1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/icu/source/i18n/rematch.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,5690 @@
1.4 +/*
1.5 +**************************************************************************
1.6 +* Copyright (C) 2002-2013 International Business Machines Corporation *
1.7 +* and others. All rights reserved. *
1.8 +**************************************************************************
1.9 +*/
1.10 +//
1.11 +// file: rematch.cpp
1.12 +//
1.13 +// Contains the implementation of class RegexMatcher,
1.14 +// which is one of the main API classes for the ICU regular expression package.
1.15 +//
1.16 +
1.17 +#include "unicode/utypes.h"
1.18 +#if !UCONFIG_NO_REGULAR_EXPRESSIONS
1.19 +
1.20 +#include "unicode/regex.h"
1.21 +#include "unicode/uniset.h"
1.22 +#include "unicode/uchar.h"
1.23 +#include "unicode/ustring.h"
1.24 +#include "unicode/rbbi.h"
1.25 +#include "unicode/utf.h"
1.26 +#include "unicode/utf16.h"
1.27 +#include "uassert.h"
1.28 +#include "cmemory.h"
1.29 +#include "uvector.h"
1.30 +#include "uvectr32.h"
1.31 +#include "uvectr64.h"
1.32 +#include "regeximp.h"
1.33 +#include "regexst.h"
1.34 +#include "regextxt.h"
1.35 +#include "ucase.h"
1.36 +
1.37 +// #include <malloc.h> // Needed for heapcheck testing
1.38 +
1.39 +
1.40 +// Find progress callback
1.41 +// ----------------------
1.42 +// Macro to inline test & call to ReportFindProgress(). Eliminates unnecessary function call.
1.43 +//
1.44 +#define REGEXFINDPROGRESS_INTERRUPT(pos, status) \
1.45 + (fFindProgressCallbackFn != NULL) && (ReportFindProgress(pos, status) == FALSE)
1.46 +
1.47 +
1.48 +// Smart Backtracking
1.49 +// ------------------
1.50 +// When a failure would go back to a LOOP_C instruction,
1.51 +// strings, characters, and setrefs scan backwards for a valid start
1.52 +// character themselves, pop the stack, and save state, emulating the
1.53 +// LOOP_C's effect but assured that the next character of input is a
1.54 +// possible matching character.
1.55 +//
1.56 +// Good idea in theory; unfortunately it only helps out a few specific
1.57 +// cases and slows the engine down a little in the rest.
1.58 +
1.59 +U_NAMESPACE_BEGIN
1.60 +
1.61 +// Default limit for the size of the back track stack, to avoid system
1.62 +// failures causedby heap exhaustion. Units are in 32 bit words, not bytes.
1.63 +// This value puts ICU's limits higher than most other regexp implementations,
1.64 +// which use recursion rather than the heap, and take more storage per
1.65 +// backtrack point.
1.66 +//
1.67 +static const int32_t DEFAULT_BACKTRACK_STACK_CAPACITY = 8000000;
1.68 +
1.69 +// Time limit counter constant.
1.70 +// Time limits for expression evaluation are in terms of quanta of work by
1.71 +// the engine, each of which is 10,000 state saves.
1.72 +// This constant determines that state saves per tick number.
1.73 +static const int32_t TIMER_INITIAL_VALUE = 10000;
1.74 +
1.75 +//-----------------------------------------------------------------------------
1.76 +//
1.77 +// Constructor and Destructor
1.78 +//
1.79 +//-----------------------------------------------------------------------------
1.80 +RegexMatcher::RegexMatcher(const RegexPattern *pat) {
1.81 + fDeferredStatus = U_ZERO_ERROR;
1.82 + init(fDeferredStatus);
1.83 + if (U_FAILURE(fDeferredStatus)) {
1.84 + return;
1.85 + }
1.86 + if (pat==NULL) {
1.87 + fDeferredStatus = U_ILLEGAL_ARGUMENT_ERROR;
1.88 + return;
1.89 + }
1.90 + fPattern = pat;
1.91 + init2(RegexStaticSets::gStaticSets->fEmptyText, fDeferredStatus);
1.92 +}
1.93 +
1.94 +
1.95 +
1.96 +RegexMatcher::RegexMatcher(const UnicodeString ®exp, const UnicodeString &input,
1.97 + uint32_t flags, UErrorCode &status) {
1.98 + init(status);
1.99 + if (U_FAILURE(status)) {
1.100 + return;
1.101 + }
1.102 + UParseError pe;
1.103 + fPatternOwned = RegexPattern::compile(regexp, flags, pe, status);
1.104 + fPattern = fPatternOwned;
1.105 +
1.106 + UText inputText = UTEXT_INITIALIZER;
1.107 + utext_openConstUnicodeString(&inputText, &input, &status);
1.108 + init2(&inputText, status);
1.109 + utext_close(&inputText);
1.110 +
1.111 + fInputUniStrMaybeMutable = TRUE;
1.112 +}
1.113 +
1.114 +
1.115 +RegexMatcher::RegexMatcher(UText *regexp, UText *input,
1.116 + uint32_t flags, UErrorCode &status) {
1.117 + init(status);
1.118 + if (U_FAILURE(status)) {
1.119 + return;
1.120 + }
1.121 + UParseError pe;
1.122 + fPatternOwned = RegexPattern::compile(regexp, flags, pe, status);
1.123 + if (U_FAILURE(status)) {
1.124 + return;
1.125 + }
1.126 +
1.127 + fPattern = fPatternOwned;
1.128 + init2(input, status);
1.129 +}
1.130 +
1.131 +
1.132 +RegexMatcher::RegexMatcher(const UnicodeString ®exp,
1.133 + uint32_t flags, UErrorCode &status) {
1.134 + init(status);
1.135 + if (U_FAILURE(status)) {
1.136 + return;
1.137 + }
1.138 + UParseError pe;
1.139 + fPatternOwned = RegexPattern::compile(regexp, flags, pe, status);
1.140 + if (U_FAILURE(status)) {
1.141 + return;
1.142 + }
1.143 + fPattern = fPatternOwned;
1.144 + init2(RegexStaticSets::gStaticSets->fEmptyText, status);
1.145 +}
1.146 +
1.147 +RegexMatcher::RegexMatcher(UText *regexp,
1.148 + uint32_t flags, UErrorCode &status) {
1.149 + init(status);
1.150 + if (U_FAILURE(status)) {
1.151 + return;
1.152 + }
1.153 + UParseError pe;
1.154 + fPatternOwned = RegexPattern::compile(regexp, flags, pe, status);
1.155 + if (U_FAILURE(status)) {
1.156 + return;
1.157 + }
1.158 +
1.159 + fPattern = fPatternOwned;
1.160 + init2(RegexStaticSets::gStaticSets->fEmptyText, status);
1.161 +}
1.162 +
1.163 +
1.164 +
1.165 +
1.166 +RegexMatcher::~RegexMatcher() {
1.167 + delete fStack;
1.168 + if (fData != fSmallData) {
1.169 + uprv_free(fData);
1.170 + fData = NULL;
1.171 + }
1.172 + if (fPatternOwned) {
1.173 + delete fPatternOwned;
1.174 + fPatternOwned = NULL;
1.175 + fPattern = NULL;
1.176 + }
1.177 +
1.178 + if (fInput) {
1.179 + delete fInput;
1.180 + }
1.181 + if (fInputText) {
1.182 + utext_close(fInputText);
1.183 + }
1.184 + if (fAltInputText) {
1.185 + utext_close(fAltInputText);
1.186 + }
1.187 +
1.188 + #if UCONFIG_NO_BREAK_ITERATION==0
1.189 + delete fWordBreakItr;
1.190 + #endif
1.191 +}
1.192 +
1.193 +//
1.194 +// init() common initialization for use by all constructors.
1.195 +// Initialize all fields, get the object into a consistent state.
1.196 +// This must be done even when the initial status shows an error,
1.197 +// so that the object is initialized sufficiently well for the destructor
1.198 +// to run safely.
1.199 +//
1.200 +void RegexMatcher::init(UErrorCode &status) {
1.201 + fPattern = NULL;
1.202 + fPatternOwned = NULL;
1.203 + fFrameSize = 0;
1.204 + fRegionStart = 0;
1.205 + fRegionLimit = 0;
1.206 + fAnchorStart = 0;
1.207 + fAnchorLimit = 0;
1.208 + fLookStart = 0;
1.209 + fLookLimit = 0;
1.210 + fActiveStart = 0;
1.211 + fActiveLimit = 0;
1.212 + fTransparentBounds = FALSE;
1.213 + fAnchoringBounds = TRUE;
1.214 + fMatch = FALSE;
1.215 + fMatchStart = 0;
1.216 + fMatchEnd = 0;
1.217 + fLastMatchEnd = -1;
1.218 + fAppendPosition = 0;
1.219 + fHitEnd = FALSE;
1.220 + fRequireEnd = FALSE;
1.221 + fStack = NULL;
1.222 + fFrame = NULL;
1.223 + fTimeLimit = 0;
1.224 + fTime = 0;
1.225 + fTickCounter = 0;
1.226 + fStackLimit = DEFAULT_BACKTRACK_STACK_CAPACITY;
1.227 + fCallbackFn = NULL;
1.228 + fCallbackContext = NULL;
1.229 + fFindProgressCallbackFn = NULL;
1.230 + fFindProgressCallbackContext = NULL;
1.231 + fTraceDebug = FALSE;
1.232 + fDeferredStatus = status;
1.233 + fData = fSmallData;
1.234 + fWordBreakItr = NULL;
1.235 +
1.236 + fStack = NULL;
1.237 + fInputText = NULL;
1.238 + fAltInputText = NULL;
1.239 + fInput = NULL;
1.240 + fInputLength = 0;
1.241 + fInputUniStrMaybeMutable = FALSE;
1.242 +
1.243 + if (U_FAILURE(status)) {
1.244 + fDeferredStatus = status;
1.245 + }
1.246 +}
1.247 +
1.248 +//
1.249 +// init2() Common initialization for use by RegexMatcher constructors, part 2.
1.250 +// This handles the common setup to be done after the Pattern is available.
1.251 +//
1.252 +void RegexMatcher::init2(UText *input, UErrorCode &status) {
1.253 + if (U_FAILURE(status)) {
1.254 + fDeferredStatus = status;
1.255 + return;
1.256 + }
1.257 +
1.258 + if (fPattern->fDataSize > (int32_t)(sizeof(fSmallData)/sizeof(fSmallData[0]))) {
1.259 + fData = (int64_t *)uprv_malloc(fPattern->fDataSize * sizeof(int64_t));
1.260 + if (fData == NULL) {
1.261 + status = fDeferredStatus = U_MEMORY_ALLOCATION_ERROR;
1.262 + return;
1.263 + }
1.264 + }
1.265 +
1.266 + fStack = new UVector64(status);
1.267 + if (fStack == NULL) {
1.268 + status = fDeferredStatus = U_MEMORY_ALLOCATION_ERROR;
1.269 + return;
1.270 + }
1.271 +
1.272 + reset(input);
1.273 + setStackLimit(DEFAULT_BACKTRACK_STACK_CAPACITY, status);
1.274 + if (U_FAILURE(status)) {
1.275 + fDeferredStatus = status;
1.276 + return;
1.277 + }
1.278 +}
1.279 +
1.280 +
1.281 +static const UChar BACKSLASH = 0x5c;
1.282 +static const UChar DOLLARSIGN = 0x24;
1.283 +//--------------------------------------------------------------------------------
1.284 +//
1.285 +// appendReplacement
1.286 +//
1.287 +//--------------------------------------------------------------------------------
1.288 +RegexMatcher &RegexMatcher::appendReplacement(UnicodeString &dest,
1.289 + const UnicodeString &replacement,
1.290 + UErrorCode &status) {
1.291 + UText replacementText = UTEXT_INITIALIZER;
1.292 +
1.293 + utext_openConstUnicodeString(&replacementText, &replacement, &status);
1.294 + if (U_SUCCESS(status)) {
1.295 + UText resultText = UTEXT_INITIALIZER;
1.296 + utext_openUnicodeString(&resultText, &dest, &status);
1.297 +
1.298 + if (U_SUCCESS(status)) {
1.299 + appendReplacement(&resultText, &replacementText, status);
1.300 + utext_close(&resultText);
1.301 + }
1.302 + utext_close(&replacementText);
1.303 + }
1.304 +
1.305 + return *this;
1.306 +}
1.307 +
1.308 +//
1.309 +// appendReplacement, UText mode
1.310 +//
1.311 +RegexMatcher &RegexMatcher::appendReplacement(UText *dest,
1.312 + UText *replacement,
1.313 + UErrorCode &status) {
1.314 + if (U_FAILURE(status)) {
1.315 + return *this;
1.316 + }
1.317 + if (U_FAILURE(fDeferredStatus)) {
1.318 + status = fDeferredStatus;
1.319 + return *this;
1.320 + }
1.321 + if (fMatch == FALSE) {
1.322 + status = U_REGEX_INVALID_STATE;
1.323 + return *this;
1.324 + }
1.325 +
1.326 + // Copy input string from the end of previous match to start of current match
1.327 + int64_t destLen = utext_nativeLength(dest);
1.328 + if (fMatchStart > fAppendPosition) {
1.329 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.330 + destLen += utext_replace(dest, destLen, destLen, fInputText->chunkContents+fAppendPosition,
1.331 + (int32_t)(fMatchStart-fAppendPosition), &status);
1.332 + } else {
1.333 + int32_t len16;
1.334 + if (UTEXT_USES_U16(fInputText)) {
1.335 + len16 = (int32_t)(fMatchStart-fAppendPosition);
1.336 + } else {
1.337 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.338 + len16 = utext_extract(fInputText, fAppendPosition, fMatchStart, NULL, 0, &lengthStatus);
1.339 + }
1.340 + UChar *inputChars = (UChar *)uprv_malloc(sizeof(UChar)*(len16+1));
1.341 + if (inputChars == NULL) {
1.342 + status = U_MEMORY_ALLOCATION_ERROR;
1.343 + return *this;
1.344 + }
1.345 + utext_extract(fInputText, fAppendPosition, fMatchStart, inputChars, len16+1, &status);
1.346 + destLen += utext_replace(dest, destLen, destLen, inputChars, len16, &status);
1.347 + uprv_free(inputChars);
1.348 + }
1.349 + }
1.350 + fAppendPosition = fMatchEnd;
1.351 +
1.352 +
1.353 + // scan the replacement text, looking for substitutions ($n) and \escapes.
1.354 + // TODO: optimize this loop by efficiently scanning for '$' or '\',
1.355 + // move entire ranges not containing substitutions.
1.356 + UTEXT_SETNATIVEINDEX(replacement, 0);
1.357 + UChar32 c = UTEXT_NEXT32(replacement);
1.358 + while (c != U_SENTINEL) {
1.359 + if (c == BACKSLASH) {
1.360 + // Backslash Escape. Copy the following char out without further checks.
1.361 + // Note: Surrogate pairs don't need any special handling
1.362 + // The second half wont be a '$' or a '\', and
1.363 + // will move to the dest normally on the next
1.364 + // loop iteration.
1.365 + c = UTEXT_CURRENT32(replacement);
1.366 + if (c == U_SENTINEL) {
1.367 + break;
1.368 + }
1.369 +
1.370 + if (c==0x55/*U*/ || c==0x75/*u*/) {
1.371 + // We have a \udddd or \Udddddddd escape sequence.
