStack.pop(); |
| 4650 | leftOperand = (UnicodeSet *)fSetStack.peek(); |
| 4651 | leftOperand->retainAll(*rightOperand); |
| 4652 | delete rightOperand; |
| 4653 | break; |
| 4654 | case setUnion: |
| 4655 | fSetStack.pop(); |
| 4656 | leftOperand = (UnicodeSet *)fSetStack.peek(); |
| 4657 | leftOperand->addAll(*rightOperand); |
| 4658 | delete rightOperand; |
| 4659 | break; |
| 4660 | default: |
| 4661 | UPRV_UNREACHABLE_EXITabort(); |
| 4662 | } |
| 4663 | } |
| 4664 | } |
| 4665 | |
| 4666 | void RegexCompile::setPushOp(int32_t op) { |
| 4667 | setEval(op); |
| 4668 | fSetOpStack.push(op, *fStatus); |
| 4669 | LocalPointer<UnicodeSet> lpSet(new UnicodeSet(), *fStatus); |
| 4670 | fSetStack.push(lpSet.orphan(), *fStatus); |
| 4671 | } |
| 4672 | |
| 4673 | U_NAMESPACE_END} |
| 4674 | #endif // !UCONFIG_NO_REGULAR_EXPRESSIONS |
| 4675 |