1.372 + int32_t offset = 0;
1.373 + struct URegexUTextUnescapeCharContext context = U_REGEX_UTEXT_UNESCAPE_CONTEXT(replacement);
1.374 + UChar32 escapedChar = u_unescapeAt(uregex_utext_unescape_charAt, &offset, INT32_MAX, &context);
1.375 + if (escapedChar != (UChar32)0xFFFFFFFF) {
1.376 + if (U_IS_BMP(escapedChar)) {
1.377 + UChar c16 = (UChar)escapedChar;
1.378 + destLen += utext_replace(dest, destLen, destLen, &c16, 1, &status);
1.379 + } else {
1.380 + UChar surrogate[2];
1.381 + surrogate[0] = U16_LEAD(escapedChar);
1.382 + surrogate[1] = U16_TRAIL(escapedChar);
1.383 + if (U_SUCCESS(status)) {
1.384 + destLen += utext_replace(dest, destLen, destLen, surrogate, 2, &status);
1.385 + }
1.386 + }
1.387 + // TODO: Report errors for mal-formed \u escapes?
1.388 + // As this is, the original sequence is output, which may be OK.
1.389 + if (context.lastOffset == offset) {
1.390 + (void)UTEXT_PREVIOUS32(replacement);
1.391 + } else if (context.lastOffset != offset-1) {
1.392 + utext_moveIndex32(replacement, offset - context.lastOffset - 1);
1.393 + }
1.394 + }
1.395 + } else {
1.396 + (void)UTEXT_NEXT32(replacement);
1.397 + // Plain backslash escape. Just put out the escaped character.
1.398 + if (U_IS_BMP(c)) {
1.399 + UChar c16 = (UChar)c;
1.400 + destLen += utext_replace(dest, destLen, destLen, &c16, 1, &status);
1.401 + } else {
1.402 + UChar surrogate[2];
1.403 + surrogate[0] = U16_LEAD(c);
1.404 + surrogate[1] = U16_TRAIL(c);
1.405 + if (U_SUCCESS(status)) {
1.406 + destLen += utext_replace(dest, destLen, destLen, surrogate, 2, &status);
1.407 + }
1.408 + }
1.409 + }
1.410 + } else if (c != DOLLARSIGN) {
1.411 + // Normal char, not a $. Copy it out without further checks.
1.412 + if (U_IS_BMP(c)) {
1.413 + UChar c16 = (UChar)c;
1.414 + destLen += utext_replace(dest, destLen, destLen, &c16, 1, &status);
1.415 + } else {
1.416 + UChar surrogate[2];
1.417 + surrogate[0] = U16_LEAD(c);
1.418 + surrogate[1] = U16_TRAIL(c);
1.419 + if (U_SUCCESS(status)) {
1.420 + destLen += utext_replace(dest, destLen, destLen, surrogate, 2, &status);
1.421 + }
1.422 + }
1.423 + } else {
1.424 + // We've got a $. Pick up a capture group number if one follows.
1.425 + // Consume at most the number of digits necessary for the largest capture
1.426 + // number that is valid for this pattern.
1.427 +
1.428 + int32_t numDigits = 0;
1.429 + int32_t groupNum = 0;
1.430 + UChar32 digitC;
1.431 + for (;;) {
1.432 + digitC = UTEXT_CURRENT32(replacement);
1.433 + if (digitC == U_SENTINEL) {
1.434 + break;
1.435 + }
1.436 + if (u_isdigit(digitC) == FALSE) {
1.437 + break;
1.438 + }
1.439 + (void)UTEXT_NEXT32(replacement);
1.440 + groupNum=groupNum*10 + u_charDigitValue(digitC);
1.441 + numDigits++;
1.442 + if (numDigits >= fPattern->fMaxCaptureDigits) {
1.443 + break;
1.444 + }
1.445 + }
1.446 +
1.447 +
1.448 + if (numDigits == 0) {
1.449 + // The $ didn't introduce a group number at all.
1.450 + // Treat it as just part of the substitution text.
1.451 + UChar c16 = DOLLARSIGN;
1.452 + destLen += utext_replace(dest, destLen, destLen, &c16, 1, &status);
1.453 + } else {
1.454 + // Finally, append the capture group data to the destination.
1.455 + destLen += appendGroup(groupNum, dest, status);
1.456 + if (U_FAILURE(status)) {
1.457 + // Can fail if group number is out of range.
1.458 + break;
1.459 + }
1.460 + }
1.461 + }
1.462 +
1.463 + if (U_FAILURE(status)) {
1.464 + break;
1.465 + } else {
1.466 + c = UTEXT_NEXT32(replacement);
1.467 + }
1.468 + }
1.469 +
1.470 + return *this;
1.471 +}
1.472 +
1.473 +
1.474 +
1.475 +//--------------------------------------------------------------------------------
1.476 +//
1.477 +// appendTail Intended to be used in conjunction with appendReplacement()
1.478 +// To the destination string, append everything following
1.479 +// the last match position from the input string.
1.480 +//
1.481 +// Note: Match ranges do not affect appendTail or appendReplacement
1.482 +//
1.483 +//--------------------------------------------------------------------------------
1.484 +UnicodeString &RegexMatcher::appendTail(UnicodeString &dest) {
1.485 + UErrorCode status = U_ZERO_ERROR;
1.486 + UText resultText = UTEXT_INITIALIZER;
1.487 + utext_openUnicodeString(&resultText, &dest, &status);
1.488 +
1.489 + if (U_SUCCESS(status)) {
1.490 + appendTail(&resultText, status);
1.491 + utext_close(&resultText);
1.492 + }
1.493 +
1.494 + return dest;
1.495 +}
1.496 +
1.497 +//
1.498 +// appendTail, UText mode
1.499 +//
1.500 +UText *RegexMatcher::appendTail(UText *dest, UErrorCode &status) {
1.501 + UBool bailOut = FALSE;
1.502 + if (U_FAILURE(status)) {
1.503 + bailOut = TRUE;
1.504 + }
1.505 + if (U_FAILURE(fDeferredStatus)) {
1.506 + status = fDeferredStatus;
1.507 + bailOut = TRUE;
1.508 + }
1.509 +
1.510 + if (bailOut) {
1.511 + // dest must not be NULL
1.512 + if (dest) {
1.513 + utext_replace(dest, utext_nativeLength(dest), utext_nativeLength(dest), NULL, 0, &status);
1.514 + return dest;
1.515 + }
1.516 + }
1.517 +
1.518 + if (fInputLength > fAppendPosition) {
1.519 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.520 + int64_t destLen = utext_nativeLength(dest);
1.521 + utext_replace(dest, destLen, destLen, fInputText->chunkContents+fAppendPosition,
1.522 + (int32_t)(fInputLength-fAppendPosition), &status);
1.523 + } else {
1.524 + int32_t len16;
1.525 + if (UTEXT_USES_U16(fInputText)) {
1.526 + len16 = (int32_t)(fInputLength-fAppendPosition);
1.527 + } else {
1.528 + len16 = utext_extract(fInputText, fAppendPosition, fInputLength, NULL, 0, &status);
1.529 + status = U_ZERO_ERROR; // buffer overflow
1.530 + }
1.531 +
1.532 + UChar *inputChars = (UChar *)uprv_malloc(sizeof(UChar)*(len16));
1.533 + if (inputChars == NULL) {
1.534 + fDeferredStatus = U_MEMORY_ALLOCATION_ERROR;
1.535 + } else {
1.536 + utext_extract(fInputText, fAppendPosition, fInputLength, inputChars, len16, &status); // unterminated
1.537 + int64_t destLen = utext_nativeLength(dest);
1.538 + utext_replace(dest, destLen, destLen, inputChars, len16, &status);
1.539 + uprv_free(inputChars);
1.540 + }
1.541 + }
1.542 + }
1.543 + return dest;
1.544 +}
1.545 +
1.546 +
1.547 +
1.548 +//--------------------------------------------------------------------------------
1.549 +//
1.550 +// end
1.551 +//
1.552 +//--------------------------------------------------------------------------------
1.553 +int32_t RegexMatcher::end(UErrorCode &err) const {
1.554 + return end(0, err);
1.555 +}
1.556 +
1.557 +int64_t RegexMatcher::end64(UErrorCode &err) const {
1.558 + return end64(0, err);
1.559 +}
1.560 +
1.561 +int64_t RegexMatcher::end64(int32_t group, UErrorCode &err) const {
1.562 + if (U_FAILURE(err)) {
1.563 + return -1;
1.564 + }
1.565 + if (fMatch == FALSE) {
1.566 + err = U_REGEX_INVALID_STATE;
1.567 + return -1;
1.568 + }
1.569 + if (group < 0 || group > fPattern->fGroupMap->size()) {
1.570 + err = U_INDEX_OUTOFBOUNDS_ERROR;
1.571 + return -1;
1.572 + }
1.573 + int64_t e = -1;
1.574 + if (group == 0) {
1.575 + e = fMatchEnd;
1.576 + } else {
1.577 + // Get the position within the stack frame of the variables for
1.578 + // this capture group.
1.579 + int32_t groupOffset = fPattern->fGroupMap->elementAti(group-1);
1.580 + U_ASSERT(groupOffset < fPattern->fFrameSize);
1.581 + U_ASSERT(groupOffset >= 0);
1.582 + e = fFrame->fExtra[groupOffset + 1];
1.583 + }
1.584 +
1.585 + return e;
1.586 +}
1.587 +
1.588 +int32_t RegexMatcher::end(int32_t group, UErrorCode &err) const {
1.589 + return (int32_t)end64(group, err);
1.590 +}
1.591 +
1.592 +
1.593 +//--------------------------------------------------------------------------------
1.594 +//
1.595 +// find()
1.596 +//
1.597 +//--------------------------------------------------------------------------------
1.598 +UBool RegexMatcher::find() {
1.599 + // Start at the position of the last match end. (Will be zero if the
1.600 + // matcher has been reset.)
1.601 + //
1.602 + if (U_FAILURE(fDeferredStatus)) {
1.603 + return FALSE;
1.604 + }
1.605 +
1.606 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.607 + return findUsingChunk();
1.608 + }
1.609 +
1.610 + int64_t startPos = fMatchEnd;
1.611 + if (startPos==0) {
1.612 + startPos = fActiveStart;
1.613 + }
1.614 +
1.615 + if (fMatch) {
1.616 + // Save the position of any previous successful match.
1.617 + fLastMatchEnd = fMatchEnd;
1.618 +
1.619 + if (fMatchStart == fMatchEnd) {
1.620 + // Previous match had zero length. Move start position up one position
1.621 + // to avoid sending find() into a loop on zero-length matches.
1.622 + if (startPos >= fActiveLimit) {
1.623 + fMatch = FALSE;
1.624 + fHitEnd = TRUE;
1.625 + return FALSE;
1.626 + }
1.627 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.628 + (void)UTEXT_NEXT32(fInputText);
1.629 + startPos = UTEXT_GETNATIVEINDEX(fInputText);
1.630 + }
1.631 + } else {
1.632 + if (fLastMatchEnd >= 0) {
1.633 + // A previous find() failed to match. Don't try again.
1.634 + // (without this test, a pattern with a zero-length match
1.635 + // could match again at the end of an input string.)
1.636 + fHitEnd = TRUE;
1.637 + return FALSE;
1.638 + }
1.639 + }
1.640 +
1.641 +
1.642 + // Compute the position in the input string beyond which a match can not begin, because
1.643 + // the minimum length match would extend past the end of the input.
1.644 + // Note: some patterns that cannot match anything will have fMinMatchLength==Max Int.
1.645 + // Be aware of possible overflows if making changes here.
1.646 + int64_t testStartLimit;
1.647 + if (UTEXT_USES_U16(fInputText)) {
1.648 + testStartLimit = fActiveLimit - fPattern->fMinMatchLen;
1.649 + if (startPos > testStartLimit) {
1.650 + fMatch = FALSE;
1.651 + fHitEnd = TRUE;
1.652 + return FALSE;
1.653 + }
1.654 + } else {
1.655 + // For now, let the matcher discover that it can't match on its own
1.656 + // We don't know how long the match len is in native characters
1.657 + testStartLimit = fActiveLimit;
1.658 + }
1.659 +
1.660 + UChar32 c;
1.661 + U_ASSERT(startPos >= 0);
1.662 +
1.663 + switch (fPattern->fStartType) {
1.664 + case START_NO_INFO:
1.665 + // No optimization was found.
1.666 + // Try a match at each input position.
1.667 + for (;;) {
1.668 + MatchAt(startPos, FALSE, fDeferredStatus);
1.669 + if (U_FAILURE(fDeferredStatus)) {
1.670 + return FALSE;
1.671 + }
1.672 + if (fMatch) {
1.673 + return TRUE;
1.674 + }
1.675 + if (startPos >= testStartLimit) {
1.676 + fHitEnd = TRUE;
1.677 + return FALSE;
1.678 + }
1.679 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.680 + (void)UTEXT_NEXT32(fInputText);
1.681 + startPos = UTEXT_GETNATIVEINDEX(fInputText);
1.682 + // Note that it's perfectly OK for a pattern to have a zero-length
1.683 + // match at the end of a string, so we must make sure that the loop
1.684 + // runs with startPos == testStartLimit the last time through.
1.685 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.686 + return FALSE;
1.687 + }
1.688 + U_ASSERT(FALSE);
1.689 +
1.690 + case START_START:
1.691 + // Matches are only possible at the start of the input string
1.692 + // (pattern begins with ^ or \A)
1.693 + if (startPos > fActiveStart) {
1.694 + fMatch = FALSE;
1.695 + return FALSE;
1.696 + }
1.697 + MatchAt(startPos, FALSE, fDeferredStatus);
1.698 + if (U_FAILURE(fDeferredStatus)) {
1.699 + return FALSE;
1.700 + }
1.701 + return fMatch;
1.702 +
1.703 +
1.704 + case START_SET:
1.705 + {
1.706 + // Match may start on any char from a pre-computed set.
1.707 + U_ASSERT(fPattern->fMinMatchLen > 0);
1.708 + int64_t pos;
1.709 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.710 + for (;;) {
1.711 + c = UTEXT_NEXT32(fInputText);
1.712 + pos = UTEXT_GETNATIVEINDEX(fInputText);
1.713 + // c will be -1 (U_SENTINEL) at end of text, in which case we
1.714 + // skip this next block (so we don't have a negative array index)
1.715 + // and handle end of text in the following block.
1.716 + if (c >= 0 && ((c<256 && fPattern->fInitialChars8->contains(c)) ||
1.717 + (c>=256 && fPattern->fInitialChars->contains(c)))) {
1.718 + MatchAt(startPos, FALSE, fDeferredStatus);
1.719 + if (U_FAILURE(fDeferredStatus)) {
1.720 + return FALSE;
1.721 + }
1.722 + if (fMatch) {
1.723 + return TRUE;
1.724 + }
1.725 + UTEXT_SETNATIVEINDEX(fInputText, pos);
1.726 + }
1.727 + if (startPos >= testStartLimit) {
1.728 + fMatch = FALSE;
1.729 + fHitEnd = TRUE;
1.730 + return FALSE;
1.731 + }
1.732 + startPos = pos;
1.733 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.734 + return FALSE;
1.735 + }
1.736 + }
1.737 + U_ASSERT(FALSE);
1.738 +
1.739 + case START_STRING:
1.740 + case START_CHAR:
1.741 + {
1.742 + // Match starts on exactly one char.
1.743 + U_ASSERT(fPattern->fMinMatchLen > 0);
1.744 + UChar32 theChar = fPattern->fInitialChar;
1.745 + int64_t pos;
1.746 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.747 + for (;;) {
1.748 + c = UTEXT_NEXT32(fInputText);
1.749 + pos = UTEXT_GETNATIVEINDEX(fInputText);
1.750 + if (c == theChar) {
1.751 + MatchAt(startPos, FALSE, fDeferredStatus);
1.752 + if (U_FAILURE(fDeferredStatus)) {
1.753 + return FALSE;
1.754 + }
1.755 + if (fMatch) {
1.756 + return TRUE;
1.757 + }
1.758 + UTEXT_SETNATIVEINDEX(fInputText, pos);
1.759 + }
1.760 + if (startPos >= testStartLimit) {
1.761 + fMatch = FALSE;
1.762 + fHitEnd = TRUE;
1.763 + return FALSE;
1.764 + }
1.765 + startPos = pos;
1.766 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.767 + return FALSE;
1.768 + }
1.769 + }
1.770 + U_ASSERT(FALSE);
1.771 +
1.772 + case START_LINE:
1.773 + {
1.774 + UChar32 c;
1.775 + if (startPos == fAnchorStart) {
1.776 + MatchAt(startPos, FALSE, fDeferredStatus);
1.777 + if (U_FAILURE(fDeferredStatus)) {
1.778 + return FALSE;
1.779 + }
1.780 + if (fMatch) {
1.781 + return TRUE;
1.782 + }
1.783 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.784 + c = UTEXT_NEXT32(fInputText);
1.785 + startPos = UTEXT_GETNATIVEINDEX(fInputText);
1.786 + } else {
1.787 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.788 + c = UTEXT_PREVIOUS32(fInputText);
1.789 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.790 + }
1.791 +
1.792 + if (fPattern->fFlags & UREGEX_UNIX_LINES) {
1.793 + for (;;) {
1.794 + if (c == 0x0a) {
1.795 + MatchAt(startPos, FALSE, fDeferredStatus);
1.796 + if (U_FAILURE(fDeferredStatus)) {
1.797 + return FALSE;
1.798 + }
1.799 + if (fMatch) {
1.800 + return TRUE;
1.801 + }
1.802 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.803 + }
1.804 + if (startPos >= testStartLimit) {
1.805 + fMatch = FALSE;
1.806 + fHitEnd = TRUE;
1.807 + return FALSE;
1.808 + }
1.809 + c = UTEXT_NEXT32(fInputText);
1.810 + startPos = UTEXT_GETNATIVEINDEX(fInputText);
1.811 + // Note that it's perfectly OK for a pattern to have a zero-length
1.812 + // match at the end of a string, so we must make sure that the loop
1.813 + // runs with startPos == testStartLimit the last time through.
1.814 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.815 + return FALSE;
1.816 + }
1.817 + } else {
1.818 + for (;;) {
1.819 + if (((c & 0x7f) <= 0x29) && // First quickly bypass as many chars as possible
1.820 + ((c<=0x0d && c>=0x0a) || c==0x85 ||c==0x2028 || c==0x2029 )) {
1.821 + if (c == 0x0d && startPos < fActiveLimit && UTEXT_CURRENT32(fInputText) == 0x0a) {
1.822 + (void)UTEXT_NEXT32(fInputText);
1.823 + startPos = UTEXT_GETNATIVEINDEX(fInputText);
1.824 + }
1.825 + MatchAt(startPos, FALSE, fDeferredStatus);
1.826 + if (U_FAILURE(fDeferredStatus)) {
1.827 + return FALSE;
1.828 + }
1.829 + if (fMatch) {
1.830 + return TRUE;
1.831 + }
1.832 + UTEXT_SETNATIVEINDEX(fInputText, startPos);
1.833 + }
1.834 + if (startPos >= testStartLimit) {
1.835 + fMatch = FALSE;
1.836 + fHitEnd = TRUE;
1.837 + return FALSE;
1.838 + }
1.839 + c = UTEXT_NEXT32(fInputText);
1.840 + startPos = UTEXT_GETNATIVEINDEX(fInputText);
1.841 + // Note that it's perfectly OK for a pattern to have a zero-length
1.842 + // match at the end of a string, so we must make sure that the loop
1.843 + // runs with startPos == testStartLimit the last time through.
1.844 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.845 + return FALSE;
1.846 + }
1.847 + }
1.848 + }
1.849 +
1.850 + default:
1.851 + U_ASSERT(FALSE);
1.852 + }
1.853 +
1.854 + U_ASSERT(FALSE);
1.855 + return FALSE;
1.856 +}
1.857 +
1.858 +
1.859 +
1.860 +UBool RegexMatcher::find(int64_t start, UErrorCode &status) {
1.861 + if (U_FAILURE(status)) {
1.862 + return FALSE;
1.863 + }
1.864 + if (U_FAILURE(fDeferredStatus)) {
1.865 + status = fDeferredStatus;
1.866 + return FALSE;
1.867 + }
1.868 + this->reset(); // Note: Reset() is specified by Java Matcher documentation.
1.869 + // This will reset the region to be the full input length.
1.870 + if (start < 0) {
1.871 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.872 + return FALSE;
1.873 + }
1.874 +
1.875 + int64_t nativeStart = start;
1.876 + if (nativeStart < fActiveStart || nativeStart > fActiveLimit) {
1.877 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.878 + return FALSE;
1.879 + }
1.880 + fMatchEnd = nativeStart;
1.881 + return find();
1.882 +}
1.883 +
1.884 +
1.885 +//--------------------------------------------------------------------------------
1.886 +//
1.887 +// findUsingChunk() -- like find(), but with the advance knowledge that the
1.888 +// entire string is available in the UText's chunk buffer.
1.889 +//
1.890 +//--------------------------------------------------------------------------------
1.891 +UBool RegexMatcher::findUsingChunk() {
1.892 + // Start at the position of the last match end. (Will be zero if the
1.893 + // matcher has been reset.
1.894 + //
1.895 +
1.896 + int32_t startPos = (int32_t)fMatchEnd;
1.897 + if (startPos==0) {
1.898 + startPos = (int32_t)fActiveStart;
1.899 + }
1.900 +
1.901 + const UChar *inputBuf = fInputText->chunkContents;
1.902 +
1.903 + if (fMatch) {
1.904 + // Save the position of any previous successful match.
1.905 + fLastMatchEnd = fMatchEnd;
1.906 +
1.907 + if (fMatchStart == fMatchEnd) {
1.908 + // Previous match had zero length. Move start position up one position
1.909 + // to avoid sending find() into a loop on zero-length matches.
1.910 + if (startPos >= fActiveLimit) {
1.911 + fMatch = FALSE;
1.912 + fHitEnd = TRUE;
1.913 + return FALSE;
1.914 + }
1.915 + U16_FWD_1(inputBuf, startPos, fInputLength);
1.916 + }
1.917 + } else {
1.918 + if (fLastMatchEnd >= 0) {
1.919 + // A previous find() failed to match. Don't try again.
1.920 + // (without this test, a pattern with a zero-length match
1.921 + // could match again at the end of an input string.)
1.922 + fHitEnd = TRUE;
1.923 + return FALSE;
1.924 + }
1.925 + }
1.926 +
1.927 +
1.928 + // Compute the position in the input string beyond which a match can not begin, because
1.929 + // the minimum length match would extend past the end of the input.
1.930 + // Note: some patterns that cannot match anything will have fMinMatchLength==Max Int.
1.931 + // Be aware of possible overflows if making changes here.
1.932 + int32_t testLen = (int32_t)(fActiveLimit - fPattern->fMinMatchLen);
1.933 + if (startPos > testLen) {
1.934 + fMatch = FALSE;
1.935 + fHitEnd = TRUE;
1.936 + return FALSE;
1.937 + }
1.938 +
1.939 + UChar32 c;
1.940 + U_ASSERT(startPos >= 0);
1.941 +
1.942 + switch (fPattern->fStartType) {
1.943 + case START_NO_INFO:
1.944 + // No optimization was found.
1.945 + // Try a match at each input position.
1.946 + for (;;) {
1.947 + MatchChunkAt(startPos, FALSE, fDeferredStatus);
1.948 + if (U_FAILURE(fDeferredStatus)) {
1.949 + return FALSE;
1.950 + }
1.951 + if (fMatch) {
1.952 + return TRUE;
1.953 + }
1.954 + if (startPos >= testLen) {
1.955 + fHitEnd = TRUE;
1.956 + return FALSE;
1.957 + }
1.958 + U16_FWD_1(inputBuf, startPos, fActiveLimit);
1.959 + // Note that it's perfectly OK for a pattern to have a zero-length
1.960 + // match at the end of a string, so we must make sure that the loop
1.961 + // runs with startPos == testLen the last time through.
1.962 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.963 + return FALSE;
1.964 + }
1.965 + U_ASSERT(FALSE);
1.966 +
1.967 + case START_START:
1.968 + // Matches are only possible at the start of the input string
1.969 + // (pattern begins with ^ or \A)
1.970 + if (startPos > fActiveStart) {
1.971 + fMatch = FALSE;
1.972 + return FALSE;
1.973 + }
1.974 + MatchChunkAt(startPos, FALSE, fDeferredStatus);
1.975 + if (U_FAILURE(fDeferredStatus)) {
1.976 + return FALSE;
1.977 + }
1.978 + return fMatch;
1.979 +
1.980 +
1.981 + case START_SET:
1.982 + {
1.983 + // Match may start on any char from a pre-computed set.
1.984 + U_ASSERT(fPattern->fMinMatchLen > 0);
1.985 + for (;;) {
1.986 + int32_t pos = startPos;
1.987 + U16_NEXT(inputBuf, startPos, fActiveLimit, c); // like c = inputBuf[startPos++];
1.988 + if ((c<256 && fPattern->fInitialChars8->contains(c)) ||
1.989 + (c>=256 && fPattern->fInitialChars->contains(c))) {
1.990 + MatchChunkAt(pos, FALSE, fDeferredStatus);
1.991 + if (U_FAILURE(fDeferredStatus)) {
1.992 + return FALSE;
1.993 + }
1.994 + if (fMatch) {
1.995 + return TRUE;
1.996 + }
1.997 + }
1.998 + if (pos >= testLen) {
1.999 + fMatch = FALSE;
1.1000 + fHitEnd = TRUE;
1.1001 + return FALSE;
1.1002 + }
1.1003 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.1004 + return FALSE;
1.1005 + }
1.1006 + }
1.1007 + U_ASSERT(FALSE);
1.1008 +
1.1009 + case START_STRING:
1.1010 + case START_CHAR:
1.1011 + {
1.1012 + // Match starts on exactly one char.
1.1013 + U_ASSERT(fPattern->fMinMatchLen > 0);
1.1014 + UChar32 theChar = fPattern->fInitialChar;
1.1015 + for (;;) {
1.1016 + int32_t pos = startPos;
1.1017 + U16_NEXT(inputBuf, startPos, fActiveLimit, c); // like c = inputBuf[startPos++];
1.1018 + if (c == theChar) {
1.1019 + MatchChunkAt(pos, FALSE, fDeferredStatus);
1.1020 + if (U_FAILURE(fDeferredStatus)) {
1.1021 + return FALSE;
1.1022 + }
1.1023 + if (fMatch) {
1.1024 + return TRUE;
1.1025 + }
1.1026 + }
1.1027 + if (pos >= testLen) {
1.1028 + fMatch = FALSE;
1.1029 + fHitEnd = TRUE;
1.1030 + return FALSE;
1.1031 + }
1.1032 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.1033 + return FALSE;
1.1034 + }
1.1035 + }
1.1036 + U_ASSERT(FALSE);
1.1037 +
1.1038 + case START_LINE:
1.1039 + {
1.1040 + UChar32 c;
1.1041 + if (startPos == fAnchorStart) {
1.1042 + MatchChunkAt(startPos, FALSE, fDeferredStatus);
1.1043 + if (U_FAILURE(fDeferredStatus)) {
1.1044 + return FALSE;
1.1045 + }
1.1046 + if (fMatch) {
1.1047 + return TRUE;
1.1048 + }
1.1049 + U16_FWD_1(inputBuf, startPos, fActiveLimit);
1.1050 + }
1.1051 +
1.1052 + if (fPattern->fFlags & UREGEX_UNIX_LINES) {
1.1053 + for (;;) {
1.1054 + c = inputBuf[startPos-1];
1.1055 + if (c == 0x0a) {
1.1056 + MatchChunkAt(startPos, FALSE, fDeferredStatus);
1.1057 + if (U_FAILURE(fDeferredStatus)) {
1.1058 + return FALSE;
1.1059 + }
1.1060 + if (fMatch) {
1.1061 + return TRUE;
1.1062 + }
1.1063 + }
1.1064 + if (startPos >= testLen) {
1.1065 + fMatch = FALSE;
1.1066 + fHitEnd = TRUE;
1.1067 + return FALSE;
1.1068 + }
1.1069 + U16_FWD_1(inputBuf, startPos, fActiveLimit);
1.1070 + // Note that it's perfectly OK for a pattern to have a zero-length
1.1071 + // match at the end of a string, so we must make sure that the loop
1.1072 + // runs with startPos == testLen the last time through.
1.1073 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.1074 + return FALSE;
1.1075 + }
1.1076 + } else {
1.1077 + for (;;) {
1.1078 + c = inputBuf[startPos-1];
1.1079 + if (((c & 0x7f) <= 0x29) && // First quickly bypass as many chars as possible
1.1080 + ((c<=0x0d && c>=0x0a) || c==0x85 ||c==0x2028 || c==0x2029 )) {
1.1081 + if (c == 0x0d && startPos < fActiveLimit && inputBuf[startPos] == 0x0a) {
1.1082 + startPos++;
1.1083 + }
1.1084 + MatchChunkAt(startPos, FALSE, fDeferredStatus);
1.1085 + if (U_FAILURE(fDeferredStatus)) {
1.1086 + return FALSE;
1.1087 + }
1.1088 + if (fMatch) {
1.1089 + return TRUE;
1.1090 + }
1.1091 + }
1.1092 + if (startPos >= testLen) {
1.1093 + fMatch = FALSE;
1.1094 + fHitEnd = TRUE;
1.1095 + return FALSE;
1.1096 + }
1.1097 + U16_FWD_1(inputBuf, startPos, fActiveLimit);
1.1098 + // Note that it's perfectly OK for a pattern to have a zero-length
1.1099 + // match at the end of a string, so we must make sure that the loop
1.1100 + // runs with startPos == testLen the last time through.
1.1101 + if (REGEXFINDPROGRESS_INTERRUPT(startPos, fDeferredStatus))
1.1102 + return FALSE;
1.1103 + }
1.1104 + }
1.1105 + }
1.1106 +
1.1107 + default:
1.1108 + U_ASSERT(FALSE);
1.1109 + }
1.1110 +
1.1111 + U_ASSERT(FALSE);
1.1112 + return FALSE;
1.1113 +}
1.1114 +
1.1115 +
1.1116 +
1.1117 +//--------------------------------------------------------------------------------
1.1118 +//
1.1119 +// group()
1.1120 +//
1.1121 +//--------------------------------------------------------------------------------
1.1122 +UnicodeString RegexMatcher::group(UErrorCode &status) const {
1.1123 + return group(0, status);
1.1124 +}
1.1125 +
1.1126 +// Return immutable shallow clone
1.1127 +UText *RegexMatcher::group(UText *dest, int64_t &group_len, UErrorCode &status) const {
1.1128 + return group(0, dest, group_len, status);
1.1129 +}
1.1130 +
1.1131 +// Return immutable shallow clone
1.1132 +UText *RegexMatcher::group(int32_t groupNum, UText *dest, int64_t &group_len, UErrorCode &status) const {
1.1133 + group_len = 0;
1.1134 + UBool bailOut = FALSE;
1.1135 + if (U_FAILURE(status)) {
1.1136 + return dest;
1.1137 + }
1.1138 + if (U_FAILURE(fDeferredStatus)) {
1.1139 + status = fDeferredStatus;
1.1140 + bailOut = TRUE;
1.1141 + }
1.1142 + if (fMatch == FALSE) {
1.1143 + status = U_REGEX_INVALID_STATE;
1.1144 + bailOut = TRUE;
1.1145 + }
1.1146 + if (groupNum < 0 || groupNum > fPattern->fGroupMap->size()) {
1.1147 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1148 + bailOut = TRUE;
1.1149 + }
1.1150 +
1.1151 + if (bailOut) {
1.1152 + return (dest) ? dest : utext_openUChars(NULL, NULL, 0, &status);
1.1153 + }
1.1154 +
1.1155 + int64_t s, e;
1.1156 + if (groupNum == 0) {
1.1157 + s = fMatchStart;
1.1158 + e = fMatchEnd;
1.1159 + } else {
1.1160 + int32_t groupOffset = fPattern->fGroupMap->elementAti(groupNum-1);
1.1161 + U_ASSERT(groupOffset < fPattern->fFrameSize);
1.1162 + U_ASSERT(groupOffset >= 0);
1.1163 + s = fFrame->fExtra[groupOffset];
1.1164 + e = fFrame->fExtra[groupOffset+1];
1.1165 + }
1.1166 +
1.1167 + if (s < 0) {
1.1168 + // A capture group wasn't part of the match
1.1169 + return utext_clone(dest, fInputText, FALSE, TRUE, &status);
1.1170 + }
1.1171 + U_ASSERT(s <= e);
1.1172 + group_len = e - s;
1.1173 +
1.1174 + dest = utext_clone(dest, fInputText, FALSE, TRUE, &status);
1.1175 + if (dest)
1.1176 + UTEXT_SETNATIVEINDEX(dest, s);
1.1177 + return dest;
1.1178 +}
1.1179 +
1.1180 +UnicodeString RegexMatcher::group(int32_t groupNum, UErrorCode &status) const {
1.1181 + UnicodeString result;
1.1182 + if (U_FAILURE(status)) {
1.1183 + return result;
1.1184 + }
1.1185 + UText resultText = UTEXT_INITIALIZER;
1.1186 + utext_openUnicodeString(&resultText, &result, &status);
1.1187 + group(groupNum, &resultText, status);
1.1188 + utext_close(&resultText);
1.1189 + return result;
1.1190 +}
1.1191 +
1.1192 +
1.1193 +// Return deep (mutable) clone
1.1194 +// Technology Preview (as an API), but note that the UnicodeString API is implemented
1.1195 +// using this function.
1.1196 +UText *RegexMatcher::group(int32_t groupNum, UText *dest, UErrorCode &status) const {
1.1197 + UBool bailOut = FALSE;
1.1198 + if (U_FAILURE(status)) {
1.1199 + return dest;
1.1200 + }
1.1201 + if (U_FAILURE(fDeferredStatus)) {
1.1202 + status = fDeferredStatus;
1.1203 + bailOut = TRUE;
1.1204 + }
1.1205 +
1.1206 + if (fMatch == FALSE) {
1.1207 + status = U_REGEX_INVALID_STATE;
1.1208 + bailOut = TRUE;
1.1209 + }
1.1210 + if (groupNum < 0 || groupNum > fPattern->fGroupMap->size()) {
1.1211 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1212 + bailOut = TRUE;
1.1213 + }
1.1214 +
1.1215 + if (bailOut) {
1.1216 + if (dest) {
1.1217 + utext_replace(dest, 0, utext_nativeLength(dest), NULL, 0, &status);
1.1218 + return dest;
1.1219 + } else {
1.1220 + return utext_openUChars(NULL, NULL, 0, &status);
1.1221 + }
1.1222 + }
1.1223 +
1.1224 + int64_t s, e;
1.1225 + if (groupNum == 0) {
1.1226 + s = fMatchStart;
1.1227 + e = fMatchEnd;
1.1228 + } else {
1.1229 + int32_t groupOffset = fPattern->fGroupMap->elementAti(groupNum-1);
1.1230 + U_ASSERT(groupOffset < fPattern->fFrameSize);
1.1231 + U_ASSERT(groupOffset >= 0);
1.1232 + s = fFrame->fExtra[groupOffset];
1.1233 + e = fFrame->fExtra[groupOffset+1];
1.1234 + }
1.1235 +
1.1236 + if (s < 0) {
1.1237 + // A capture group wasn't part of the match
1.1238 + if (dest) {
1.1239 + utext_replace(dest, 0, utext_nativeLength(dest), NULL, 0, &status);
1.1240 + return dest;
1.1241 + } else {
1.1242 + return utext_openUChars(NULL, NULL, 0, &status);
1.1243 + }
1.1244 + }
1.1245 + U_ASSERT(s <= e);
1.1246 +
1.1247 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1248 + U_ASSERT(e <= fInputLength);
1.1249 + if (dest) {
1.1250 + utext_replace(dest, 0, utext_nativeLength(dest), fInputText->chunkContents+s, (int32_t)(e-s), &status);
1.1251 + } else {
1.1252 + UText groupText = UTEXT_INITIALIZER;
1.1253 + utext_openUChars(&groupText, fInputText->chunkContents+s, e-s, &status);
1.1254 + dest = utext_clone(NULL, &groupText, TRUE, FALSE, &status);
1.1255 + utext_close(&groupText);
1.1256 + }
1.1257 + } else {
1.1258 + int32_t len16;
1.1259 + if (UTEXT_USES_U16(fInputText)) {
1.1260 + len16 = (int32_t)(e-s);
1.1261 + } else {
1.1262 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.1263 + len16 = utext_extract(fInputText, s, e, NULL, 0, &lengthStatus);
1.1264 + }
1.1265 + UChar *groupChars = (UChar *)uprv_malloc(sizeof(UChar)*(len16+1));
1.1266 + if (groupChars == NULL) {
1.1267 + status = U_MEMORY_ALLOCATION_ERROR;
1.1268 + return dest;
1.1269 + }
1.1270 + utext_extract(fInputText, s, e, groupChars, len16+1, &status);
1.1271 +
1.1272 + if (dest) {
1.1273 + utext_replace(dest, 0, utext_nativeLength(dest), groupChars, len16, &status);
1.1274 + } else {
1.1275 + UText groupText = UTEXT_INITIALIZER;
1.1276 + utext_openUChars(&groupText, groupChars, len16, &status);
1.1277 + dest = utext_clone(NULL, &groupText, TRUE, FALSE, &status);
1.1278 + utext_close(&groupText);
1.1279 + }
1.1280 +
1.1281 + uprv_free(groupChars);
1.1282 + }
1.1283 + return dest;
1.1284 +}
1.1285 +
1.1286 +//--------------------------------------------------------------------------------
1.1287 +//
1.1288 +// appendGroup() -- currently internal only, appends a group to a UText rather
1.1289 +// than replacing its contents
1.1290 +//
1.1291 +//--------------------------------------------------------------------------------
1.1292 +
1.1293 +int64_t RegexMatcher::appendGroup(int32_t groupNum, UText *dest, UErrorCode &status) const {
1.1294 + if (U_FAILURE(status)) {
1.1295 + return 0;
1.1296 + }
1.1297 + if (U_FAILURE(fDeferredStatus)) {
1.1298 + status = fDeferredStatus;
1.1299 + return 0;
1.1300 + }
1.1301 + int64_t destLen = utext_nativeLength(dest);
1.1302 +
1.1303 + if (fMatch == FALSE) {
1.1304 + status = U_REGEX_INVALID_STATE;
1.1305 + return utext_replace(dest, destLen, destLen, NULL, 0, &status);
1.1306 + }
1.1307 + if (groupNum < 0 || groupNum > fPattern->fGroupMap->size()) {
1.1308 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1309 + return utext_replace(dest, destLen, destLen, NULL, 0, &status);
1.1310 + }
1.1311 +
1.1312 + int64_t s, e;
1.1313 + if (groupNum == 0) {
1.1314 + s = fMatchStart;
1.1315 + e = fMatchEnd;
1.1316 + } else {
1.1317 + int32_t groupOffset = fPattern->fGroupMap->elementAti(groupNum-1);
1.1318 + U_ASSERT(groupOffset < fPattern->fFrameSize);
1.1319 + U_ASSERT(groupOffset >= 0);
1.1320 + s = fFrame->fExtra[groupOffset];
1.1321 + e = fFrame->fExtra[groupOffset+1];
1.1322 + }
1.1323 +
1.1324 + if (s < 0) {
1.1325 + // A capture group wasn't part of the match
1.1326 + return utext_replace(dest, destLen, destLen, NULL, 0, &status);
1.1327 + }
1.1328 + U_ASSERT(s <= e);
1.1329 +
1.1330 + int64_t deltaLen;
1.1331 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1332 + U_ASSERT(e <= fInputLength);
1.1333 + deltaLen = utext_replace(dest, destLen, destLen, fInputText->chunkContents+s, (int32_t)(e-s), &status);
1.1334 + } else {
1.1335 + int32_t len16;
1.1336 + if (UTEXT_USES_U16(fInputText)) {
1.1337 + len16 = (int32_t)(e-s);
1.1338 + } else {
1.1339 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.1340 + len16 = utext_extract(fInputText, s, e, NULL, 0, &lengthStatus);
1.1341 + }
1.1342 + UChar *groupChars = (UChar *)uprv_malloc(sizeof(UChar)*(len16+1));
1.1343 + if (groupChars == NULL) {
1.1344 + status = U_MEMORY_ALLOCATION_ERROR;
1.1345 + return 0;
1.1346 + }
1.1347 + utext_extract(fInputText, s, e, groupChars, len16+1, &status);
1.1348 +
1.1349 + deltaLen = utext_replace(dest, destLen, destLen, groupChars, len16, &status);
1.1350 + uprv_free(groupChars);
1.1351 + }
1.1352 + return deltaLen;
1.1353 +}
1.1354 +
1.1355 +
1.1356 +
1.1357 +//--------------------------------------------------------------------------------
1.1358 +//
1.1359 +// groupCount()
1.1360 +//
1.1361 +//--------------------------------------------------------------------------------
1.1362 +int32_t RegexMatcher::groupCount() const {
1.1363 + return fPattern->fGroupMap->size();
1.1364 +}
1.1365 +
1.1366 +
1.1367 +
1.1368 +//--------------------------------------------------------------------------------
1.1369 +//
1.1370 +// hasAnchoringBounds()
1.1371 +//
1.1372 +//--------------------------------------------------------------------------------
1.1373 +UBool RegexMatcher::hasAnchoringBounds() const {
1.1374 + return fAnchoringBounds;
1.1375 +}
1.1376 +
1.1377 +
1.1378 +//--------------------------------------------------------------------------------
1.1379 +//
1.1380 +// hasTransparentBounds()
1.1381 +//
1.1382 +//--------------------------------------------------------------------------------
1.1383 +UBool RegexMatcher::hasTransparentBounds() const {
1.1384 + return fTransparentBounds;
1.1385 +}
1.1386 +
1.1387 +
1.1388 +
1.1389 +//--------------------------------------------------------------------------------
1.1390 +//
1.1391 +// hitEnd()
1.1392 +//
1.1393 +//--------------------------------------------------------------------------------
1.1394 +UBool RegexMatcher::hitEnd() const {
1.1395 + return fHitEnd;
1.1396 +}
1.1397 +
1.1398 +
1.1399 +//--------------------------------------------------------------------------------
1.1400 +//
1.1401 +// input()
1.1402 +//
1.1403 +//--------------------------------------------------------------------------------
1.1404 +const UnicodeString &RegexMatcher::input() const {
1.1405 + if (!fInput) {
1.1406 + UErrorCode status = U_ZERO_ERROR;
1.1407 + int32_t len16;
1.1408 + if (UTEXT_USES_U16(fInputText)) {
1.1409 + len16 = (int32_t)fInputLength;
1.1410 + } else {
1.1411 + len16 = utext_extract(fInputText, 0, fInputLength, NULL, 0, &status);
1.1412 + status = U_ZERO_ERROR; // overflow, length status
1.1413 + }
1.1414 + UnicodeString *result = new UnicodeString(len16, 0, 0);
1.1415 +
1.1416 + UChar *inputChars = result->getBuffer(len16);
1.1417 + utext_extract(fInputText, 0, fInputLength, inputChars, len16, &status); // unterminated warning
1.1418 + result->releaseBuffer(len16);
1.1419 +
1.1420 + (*(const UnicodeString **)&fInput) = result; // pointer assignment, rather than operator=
1.1421 + }
1.1422 +
1.1423 + return *fInput;
1.1424 +}
1.1425 +
1.1426 +//--------------------------------------------------------------------------------
1.1427 +//
1.1428 +// inputText()
1.1429 +//
1.1430 +//--------------------------------------------------------------------------------
1.1431 +UText *RegexMatcher::inputText() const {
1.1432 + return fInputText;
1.1433 +}
1.1434 +
1.1435 +
1.1436 +//--------------------------------------------------------------------------------
1.1437 +//
1.1438 +// getInput() -- like inputText(), but makes a clone or copies into another UText
1.1439 +//
1.1440 +//--------------------------------------------------------------------------------
1.1441 +UText *RegexMatcher::getInput (UText *dest, UErrorCode &status) const {
1.1442 + UBool bailOut = FALSE;
1.1443 + if (U_FAILURE(status)) {
1.1444 + return dest;
1.1445 + }
1.1446 + if (U_FAILURE(fDeferredStatus)) {
1.1447 + status = fDeferredStatus;
1.1448 + bailOut = TRUE;
1.1449 + }
1.1450 +
1.1451 + if (bailOut) {
1.1452 + if (dest) {
1.1453 + utext_replace(dest, 0, utext_nativeLength(dest), NULL, 0, &status);
1.1454 + return dest;
1.1455 + } else {
1.1456 + return utext_clone(NULL, fInputText, FALSE, TRUE, &status);
1.1457 + }
1.1458 + }
1.1459 +
1.1460 + if (dest) {
1.1461 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1462 + utext_replace(dest, 0, utext_nativeLength(dest), fInputText->chunkContents, (int32_t)fInputLength, &status);
1.1463 + } else {
1.1464 + int32_t input16Len;
1.1465 + if (UTEXT_USES_U16(fInputText)) {
1.1466 + input16Len = (int32_t)fInputLength;
1.1467 + } else {
1.1468 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.1469 + input16Len = utext_extract(fInputText, 0, fInputLength, NULL, 0, &lengthStatus); // buffer overflow error
1.1470 + }
1.1471 + UChar *inputChars = (UChar *)uprv_malloc(sizeof(UChar)*(input16Len));
1.1472 + if (inputChars == NULL) {
1.1473 + return dest;
1.1474 + }
1.1475 +
1.1476 + status = U_ZERO_ERROR;
1.1477 + utext_extract(fInputText, 0, fInputLength, inputChars, input16Len, &status); // not terminated warning
1.1478 + status = U_ZERO_ERROR;
1.1479 + utext_replace(dest, 0, utext_nativeLength(dest), inputChars, input16Len, &status);
1.1480 +
1.1481 + uprv_free(inputChars);
1.1482 + }
1.1483 + return dest;
1.1484 + } else {
1.1485 + return utext_clone(NULL, fInputText, FALSE, TRUE, &status);
1.1486 + }
1.1487 +}
1.1488 +
1.1489 +
1.1490 +static UBool compat_SyncMutableUTextContents(UText *ut);
1.1491 +static UBool compat_SyncMutableUTextContents(UText *ut) {
1.1492 + UBool retVal = FALSE;
1.1493 +
1.1494 + // In the following test, we're really only interested in whether the UText should switch
1.1495 + // between heap and stack allocation. If length hasn't changed, we won't, so the chunkContents
1.1496 + // will still point to the correct data.
1.1497 + if (utext_nativeLength(ut) != ut->nativeIndexingLimit) {
1.1498 + UnicodeString *us=(UnicodeString *)ut->context;
1.1499 +
1.1500 + // Update to the latest length.
1.1501 + // For example, (utext_nativeLength(ut) != ut->nativeIndexingLimit).
1.1502 + int32_t newLength = us->length();
1.1503 +
1.1504 + // Update the chunk description.
1.1505 + // The buffer may have switched between stack- and heap-based.
1.1506 + ut->chunkContents = us->getBuffer();
1.1507 + ut->chunkLength = newLength;
1.1508 + ut->chunkNativeLimit = newLength;
1.1509 + ut->nativeIndexingLimit = newLength;
1.1510 + retVal = TRUE;
1.1511 + }
1.1512 +
1.1513 + return retVal;
1.1514 +}
1.1515 +
1.1516 +//--------------------------------------------------------------------------------
1.1517 +//
1.1518 +// lookingAt()
1.1519 +//
1.1520 +//--------------------------------------------------------------------------------
1.1521 +UBool RegexMatcher::lookingAt(UErrorCode &status) {
1.1522 + if (U_FAILURE(status)) {
1.1523 + return FALSE;
1.1524 + }
1.1525 + if (U_FAILURE(fDeferredStatus)) {
1.1526 + status = fDeferredStatus;
1.1527 + return FALSE;
1.1528 + }
1.1529 +
1.1530 + if (fInputUniStrMaybeMutable) {
1.1531 + if (compat_SyncMutableUTextContents(fInputText)) {
1.1532 + fInputLength = utext_nativeLength(fInputText);
1.1533 + reset();
1.1534 + }
1.1535 + }
1.1536 + else {
1.1537 + resetPreserveRegion();
1.1538 + }
1.1539 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1540 + MatchChunkAt((int32_t)fActiveStart, FALSE, status);
1.1541 + } else {
1.1542 + MatchAt(fActiveStart, FALSE, status);
1.1543 + }
1.1544 + return fMatch;
1.1545 +}
1.1546 +
1.1547 +
1.1548 +UBool RegexMatcher::lookingAt(int64_t start, UErrorCode &status) {
1.1549 + if (U_FAILURE(status)) {
1.1550 + return FALSE;
1.1551 + }
1.1552 + if (U_FAILURE(fDeferredStatus)) {
1.1553 + status = fDeferredStatus;
1.1554 + return FALSE;
1.1555 + }
1.1556 + reset();
1.1557 +
1.1558 + if (start < 0) {
1.1559 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1560 + return FALSE;
1.1561 + }
1.1562 +
1.1563 + if (fInputUniStrMaybeMutable) {
1.1564 + if (compat_SyncMutableUTextContents(fInputText)) {
1.1565 + fInputLength = utext_nativeLength(fInputText);
1.1566 + reset();
1.1567 + }
1.1568 + }
1.1569 +
1.1570 + int64_t nativeStart;
1.1571 + nativeStart = start;
1.1572 + if (nativeStart < fActiveStart || nativeStart > fActiveLimit) {
1.1573 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1574 + return FALSE;
1.1575 + }
1.1576 +
1.1577 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1578 + MatchChunkAt((int32_t)nativeStart, FALSE, status);
1.1579 + } else {
1.1580 + MatchAt(nativeStart, FALSE, status);
1.1581 + }
1.1582 + return fMatch;
1.1583 +}
1.1584 +
1.1585 +
1.1586 +
1.1587 +//--------------------------------------------------------------------------------
1.1588 +//
1.1589 +// matches()
1.1590 +//
1.1591 +//--------------------------------------------------------------------------------
1.1592 +UBool RegexMatcher::matches(UErrorCode &status) {
1.1593 + if (U_FAILURE(status)) {
1.1594 + return FALSE;
1.1595 + }
1.1596 + if (U_FAILURE(fDeferredStatus)) {
1.1597 + status = fDeferredStatus;
1.1598 + return FALSE;
1.1599 + }
1.1600 +
1.1601 + if (fInputUniStrMaybeMutable) {
1.1602 + if (compat_SyncMutableUTextContents(fInputText)) {
1.1603 + fInputLength = utext_nativeLength(fInputText);
1.1604 + reset();
1.1605 + }
1.1606 + }
1.1607 + else {
1.1608 + resetPreserveRegion();
1.1609 + }
1.1610 +
1.1611 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1612 + MatchChunkAt((int32_t)fActiveStart, TRUE, status);
1.1613 + } else {
1.1614 + MatchAt(fActiveStart, TRUE, status);
1.1615 + }
1.1616 + return fMatch;
1.1617 +}
1.1618 +
1.1619 +
1.1620 +UBool RegexMatcher::matches(int64_t start, UErrorCode &status) {
1.1621 + if (U_FAILURE(status)) {
1.1622 + return FALSE;
1.1623 + }
1.1624 + if (U_FAILURE(fDeferredStatus)) {
1.1625 + status = fDeferredStatus;
1.1626 + return FALSE;
1.1627 + }
1.1628 + reset();
1.1629 +
1.1630 + if (start < 0) {
1.1631 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1632 + return FALSE;
1.1633 + }
1.1634 +
1.1635 + if (fInputUniStrMaybeMutable) {
1.1636 + if (compat_SyncMutableUTextContents(fInputText)) {
1.1637 + fInputLength = utext_nativeLength(fInputText);
1.1638 + reset();
1.1639 + }
1.1640 + }
1.1641 +
1.1642 + int64_t nativeStart;
1.1643 + nativeStart = start;
1.1644 + if (nativeStart < fActiveStart || nativeStart > fActiveLimit) {
1.1645 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1646 + return FALSE;
1.1647 + }
1.1648 +
1.1649 + if (UTEXT_FULL_TEXT_IN_CHUNK(fInputText, fInputLength)) {
1.1650 + MatchChunkAt((int32_t)nativeStart, TRUE, status);
1.1651 + } else {
1.1652 + MatchAt(nativeStart, TRUE, status);
1.1653 + }
1.1654 + return fMatch;
1.1655 +}
1.1656 +
1.1657 +
1.1658 +
1.1659 +//--------------------------------------------------------------------------------
1.1660 +//
1.1661 +// pattern
1.1662 +//
1.1663 +//--------------------------------------------------------------------------------
1.1664 +const RegexPattern &RegexMatcher::pattern() const {
1.1665 + return *fPattern;
1.1666 +}
1.1667 +
1.1668 +
1.1669 +
1.1670 +//--------------------------------------------------------------------------------
1.1671 +//
1.1672 +// region
1.1673 +//
1.1674 +//--------------------------------------------------------------------------------
1.1675 +RegexMatcher &RegexMatcher::region(int64_t regionStart, int64_t regionLimit, int64_t startIndex, UErrorCode &status) {
1.1676 + if (U_FAILURE(status)) {
1.1677 + return *this;
1.1678 + }
1.1679 +
1.1680 + if (regionStart>regionLimit || regionStart<0 || regionLimit<0) {
1.1681 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.1682 + }
1.1683 +
1.1684 + int64_t nativeStart = regionStart;
1.1685 + int64_t nativeLimit = regionLimit;
1.1686 + if (nativeStart > fInputLength || nativeLimit > fInputLength) {
1.1687 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.1688 + }
1.1689 +
1.1690 + if (startIndex == -1)
1.1691 + this->reset();
1.1692 + else
1.1693 + resetPreserveRegion();
1.1694 +
1.1695 + fRegionStart = nativeStart;
1.1696 + fRegionLimit = nativeLimit;
1.1697 + fActiveStart = nativeStart;
1.1698 + fActiveLimit = nativeLimit;
1.1699 +
1.1700 + if (startIndex != -1) {
1.1701 + if (startIndex < fActiveStart || startIndex > fActiveLimit) {
1.1702 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1703 + }
1.1704 + fMatchEnd = startIndex;
1.1705 + }
1.1706 +
1.1707 + if (!fTransparentBounds) {
1.1708 + fLookStart = nativeStart;
1.1709 + fLookLimit = nativeLimit;
1.1710 + }
1.1711 + if (fAnchoringBounds) {
1.1712 + fAnchorStart = nativeStart;
1.1713 + fAnchorLimit = nativeLimit;
1.1714 + }
1.1715 + return *this;
1.1716 +}
1.1717 +
1.1718 +RegexMatcher &RegexMatcher::region(int64_t start, int64_t limit, UErrorCode &status) {
1.1719 + return region(start, limit, -1, status);
1.1720 +}
1.1721 +
1.1722 +//--------------------------------------------------------------------------------
1.1723 +//
1.1724 +// regionEnd
1.1725 +//
1.1726 +//--------------------------------------------------------------------------------
1.1727 +int32_t RegexMatcher::regionEnd() const {
1.1728 + return (int32_t)fRegionLimit;
1.1729 +}
1.1730 +
1.1731 +int64_t RegexMatcher::regionEnd64() const {
1.1732 + return fRegionLimit;
1.1733 +}
1.1734 +
1.1735 +//--------------------------------------------------------------------------------
1.1736 +//
1.1737 +// regionStart
1.1738 +//
1.1739 +//--------------------------------------------------------------------------------
1.1740 +int32_t RegexMatcher::regionStart() const {
1.1741 + return (int32_t)fRegionStart;
1.1742 +}
1.1743 +
1.1744 +int64_t RegexMatcher::regionStart64() const {
1.1745 + return fRegionStart;
1.1746 +}
1.1747 +
1.1748 +
1.1749 +//--------------------------------------------------------------------------------
1.1750 +//
1.1751 +// replaceAll
1.1752 +//
1.1753 +//--------------------------------------------------------------------------------
1.1754 +UnicodeString RegexMatcher::replaceAll(const UnicodeString &replacement, UErrorCode &status) {
1.1755 + UText replacementText = UTEXT_INITIALIZER;
1.1756 + UText resultText = UTEXT_INITIALIZER;
1.1757 + UnicodeString resultString;
1.1758 + if (U_FAILURE(status)) {
1.1759 + return resultString;
1.1760 + }
1.1761 +
1.1762 + utext_openConstUnicodeString(&replacementText, &replacement, &status);
1.1763 + utext_openUnicodeString(&resultText, &resultString, &status);
1.1764 +
1.1765 + replaceAll(&replacementText, &resultText, status);
1.1766 +
1.1767 + utext_close(&resultText);
1.1768 + utext_close(&replacementText);
1.1769 +
1.1770 + return resultString;
1.1771 +}
1.1772 +
1.1773 +
1.1774 +//
1.1775 +// replaceAll, UText mode
1.1776 +//
1.1777 +UText *RegexMatcher::replaceAll(UText *replacement, UText *dest, UErrorCode &status) {
1.1778 + if (U_FAILURE(status)) {
1.1779 + return dest;
1.1780 + }
1.1781 + if (U_FAILURE(fDeferredStatus)) {
1.1782 + status = fDeferredStatus;
1.1783 + return dest;
1.1784 + }
1.1785 +
1.1786 + if (dest == NULL) {
1.1787 + UnicodeString emptyString;
1.1788 + UText empty = UTEXT_INITIALIZER;
1.1789 +
1.1790 + utext_openUnicodeString(&empty, &emptyString, &status);
1.1791 + dest = utext_clone(NULL, &empty, TRUE, FALSE, &status);
1.1792 + utext_close(&empty);
1.1793 + }
1.1794 +
1.1795 + if (U_SUCCESS(status)) {
1.1796 + reset();
1.1797 + while (find()) {
1.1798 + appendReplacement(dest, replacement, status);
1.1799 + if (U_FAILURE(status)) {
1.1800 + break;
1.1801 + }
1.1802 + }
1.1803 + appendTail(dest, status);
1.1804 + }
1.1805 +
1.1806 + return dest;
1.1807 +}
1.1808 +
1.1809 +
1.1810 +//--------------------------------------------------------------------------------
1.1811 +//
1.1812 +// replaceFirst
1.1813 +//
1.1814 +//--------------------------------------------------------------------------------
1.1815 +UnicodeString RegexMatcher::replaceFirst(const UnicodeString &replacement, UErrorCode &status) {
1.1816 + UText replacementText = UTEXT_INITIALIZER;
1.1817 + UText resultText = UTEXT_INITIALIZER;
1.1818 + UnicodeString resultString;
1.1819 +
1.1820 + utext_openConstUnicodeString(&replacementText, &replacement, &status);
1.1821 + utext_openUnicodeString(&resultText, &resultString, &status);
1.1822 +
1.1823 + replaceFirst(&replacementText, &resultText, status);
1.1824 +
1.1825 + utext_close(&resultText);
1.1826 + utext_close(&replacementText);
1.1827 +
1.1828 + return resultString;
1.1829 +}
1.1830 +
1.1831 +//
1.1832 +// replaceFirst, UText mode
1.1833 +//
1.1834 +UText *RegexMatcher::replaceFirst(UText *replacement, UText *dest, UErrorCode &status) {
1.1835 + if (U_FAILURE(status)) {
1.1836 + return dest;
1.1837 + }
1.1838 + if (U_FAILURE(fDeferredStatus)) {
1.1839 + status = fDeferredStatus;
1.1840 + return dest;
1.1841 + }
1.1842 +
1.1843 + reset();
1.1844 + if (!find()) {
1.1845 + return getInput(dest, status);
1.1846 + }
1.1847 +
1.1848 + if (dest == NULL) {
1.1849 + UnicodeString emptyString;
1.1850 + UText empty = UTEXT_INITIALIZER;
1.1851 +
1.1852 + utext_openUnicodeString(&empty, &emptyString, &status);
1.1853 + dest = utext_clone(NULL, &empty, TRUE, FALSE, &status);
1.1854 + utext_close(&empty);
1.1855 + }
1.1856 +
1.1857 + appendReplacement(dest, replacement, status);
1.1858 + appendTail(dest, status);
1.1859 +
1.1860 + return dest;
1.1861 +}
1.1862 +
1.1863 +
1.1864 +//--------------------------------------------------------------------------------
1.1865 +//
1.1866 +// requireEnd
1.1867 +//
1.1868 +//--------------------------------------------------------------------------------
1.1869 +UBool RegexMatcher::requireEnd() const {
1.1870 + return fRequireEnd;
1.1871 +}
1.1872 +
1.1873 +
1.1874 +//--------------------------------------------------------------------------------
1.1875 +//
1.1876 +// reset
1.1877 +//
1.1878 +//--------------------------------------------------------------------------------
1.1879 +RegexMatcher &RegexMatcher::reset() {
1.1880 + fRegionStart = 0;
1.1881 + fRegionLimit = fInputLength;
1.1882 + fActiveStart = 0;
1.1883 + fActiveLimit = fInputLength;
1.1884 + fAnchorStart = 0;
1.1885 + fAnchorLimit = fInputLength;
1.1886 + fLookStart = 0;
1.1887 + fLookLimit = fInputLength;
1.1888 + resetPreserveRegion();
1.1889 + return *this;
1.1890 +}
1.1891 +
1.1892 +
1.1893 +
1.1894 +void RegexMatcher::resetPreserveRegion() {
1.1895 + fMatchStart = 0;
1.1896 + fMatchEnd = 0;
1.1897 + fLastMatchEnd = -1;
1.1898 + fAppendPosition = 0;
1.1899 + fMatch = FALSE;
1.1900 + fHitEnd = FALSE;
1.1901 + fRequireEnd = FALSE;
1.1902 + fTime = 0;
1.1903 + fTickCounter = TIMER_INITIAL_VALUE;
1.1904 + //resetStack(); // more expensive than it looks...
1.1905 +}
1.1906 +
1.1907 +
1.1908 +RegexMatcher &RegexMatcher::reset(const UnicodeString &input) {
1.1909 + fInputText = utext_openConstUnicodeString(fInputText, &input, &fDeferredStatus);
1.1910 + if (fPattern->fNeedsAltInput) {
1.1911 + fAltInputText = utext_clone(fAltInputText, fInputText, FALSE, TRUE, &fDeferredStatus);
1.1912 + }
1.1913 + fInputLength = utext_nativeLength(fInputText);
1.1914 +
1.1915 + reset();
1.1916 + delete fInput;
1.1917 + fInput = NULL;
1.1918 +
1.1919 + // Do the following for any UnicodeString.
1.1920 + // This is for compatibility for those clients who modify the input string "live" during regex operations.
1.1921 + fInputUniStrMaybeMutable = TRUE;
1.1922 +
1.1923 + if (fWordBreakItr != NULL) {
1.1924 +#if UCONFIG_NO_BREAK_ITERATION==0
1.1925 + UErrorCode status = U_ZERO_ERROR;
1.1926 + fWordBreakItr->setText(fInputText, status);
1.1927 +#endif
1.1928 + }
1.1929 + return *this;
1.1930 +}
1.1931 +
1.1932 +
1.1933 +RegexMatcher &RegexMatcher::reset(UText *input) {
1.1934 + if (fInputText != input) {
1.1935 + fInputText = utext_clone(fInputText, input, FALSE, TRUE, &fDeferredStatus);
1.1936 + if (fPattern->fNeedsAltInput) fAltInputText = utext_clone(fAltInputText, fInputText, FALSE, TRUE, &fDeferredStatus);
1.1937 + fInputLength = utext_nativeLength(fInputText);
1.1938 +
1.1939 + delete fInput;
1.1940 + fInput = NULL;
1.1941 +
1.1942 + if (fWordBreakItr != NULL) {
1.1943 +#if UCONFIG_NO_BREAK_ITERATION==0
1.1944 + UErrorCode status = U_ZERO_ERROR;
1.1945 + fWordBreakItr->setText(input, status);
1.1946 +#endif
1.1947 + }
1.1948 + }
1.1949 + reset();
1.1950 + fInputUniStrMaybeMutable = FALSE;
1.1951 +
1.1952 + return *this;
1.1953 +}
1.1954 +
1.1955 +/*RegexMatcher &RegexMatcher::reset(const UChar *) {
1.1956 + fDeferredStatus = U_INTERNAL_PROGRAM_ERROR;
1.1957 + return *this;
1.1958 +}*/
1.1959 +
1.1960 +RegexMatcher &RegexMatcher::reset(int64_t position, UErrorCode &status) {
1.1961 + if (U_FAILURE(status)) {
1.1962 + return *this;
1.1963 + }
1.1964 + reset(); // Reset also resets the region to be the entire string.
1.1965 +
1.1966 + if (position < 0 || position > fActiveLimit) {
1.1967 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.1968 + return *this;
1.1969 + }
1.1970 + fMatchEnd = position;
1.1971 + return *this;
1.1972 +}
1.1973 +
1.1974 +
1.1975 +//--------------------------------------------------------------------------------
1.1976 +//
1.1977 +// refresh
1.1978 +//
1.1979 +//--------------------------------------------------------------------------------
1.1980 +RegexMatcher &RegexMatcher::refreshInputText(UText *input, UErrorCode &status) {
1.1981 + if (U_FAILURE(status)) {
1.1982 + return *this;
1.1983 + }
1.1984 + if (input == NULL) {
1.1985 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.1986 + return *this;
1.1987 + }
1.1988 + if (utext_nativeLength(fInputText) != utext_nativeLength(input)) {
1.1989 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.1990 + return *this;
1.1991 + }
1.1992 + int64_t pos = utext_getNativeIndex(fInputText);
1.1993 + // Shallow read-only clone of the new UText into the existing input UText
1.1994 + fInputText = utext_clone(fInputText, input, FALSE, TRUE, &status);
1.1995 + if (U_FAILURE(status)) {
1.1996 + return *this;
1.1997 + }
1.1998 + utext_setNativeIndex(fInputText, pos);
1.1999 +
1.2000 + if (fAltInputText != NULL) {
1.2001 + pos = utext_getNativeIndex(fAltInputText);
1.2002 + fAltInputText = utext_clone(fAltInputText, input, FALSE, TRUE, &status);
1.2003 + if (U_FAILURE(status)) {
1.2004 + return *this;
1.2005 + }
1.2006 + utext_setNativeIndex(fAltInputText, pos);
1.2007 + }
1.2008 + return *this;
1.2009 +}
1.2010 +
1.2011 +
1.2012 +
1.2013 +//--------------------------------------------------------------------------------
1.2014 +//
1.2015 +// setTrace
1.2016 +//
1.2017 +//--------------------------------------------------------------------------------
1.2018 +void RegexMatcher::setTrace(UBool state) {
1.2019 + fTraceDebug = state;
1.2020 +}
1.2021 +
1.2022 +
1.2023 +
1.2024 +//---------------------------------------------------------------------
1.2025 +//
1.2026 +// split
1.2027 +//
1.2028 +//---------------------------------------------------------------------
1.2029 +int32_t RegexMatcher::split(const UnicodeString &input,
1.2030 + UnicodeString dest[],
1.2031 + int32_t destCapacity,
1.2032 + UErrorCode &status)
1.2033 +{
1.2034 + UText inputText = UTEXT_INITIALIZER;
1.2035 + utext_openConstUnicodeString(&inputText, &input, &status);
1.2036 + if (U_FAILURE(status)) {
1.2037 + return 0;
1.2038 + }
1.2039 +
1.2040 + UText **destText = (UText **)uprv_malloc(sizeof(UText*)*destCapacity);
1.2041 + if (destText == NULL) {
1.2042 + status = U_MEMORY_ALLOCATION_ERROR;
1.2043 + return 0;
1.2044 + }
1.2045 + int32_t i;
1.2046 + for (i = 0; i < destCapacity; i++) {
1.2047 + destText[i] = utext_openUnicodeString(NULL, &dest[i], &status);
1.2048 + }
1.2049 +
1.2050 + int32_t fieldCount = split(&inputText, destText, destCapacity, status);
1.2051 +
1.2052 + for (i = 0; i < destCapacity; i++) {
1.2053 + utext_close(destText[i]);
1.2054 + }
1.2055 +
1.2056 + uprv_free(destText);
1.2057 + utext_close(&inputText);
1.2058 + return fieldCount;
1.2059 +}
1.2060 +
1.2061 +//
1.2062 +// split, UText mode
1.2063 +//
1.2064 +int32_t RegexMatcher::split(UText *input,
1.2065 + UText *dest[],
1.2066 + int32_t destCapacity,
1.2067 + UErrorCode &status)
1.2068 +{
1.2069 + //
1.2070 + // Check arguements for validity
1.2071 + //
1.2072 + if (U_FAILURE(status)) {
1.2073 + return 0;
1.2074 + };
1.2075 +
1.2076 + if (destCapacity < 1) {
1.2077 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.2078 + return 0;
1.2079 + }
1.2080 +
1.2081 + //
1.2082 + // Reset for the input text
1.2083 + //
1.2084 + reset(input);
1.2085 + int64_t nextOutputStringStart = 0;
1.2086 + if (fActiveLimit == 0) {
1.2087 + return 0;
1.2088 + }
1.2089 +
1.2090 + //
1.2091 + // Loop through the input text, searching for the delimiter pattern
1.2092 + //
1.2093 + int32_t i;
1.2094 + int32_t numCaptureGroups = fPattern->fGroupMap->size();
1.2095 + for (i=0; ; i++) {
1.2096 + if (i>=destCapacity-1) {
1.2097 + // There is one or zero output string left.
1.2098 + // Fill the last output string with whatever is left from the input, then exit the loop.
1.2099 + // ( i will be == destCapacity if we filled the output array while processing
1.2100 + // capture groups of the delimiter expression, in which case we will discard the
1.2101 + // last capture group saved in favor of the unprocessed remainder of the
1.2102 + // input string.)
1.2103 + i = destCapacity-1;
1.2104 + if (fActiveLimit > nextOutputStringStart) {
1.2105 + if (UTEXT_FULL_TEXT_IN_CHUNK(input, fInputLength)) {
1.2106 + if (dest[i]) {
1.2107 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]),
1.2108 + input->chunkContents+nextOutputStringStart,
1.2109 + (int32_t)(fActiveLimit-nextOutputStringStart), &status);
1.2110 + } else {
1.2111 + UText remainingText = UTEXT_INITIALIZER;
1.2112 + utext_openUChars(&remainingText, input->chunkContents+nextOutputStringStart,
1.2113 + fActiveLimit-nextOutputStringStart, &status);
1.2114 + dest[i] = utext_clone(NULL, &remainingText, TRUE, FALSE, &status);
1.2115 + utext_close(&remainingText);
1.2116 + }
1.2117 + } else {
1.2118 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.2119 + int32_t remaining16Length =
1.2120 + utext_extract(input, nextOutputStringStart, fActiveLimit, NULL, 0, &lengthStatus);
1.2121 + UChar *remainingChars = (UChar *)uprv_malloc(sizeof(UChar)*(remaining16Length+1));
1.2122 + if (remainingChars == NULL) {
1.2123 + status = U_MEMORY_ALLOCATION_ERROR;
1.2124 + break;
1.2125 + }
1.2126 +
1.2127 + utext_extract(input, nextOutputStringStart, fActiveLimit, remainingChars, remaining16Length+1, &status);
1.2128 + if (dest[i]) {
1.2129 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]), remainingChars, remaining16Length, &status);
1.2130 + } else {
1.2131 + UText remainingText = UTEXT_INITIALIZER;
1.2132 + utext_openUChars(&remainingText, remainingChars, remaining16Length, &status);
1.2133 + dest[i] = utext_clone(NULL, &remainingText, TRUE, FALSE, &status);
1.2134 + utext_close(&remainingText);
1.2135 + }
1.2136 +
1.2137 + uprv_free(remainingChars);
1.2138 + }
1.2139 + }
1.2140 + break;
1.2141 + }
1.2142 + if (find()) {
1.2143 + // We found another delimiter. Move everything from where we started looking
1.2144 + // up until the start of the delimiter into the next output string.
1.2145 + if (UTEXT_FULL_TEXT_IN_CHUNK(input, fInputLength)) {
1.2146 + if (dest[i]) {
1.2147 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]),
1.2148 + input->chunkContents+nextOutputStringStart,
1.2149 + (int32_t)(fMatchStart-nextOutputStringStart), &status);
1.2150 + } else {
1.2151 + UText remainingText = UTEXT_INITIALIZER;
1.2152 + utext_openUChars(&remainingText, input->chunkContents+nextOutputStringStart,
1.2153 + fMatchStart-nextOutputStringStart, &status);
1.2154 + dest[i] = utext_clone(NULL, &remainingText, TRUE, FALSE, &status);
1.2155 + utext_close(&remainingText);
1.2156 + }
1.2157 + } else {
1.2158 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.2159 + int32_t remaining16Length = utext_extract(input, nextOutputStringStart, fMatchStart, NULL, 0, &lengthStatus);
1.2160 + UChar *remainingChars = (UChar *)uprv_malloc(sizeof(UChar)*(remaining16Length+1));
1.2161 + if (remainingChars == NULL) {
1.2162 + status = U_MEMORY_ALLOCATION_ERROR;
1.2163 + break;
1.2164 + }
1.2165 + utext_extract(input, nextOutputStringStart, fMatchStart, remainingChars, remaining16Length+1, &status);
1.2166 + if (dest[i]) {
1.2167 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]), remainingChars, remaining16Length, &status);
1.2168 + } else {
1.2169 + UText remainingText = UTEXT_INITIALIZER;
1.2170 + utext_openUChars(&remainingText, remainingChars, remaining16Length, &status);
1.2171 + dest[i] = utext_clone(NULL, &remainingText, TRUE, FALSE, &status);
1.2172 + utext_close(&remainingText);
1.2173 + }
1.2174 +
1.2175 + uprv_free(remainingChars);
1.2176 + }
1.2177 + nextOutputStringStart = fMatchEnd;
1.2178 +
1.2179 + // If the delimiter pattern has capturing parentheses, the captured
1.2180 + // text goes out into the next n destination strings.
1.2181 + int32_t groupNum;
1.2182 + for (groupNum=1; groupNum<=numCaptureGroups; groupNum++) {
1.2183 + if (i >= destCapacity-2) {
1.2184 + // Never fill the last available output string with capture group text.
1.2185 + // It will filled with the last field, the remainder of the
1.2186 + // unsplit input text.
1.2187 + break;
1.2188 + }
1.2189 + i++;
1.2190 + dest[i] = group(groupNum, dest[i], status);
1.2191 + }
1.2192 +
1.2193 + if (nextOutputStringStart == fActiveLimit) {
1.2194 + // The delimiter was at the end of the string. We're done, but first
1.2195 + // we output one last empty string, for the empty field following
1.2196 + // the delimiter at the end of input.
1.2197 + if (i+1 < destCapacity) {
1.2198 + ++i;
1.2199 + if (dest[i] == NULL) {
1.2200 + dest[i] = utext_openUChars(NULL, NULL, 0, &status);
1.2201 + } else {
1.2202 + static UChar emptyString[] = {(UChar)0};
1.2203 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]), emptyString, 0, &status);
1.2204 + }
1.2205 + }
1.2206 + break;
1.2207 +
1.2208 + }
1.2209 + }
1.2210 + else
1.2211 + {
1.2212 + // We ran off the end of the input while looking for the next delimiter.
1.2213 + // All the remaining text goes into the current output string.
1.2214 + if (UTEXT_FULL_TEXT_IN_CHUNK(input, fInputLength)) {
1.2215 + if (dest[i]) {
1.2216 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]),
1.2217 + input->chunkContents+nextOutputStringStart,
1.2218 + (int32_t)(fActiveLimit-nextOutputStringStart), &status);
1.2219 + } else {
1.2220 + UText remainingText = UTEXT_INITIALIZER;
1.2221 + utext_openUChars(&remainingText, input->chunkContents+nextOutputStringStart,
1.2222 + fActiveLimit-nextOutputStringStart, &status);
1.2223 + dest[i] = utext_clone(NULL, &remainingText, TRUE, FALSE, &status);
1.2224 + utext_close(&remainingText);
1.2225 + }
1.2226 + } else {
1.2227 + UErrorCode lengthStatus = U_ZERO_ERROR;
1.2228 + int32_t remaining16Length = utext_extract(input, nextOutputStringStart, fActiveLimit, NULL, 0, &lengthStatus);
1.2229 + UChar *remainingChars = (UChar *)uprv_malloc(sizeof(UChar)*(remaining16Length+1));
1.2230 + if (remainingChars == NULL) {
1.2231 + status = U_MEMORY_ALLOCATION_ERROR;
1.2232 + break;
1.2233 + }
1.2234 +
1.2235 + utext_extract(input, nextOutputStringStart, fActiveLimit, remainingChars, remaining16Length+1, &status);
1.2236 + if (dest[i]) {
1.2237 + utext_replace(dest[i], 0, utext_nativeLength(dest[i]), remainingChars, remaining16Length, &status);
1.2238 + } else {
1.2239 + UText remainingText = UTEXT_INITIALIZER;
1.2240 + utext_openUChars(&remainingText, remainingChars, remaining16Length, &status);
1.2241 + dest[i] = utext_clone(NULL, &remainingText, TRUE, FALSE, &status);
1.2242 + utext_close(&remainingText);
1.2243 + }
1.2244 +
1.2245 + uprv_free(remainingChars);
1.2246 + }
1.2247 + break;
1.2248 + }
1.2249 + if (U_FAILURE(status)) {
1.2250 + break;
1.2251 + }
1.2252 + } // end of for loop
1.2253 + return i+1;
1.2254 +}
1.2255 +
1.2256 +
1.2257 +//--------------------------------------------------------------------------------
1.2258 +//
1.2259 +// start
1.2260 +//
1.2261 +//--------------------------------------------------------------------------------
1.2262 +int32_t RegexMatcher::start(UErrorCode &status) const {
1.2263 + return start(0, status);
1.2264 +}
1.2265 +
1.2266 +int64_t RegexMatcher::start64(UErrorCode &status) const {
1.2267 + return start64(0, status);
1.2268 +}
1.2269 +
1.2270 +//--------------------------------------------------------------------------------
1.2271 +//
1.2272 +// start(int32_t group, UErrorCode &status)
1.2273 +//
1.2274 +//--------------------------------------------------------------------------------
1.2275 +
1.2276 +int64_t RegexMatcher::start64(int32_t group, UErrorCode &status) const {
1.2277 + if (U_FAILURE(status)) {
1.2278 + return -1;
1.2279 + }
1.2280 + if (U_FAILURE(fDeferredStatus)) {
1.2281 + status = fDeferredStatus;
1.2282 + return -1;
1.2283 + }
1.2284 + if (fMatch == FALSE) {
1.2285 + status = U_REGEX_INVALID_STATE;
1.2286 + return -1;
1.2287 + }
1.2288 + if (group < 0 || group > fPattern->fGroupMap->size()) {
1.2289 + status = U_INDEX_OUTOFBOUNDS_ERROR;
1.2290 + return -1;
1.2291 + }
1.2292 + int64_t s;
1.2293 + if (group == 0) {
1.2294 + s = fMatchStart;
1.2295 + } else {
1.2296 + int32_t groupOffset = fPattern->fGroupMap->elementAti(group-1);
1.2297 + U_ASSERT(groupOffset < fPattern->fFrameSize);
1.2298 + U_ASSERT(groupOffset >= 0);
1.2299 + s = fFrame->fExtra[groupOffset];
1.2300 + }
1.2301 +
1.2302 + return s;
1.2303 +}
1.2304 +
1.2305 +
1.2306 +int32_t RegexMatcher::start(int32_t group, UErrorCode &status) const {
1.2307 + return (int32_t)start64(group, status);
1.2308 +}
1.2309 +
1.2310 +//--------------------------------------------------------------------------------
1.2311 +//
1.2312 +// useAnchoringBounds
1.2313 +//
1.2314 +//--------------------------------------------------------------------------------
1.2315 +RegexMatcher &RegexMatcher::useAnchoringBounds(UBool b) {
1.2316 + fAnchoringBounds = b;
1.2317 + fAnchorStart = (fAnchoringBounds ? fRegionStart : 0);
1.2318 + fAnchorLimit = (fAnchoringBounds ? fRegionLimit : fInputLength);
1.2319 + return *this;
1.2320 +}
1.2321 +
1.2322 +
1.2323 +//--------------------------------------------------------------------------------
1.2324 +//
1.2325 +// useTransparentBounds
1.2326 +//
1.2327 +//--------------------------------------------------------------------------------
1.2328 +RegexMatcher &RegexMatcher::useTransparentBounds(UBool b) {
1.2329 + fTransparentBounds = b;
1.2330 + fLookStart = (fTransparentBounds ? 0 : fRegionStart);
1.2331 + fLookLimit = (fTransparentBounds ? fInputLength : fRegionLimit);
1.2332 + return *this;
1.2333 +}
1.2334 +
1.2335 +//--------------------------------------------------------------------------------
1.2336 +//
1.2337 +// setTimeLimit
1.2338 +//
1.2339 +//--------------------------------------------------------------------------------
1.2340 +void RegexMatcher::setTimeLimit(int32_t limit, UErrorCode &status) {
1.2341 + if (U_FAILURE(status)) {
1.2342 + return;
1.2343 + }
1.2344 + if (U_FAILURE(fDeferredStatus)) {
1.2345 + status = fDeferredStatus;
1.2346 + return;
1.2347 + }
1.2348 + if (limit < 0) {
1.2349 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.2350 + return;
1.2351 + }
1.2352 + fTimeLimit = limit;
1.2353 +}
1.2354 +
1.2355 +
1.2356 +//--------------------------------------------------------------------------------
1.2357 +//
1.2358 +// getTimeLimit
1.2359 +//
1.2360 +//--------------------------------------------------------------------------------
1.2361 +int32_t RegexMatcher::getTimeLimit() const {
1.2362 + return fTimeLimit;
1.2363 +}
1.2364 +
1.2365 +
1.2366 +//--------------------------------------------------------------------------------
1.2367 +//
1.2368 +// setStackLimit
1.2369 +//
1.2370 +//--------------------------------------------------------------------------------
1.2371 +void RegexMatcher::setStackLimit(int32_t limit, UErrorCode &status) {
1.2372 + if (U_FAILURE(status)) {
1.2373 + return;
1.2374 + }
1.2375 + if (U_FAILURE(fDeferredStatus)) {
1.2376 + status = fDeferredStatus;
1.2377 + return;
1.2378 + }
1.2379 + if (limit < 0) {
1.2380 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.2381 + return;
1.2382 + }
1.2383 +
1.2384 + // Reset the matcher. This is needed here in case there is a current match
1.2385 + // whose final stack frame (containing the match results, pointed to by fFrame)
1.2386 + // would be lost by resizing to a smaller stack size.
1.2387 + reset();
1.2388 +
1.2389 + if (limit == 0) {
1.2390 + // Unlimited stack expansion
1.2391 + fStack->setMaxCapacity(0);
1.2392 + } else {
1.2393 + // Change the units of the limit from bytes to ints, and bump the size up
1.2394 + // to be big enough to hold at least one stack frame for the pattern,
1.2395 + // if it isn't there already.
1.2396 + int32_t adjustedLimit = limit / sizeof(int32_t);
1.2397 + if (adjustedLimit < fPattern->fFrameSize) {
1.2398 + adjustedLimit = fPattern->fFrameSize;
1.2399 + }
1.2400 + fStack->setMaxCapacity(adjustedLimit);
1.2401 + }
1.2402 + fStackLimit = limit;
1.2403 +}
1.2404 +
1.2405 +
1.2406 +//--------------------------------------------------------------------------------
1.2407 +//
1.2408 +// getStackLimit
1.2409 +//
1.2410 +//--------------------------------------------------------------------------------
1.2411 +int32_t RegexMatcher::getStackLimit() const {
1.2412 + return fStackLimit;
1.2413 +}
1.2414 +
1.2415 +
1.2416 +//--------------------------------------------------------------------------------
1.2417 +//
1.2418 +// setMatchCallback
1.2419 +//
1.2420 +//--------------------------------------------------------------------------------
1.2421 +void RegexMatcher::setMatchCallback(URegexMatchCallback *callback,
1.2422 + const void *context,
1.2423 + UErrorCode &status) {
1.2424 + if (U_FAILURE(status)) {
1.2425 + return;
1.2426 + }
1.2427 + fCallbackFn = callback;
1.2428 + fCallbackContext = context;
1.2429 +}
1.2430 +
1.2431 +
1.2432 +//--------------------------------------------------------------------------------
1.2433 +//
1.2434 +// getMatchCallback
1.2435 +//
1.2436 +//--------------------------------------------------------------------------------
1.2437 +void RegexMatcher::getMatchCallback(URegexMatchCallback *&callback,
1.2438 + const void *&context,
1.2439 + UErrorCode &status) {
1.2440 + if (U_FAILURE(status)) {
1.2441 + return;
1.2442 + }
1.2443 + callback = fCallbackFn;
1.2444 + context = fCallbackContext;
1.2445 +}
1.2446 +
1.2447 +
1.2448 +//--------------------------------------------------------------------------------
1.2449 +//
1.2450 +// setMatchCallback
1.2451 +//
1.2452 +//--------------------------------------------------------------------------------
1.2453 +void RegexMatcher::setFindProgressCallback(URegexFindProgressCallback *callback,
1.2454 + const void *context,
1.2455 + UErrorCode &status) {
1.2456 + if (U_FAILURE(status)) {
1.2457 + return;
1.2458 + }
1.2459 + fFindProgressCallbackFn = callback;
1.2460 + fFindProgressCallbackContext = context;
1.2461 +}
1.2462 +
1.2463 +
1.2464 +//--------------------------------------------------------------------------------
1.2465 +//
1.2466 +// getMatchCallback
1.2467 +//
1.2468 +//--------------------------------------------------------------------------------
1.2469 +void RegexMatcher::getFindProgressCallback(URegexFindProgressCallback *&callback,
1.2470 + const void *&context,
1.2471 + UErrorCode &status) {
1.2472 + if (U_FAILURE(status)) {
1.2473 + return;
1.2474 + }
1.2475 + callback = fFindProgressCallbackFn;
1.2476 + context = fFindProgressCallbackContext;
1.2477 +}
1.2478 +
1.2479 +
1.2480 +//================================================================================
1.2481 +//
1.2482 +// Code following this point in this file is the internal
1.2483 +// Match Engine Implementation.
1.2484 +//
1.2485 +//================================================================================
1.2486 +
1.2487 +
1.2488 +//--------------------------------------------------------------------------------
1.2489 +//
1.2490 +// resetStack
1.2491 +// Discard any previous contents of the state save stack, and initialize a
1.2492 +// new stack frame to all -1. The -1s are needed for capture group limits,
1.2493 +// where they indicate that a group has not yet matched anything.
1.2494 +//--------------------------------------------------------------------------------
1.2495 +REStackFrame *RegexMatcher::resetStack() {
1.2496 + // Discard any previous contents of the state save stack, and initialize a
1.2497 + // new stack frame with all -1 data. The -1s are needed for capture group limits,
1.2498 + // where they indicate that a group has not yet matched anything.
1.2499 + fStack->removeAllElements();
1.2500 +
1.2501 + REStackFrame *iFrame = (REStackFrame *)fStack->reserveBlock(fPattern->fFrameSize, fDeferredStatus);
1.2502 + int32_t i;
1.2503 + for (i=0; i<fPattern->fFrameSize-RESTACKFRAME_HDRCOUNT; i++) {
1.2504 + iFrame->fExtra[i] = -1;
1.2505 + }
1.2506 + return iFrame;
1.2507 +}
1.2508 +
1.2509 +
1.2510 +
1.2511 +//--------------------------------------------------------------------------------
1.2512 +//
1.2513 +// isWordBoundary
1.2514 +// in perl, "xab..cd..", \b is true at positions 0,3,5,7
1.2515 +// For us,
1.2516 +// If the current char is a combining mark,
1.2517 +// \b is FALSE.
1.2518 +// Else Scan backwards to the first non-combining char.
1.2519 +// We are at a boundary if the this char and the original chars are
1.2520 +// opposite in membership in \w set
1.2521 +//
1.2522 +// parameters: pos - the current position in the input buffer
1.2523 +//
1.2524 +// TODO: double-check edge cases at region boundaries.
1.2525 +//
1.2526 +//--------------------------------------------------------------------------------
1.2527 +UBool RegexMatcher::isWordBoundary(int64_t pos) {
1.2528 + UBool isBoundary = FALSE;
1.2529 + UBool cIsWord = FALSE;
1.2530 +
1.2531 + if (pos >= fLookLimit) {
1.2532 + fHitEnd = TRUE;
1.2533 + } else {
1.2534 + // Determine whether char c at current position is a member of the word set of chars.
1.2535 + // If we're off the end of the string, behave as though we're not at a word char.
1.2536 + UTEXT_SETNATIVEINDEX(fInputText, pos);
1.2537 + UChar32 c = UTEXT_CURRENT32(fInputText);
1.2538 + if (u_hasBinaryProperty(c, UCHAR_GRAPHEME_EXTEND) || u_charType(c) == U_FORMAT_CHAR) {
1.2539 + // Current char is a combining one. Not a boundary.
1.2540 + return FALSE;
1.2541 + }
1.2542 + cIsWord = fPattern->fStaticSets[URX_ISWORD_SET]->contains(c);
1.2543 + }
1.2544 +
1.2545 + // Back up until we come to a non-combining char, determine whether
1.2546 + // that char is a word char.
1.2547 + UBool prevCIsWord = FALSE;
1.2548 + for (;;) {
1.2549 + if (UTEXT_GETNATIVEINDEX(fInputText) <= fLookStart) {
1.2550 + break;
1.2551 + }
1.2552 + UChar32 prevChar = UTEXT_PREVIOUS32(fInputText);
1.2553 + if (!(u_hasBinaryProperty(prevChar, UCHAR_GRAPHEME_EXTEND)
1.2554 + || u_charType(prevChar) == U_FORMAT_CHAR)) {
1.2555 + prevCIsWord = fPattern->fStaticSets[URX_ISWORD_SET]->contains(prevChar);
1.2556 + break;
1.2557 + }
1.2558 + }
1.2559 + isBoundary = cIsWord ^ prevCIsWord;
1.2560 + return isBoundary;
1.2561 +}
1.2562 +
1.2563 +UBool RegexMatcher::isChunkWordBoundary(int32_t pos) {
1.2564 + UBool isBoundary = FALSE;
1.2565 + UBool cIsWord = FALSE;
1.2566 +
1.2567 + const UChar *inputBuf = fInputText->chunkContents;
1.2568 +
1.2569 + if (pos >= fLookLimit) {
1.2570 + fHitEnd = TRUE;
1.2571 + } else {
1.2572 + // Determine whether char c at current position is a member of the word set of chars.
1.2573 + // If we're off the end of the string, behave as though we're not at a word char.
1.2574 + UChar32 c;
1.2575 + U16_GET(inputBuf, fLookStart, pos, fLookLimit, c);
1.2576 + if (u_hasBinaryProperty(c, UCHAR_GRAPHEME_EXTEND) || u_charType(c) == U_FORMAT_CHAR) {
1.2577 + // Current char is a combining one. Not a boundary.
1.2578 + return FALSE;
1.2579 + }
1.2580 + cIsWord = fPattern->fStaticSets[URX_ISWORD_SET]->contains(c);
1.2581 + }
1.2582 +
1.2583 + // Back up until we come to a non-combining char, determine whether
1.2584 + // that char is a word char.
1.2585 + UBool prevCIsWord = FALSE;
1.2586 + for (;;) {
1.2587 + if (pos <= fLookStart) {
1.2588 + break;
1.2589 + }
1.2590 + UChar32 prevChar;
1.2591 + U16_PREV(inputBuf, fLookStart, pos, prevChar);
1.2592 + if (!(u_hasBinaryProperty(prevChar, UCHAR_GRAPHEME_EXTEND)
1.2593 + || u_charType(prevChar) == U_FORMAT_CHAR)) {
1.2594 + prevCIsWord = fPattern->fStaticSets[URX_ISWORD_SET]->contains(prevChar);
1.2595 + break;
1.2596 + }
1.2597 + }
1.2598 + isBoundary = cIsWord ^ prevCIsWord;
1.2599 + return isBoundary;
1.2600 +}
1.2601 +
1.2602 +//--------------------------------------------------------------------------------
1.2603 +//
1.2604 +// isUWordBoundary
1.2605 +//
1.2606 +// Test for a word boundary using RBBI word break.
1.2607 +//
1.2608 +// parameters: pos - the current position in the input buffer
1.2609 +//
1.2610 +//--------------------------------------------------------------------------------
1.2611 +UBool RegexMatcher::isUWordBoundary(int64_t pos) {
1.2612 + UBool returnVal = FALSE;
1.2613 +#if UCONFIG_NO_BREAK_ITERATION==0
1.2614 +
1.2615 + // If we haven't yet created a break iterator for this matcher, do it now.
1.2616 + if (fWordBreakItr == NULL) {
1.2617 + fWordBreakItr =
1.2618 + (RuleBasedBreakIterator *)BreakIterator::createWordInstance(Locale::getEnglish(), fDeferredStatus);
1.2619 + if (U_FAILURE(fDeferredStatus)) {
1.2620 + return FALSE;
1.2621 + }
1.2622 + fWordBreakItr->setText(fInputText, fDeferredStatus);
1.2623 + }
1.2624 +
1.2625 + if (pos >= fLookLimit) {
1.2626 + fHitEnd = TRUE;
1.2627 + returnVal = TRUE; // With Unicode word rules, only positions within the interior of "real"
1.2628 + // words are not boundaries. All non-word chars stand by themselves,
1.2629 + // with word boundaries on both sides.
1.2630 + } else {
1.2631 + if (!UTEXT_USES_U16(fInputText)) {
1.2632 + // !!!: Would like a better way to do this!
1.2633 + UErrorCode status = U_ZERO_ERROR;
1.2634 + pos = utext_extract(fInputText, 0, pos, NULL, 0, &status);
1.2635 + }
1.2636 + returnVal = fWordBreakItr->isBoundary((int32_t)pos);
1.2637 + }
1.2638 +#endif
1.2639 + return returnVal;
1.2640 +}
1.2641 +
1.2642 +//--------------------------------------------------------------------------------
1.2643 +//
1.2644 +// IncrementTime This function is called once each TIMER_INITIAL_VALUE state
1.2645 +// saves. Increment the "time" counter, and call the
1.2646 +// user callback function if there is one installed.
1.2647 +//
1.2648 +// If the match operation needs to be aborted, either for a time-out
1.2649 +// or because the user callback asked for it, just set an error status.
1.2650 +// The engine will pick that up and stop in its outer loop.
1.2651 +//
1.2652 +//--------------------------------------------------------------------------------
1.2653 +void RegexMatcher::IncrementTime(UErrorCode &status) {
1.2654 + fTickCounter = TIMER_INITIAL_VALUE;
1.2655 + fTime++;
1.2656 + if (fCallbackFn != NULL) {
1.2657 + if ((*fCallbackFn)(fCallbackContext, fTime) == FALSE) {
1.2658 + status = U_REGEX_STOPPED_BY_CALLER;
1.2659 + return;
1.2660 + }
1.2661 + }
1.2662 + if (fTimeLimit > 0 && fTime >= fTimeLimit) {
1.2663 + status = U_REGEX_TIME_OUT;
1.2664 + }
1.2665 +}
1.2666 +
1.2667 +//--------------------------------------------------------------------------------
1.2668 +//
1.2669 +// ReportFindProgress This function is called once for each advance in the target
1.2670 +// string from the find() function, and calls the user progress callback
1.2671 +// function if there is one installed.
1.2672 +//
1.2673 +// NOTE:
1.2674 +//
1.2675 +// If the match operation needs to be aborted because the user
1.2676 +// callback asked for it, just set an error status.
1.2677 +// The engine will pick that up and stop in its outer loop.
1.2678 +//
1.2679 +//--------------------------------------------------------------------------------
1.2680 +UBool RegexMatcher::ReportFindProgress(int64_t matchIndex, UErrorCode &status) {
1.2681 + if (fFindProgressCallbackFn != NULL) {
1.2682 + if ((*fFindProgressCallbackFn)(fFindProgressCallbackContext, matchIndex) == FALSE) {
1.2683 + status = U_ZERO_ERROR /*U_REGEX_STOPPED_BY_CALLER*/;
1.2684 + return FALSE;
1.2685 + }
1.2686 + }
1.2687 + return TRUE;
1.2688 +}
1.2689 +
1.2690 +//--------------------------------------------------------------------------------
1.2691 +//
1.2692 +// StateSave
1.2693 +// Make a new stack frame, initialized as a copy of the current stack frame.
1.2694 +// Set the pattern index in the original stack frame from the operand value
1.2695 +// in the opcode. Execution of the engine continues with the state in
1.2696 +// the newly created stack frame
1.2697 +//
1.2698 +// Note that reserveBlock() may grow the stack, resulting in the
1.2699 +// whole thing being relocated in memory.
1.2700 +//
1.2701 +// Parameters:
1.2702 +// fp The top frame pointer when called. At return, a new
1.2703 +// fame will be present
1.2704 +// savePatIdx An index into the compiled pattern. Goes into the original
1.2705 +// (not new) frame. If execution ever back-tracks out of the
1.2706 +// new frame, this will be where we continue from in the pattern.
1.2707 +// Return
1.2708 +// The new frame pointer.
1.2709 +//
1.2710 +//--------------------------------------------------------------------------------
1.2711 +inline REStackFrame *RegexMatcher::StateSave(REStackFrame *fp, int64_t savePatIdx, UErrorCode &status) {
1.2712 + // push storage for a new frame.
1.2713 + int64_t *newFP = fStack->reserveBlock(fFrameSize, status);
1.2714 + if (newFP == NULL) {
1.2715 + // Failure on attempted stack expansion.
1.2716 + // Stack function set some other error code, change it to a more
1.2717 + // specific one for regular expressions.
1.2718 + status = U_REGEX_STACK_OVERFLOW;
1.2719 + // We need to return a writable stack frame, so just return the
1.2720 + // previous frame. The match operation will stop quickly
1.2721 + // because of the error status, after which the frame will never
1.2722 + // be looked at again.
1.2723 + return fp;
1.2724 + }
1.2725 + fp = (REStackFrame *)(newFP - fFrameSize); // in case of realloc of stack.
1.2726 +
1.2727 + // New stack frame = copy of old top frame.
1.2728 + int64_t *source = (int64_t *)fp;
1.2729 + int64_t *dest = newFP;
1.2730 + for (;;) {
1.2731 + *dest++ = *source++;
1.2732 + if (source == newFP) {
1.2733 + break;
1.2734 + }
1.2735 + }
1.2736 +
1.2737 + fTickCounter--;
1.2738 + if (fTickCounter <= 0) {
1.2739 + IncrementTime(status); // Re-initializes fTickCounter
1.2740 + }
1.2741 + fp->fPatIdx = savePatIdx;
1.2742 + return (REStackFrame *)newFP;
1.2743 +}
1.2744 +
1.2745 +
1.2746 +//--------------------------------------------------------------------------------
1.2747 +//
1.2748 +// MatchAt This is the actual matching engine.
1.2749 +//
1.2750 +// startIdx: begin matching a this index.
1.2751 +// toEnd: if true, match must extend to end of the input region
1.2752 +//
1.2753 +//--------------------------------------------------------------------------------
1.2754 +void RegexMatcher::MatchAt(int64_t startIdx, UBool toEnd, UErrorCode &status) {
1.2755 + UBool isMatch = FALSE; // True if the we have a match.
1.2756 +
1.2757 + int64_t backSearchIndex = U_INT64_MAX; // used after greedy single-character matches for searching backwards
1.2758 +
1.2759 + int32_t op; // Operation from the compiled pattern, split into
1.2760 + int32_t opType; // the opcode
1.2761 + int32_t opValue; // and the operand value.
1.2762 +
1.2763 + #ifdef REGEX_RUN_DEBUG
1.2764 + if (fTraceDebug)
1.2765 + {
1.2766 + printf("MatchAt(startIdx=%ld)\n", startIdx);
1.2767 + printf("Original Pattern: ");
1.2768 + UChar32 c = utext_next32From(fPattern->fPattern, 0);
1.2769 + while (c != U_SENTINEL) {
1.2770 + if (c<32 || c>256) {
1.2771 + c = '.';
1.2772 + }
1.2773 + REGEX_DUMP_DEBUG_PRINTF(("%c", c));
1.2774 +
1.2775 + c = UTEXT_NEXT32(fPattern->fPattern);
1.2776 + }
1.2777 + printf("\n");
1.2778 + printf("Input String: ");
1.2779 + c = utext_next32From(fInputText, 0);
1.2780 + while (c != U_SENTINEL) {
1.2781 + if (c<32 || c>256) {
1.2782 + c = '.';
1.2783 + }
1.2784 + printf("%c", c);
1.2785 +
1.2786 + c = UTEXT_NEXT32(fInputText);
1.2787 + }
1.2788 + printf("\n");
1.2789 + printf("\n");
1.2790 + }
1.2791 + #endif
1.2792 +
1.2793 + if (U_FAILURE(status)) {
1.2794 + return;
1.2795 + }
1.2796 +
1.2797 + // Cache frequently referenced items from the compiled pattern
1.2798 + //
1.2799 + int64_t *pat = fPattern->fCompiledPat->getBuffer();
1.2800 +
1.2801 + const UChar *litText = fPattern->fLiteralText.getBuffer();
1.2802 + UVector *sets = fPattern->fSets;
1.2803 +
1.2804 + fFrameSize = fPattern->fFrameSize;
1.2805 + REStackFrame *fp = resetStack();
1.2806 +
1.2807 + fp->fPatIdx = 0;
1.2808 + fp->fInputIdx = startIdx;
1.2809 +
1.2810 + // Zero out the pattern's static data
1.2811 + int32_t i;
1.2812 + for (i = 0; i<fPattern->fDataSize; i++) {
1.2813 + fData[i] = 0;
1.2814 + }
1.2815 +
1.2816 + //
1.2817 + // Main loop for interpreting the compiled pattern.
1.2818 + // One iteration of the loop per pattern operation performed.
1.2819 + //
1.2820 + for (;;) {
1.2821 +#if 0
1.2822 + if (_heapchk() != _HEAPOK) {
1.2823 + fprintf(stderr, "Heap Trouble\n");
1.2824 + }
1.2825 +#endif
1.2826 +
1.2827 + op = (int32_t)pat[fp->fPatIdx];
1.2828 + opType = URX_TYPE(op);
1.2829 + opValue = URX_VAL(op);
1.2830 + #ifdef REGEX_RUN_DEBUG
1.2831 + if (fTraceDebug) {
1.2832 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.2833 + printf("inputIdx=%ld inputChar=%x sp=%3ld activeLimit=%ld ", fp->fInputIdx,
1.2834 + UTEXT_CURRENT32(fInputText), (int64_t *)fp-fStack->getBuffer(), fActiveLimit);
1.2835 + fPattern->dumpOp(fp->fPatIdx);
1.2836 + }
1.2837 + #endif
1.2838 + fp->fPatIdx++;
1.2839 +
1.2840 + switch (opType) {
1.2841 +
1.2842 +
1.2843 + case URX_NOP:
1.2844 + break;
1.2845 +
1.2846 +
1.2847 + case URX_BACKTRACK:
1.2848 + // Force a backtrack. In some circumstances, the pattern compiler
1.2849 + // will notice that the pattern can't possibly match anything, and will
1.2850 + // emit one of these at that point.
1.2851 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.2852 + break;
1.2853 +
1.2854 +
1.2855 + case URX_ONECHAR:
1.2856 + if (fp->fInputIdx < fActiveLimit) {
1.2857 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.2858 + UChar32 c = UTEXT_NEXT32(fInputText);
1.2859 + if (c == opValue) {
1.2860 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.2861 + break;
1.2862 + }
1.2863 + } else {
1.2864 + fHitEnd = TRUE;
1.2865 + }
1.2866 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.2867 + break;
1.2868 +
1.2869 +
1.2870 + case URX_STRING:
1.2871 + {
1.2872 + // Test input against a literal string.
1.2873 + // Strings require two slots in the compiled pattern, one for the
1.2874 + // offset to the string text, and one for the length.
1.2875 +
1.2876 + int32_t stringStartIdx = opValue;
1.2877 + op = (int32_t)pat[fp->fPatIdx]; // Fetch the second operand
1.2878 + fp->fPatIdx++;
1.2879 + opType = URX_TYPE(op);
1.2880 + int32_t stringLen = URX_VAL(op);
1.2881 + U_ASSERT(opType == URX_STRING_LEN);
1.2882 + U_ASSERT(stringLen >= 2);
1.2883 +
1.2884 + const UChar *patternString = litText+stringStartIdx;
1.2885 + int32_t patternStringIndex = 0;
1.2886 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.2887 + UChar32 inputChar;
1.2888 + UChar32 patternChar;
1.2889 + UBool success = TRUE;
1.2890 + while (patternStringIndex < stringLen) {
1.2891 + if (UTEXT_GETNATIVEINDEX(fInputText) >= fActiveLimit) {
1.2892 + success = FALSE;
1.2893 + fHitEnd = TRUE;
1.2894 + break;
1.2895 + }
1.2896 + inputChar = UTEXT_NEXT32(fInputText);
1.2897 + U16_NEXT(patternString, patternStringIndex, stringLen, patternChar);
1.2898 + if (patternChar != inputChar) {
1.2899 + success = FALSE;
1.2900 + break;
1.2901 + }
1.2902 + }
1.2903 +
1.2904 + if (success) {
1.2905 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.2906 + } else {
1.2907 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.2908 + }
1.2909 + }
1.2910 + break;
1.2911 +
1.2912 +
1.2913 + case URX_STATE_SAVE:
1.2914 + fp = StateSave(fp, opValue, status);
1.2915 + break;
1.2916 +
1.2917 +
1.2918 + case URX_END:
1.2919 + // The match loop will exit via this path on a successful match,
1.2920 + // when we reach the end of the pattern.
1.2921 + if (toEnd && fp->fInputIdx != fActiveLimit) {
1.2922 + // The pattern matched, but not to the end of input. Try some more.
1.2923 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.2924 + break;
1.2925 + }
1.2926 + isMatch = TRUE;
1.2927 + goto breakFromLoop;
1.2928 +
1.2929 + // Start and End Capture stack frame variables are laid out out like this:
1.2930 + // fp->fExtra[opValue] - The start of a completed capture group
1.2931 + // opValue+1 - The end of a completed capture group
1.2932 + // opValue+2 - the start of a capture group whose end
1.2933 + // has not yet been reached (and might not ever be).
1.2934 + case URX_START_CAPTURE:
1.2935 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-3);
1.2936 + fp->fExtra[opValue+2] = fp->fInputIdx;
1.2937 + break;
1.2938 +
1.2939 +
1.2940 + case URX_END_CAPTURE:
1.2941 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-3);
1.2942 + U_ASSERT(fp->fExtra[opValue+2] >= 0); // Start pos for this group must be set.
1.2943 + fp->fExtra[opValue] = fp->fExtra[opValue+2]; // Tentative start becomes real.
1.2944 + fp->fExtra[opValue+1] = fp->fInputIdx; // End position
1.2945 + U_ASSERT(fp->fExtra[opValue] <= fp->fExtra[opValue+1]);
1.2946 + break;
1.2947 +
1.2948 +
1.2949 + case URX_DOLLAR: // $, test for End of line
1.2950 + // or for position before new line at end of input
1.2951 + {
1.2952 + if (fp->fInputIdx >= fAnchorLimit) {
1.2953 + // We really are at the end of input. Success.
1.2954 + fHitEnd = TRUE;
1.2955 + fRequireEnd = TRUE;
1.2956 + break;
1.2957 + }
1.2958 +
1.2959 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.2960 +
1.2961 + // If we are positioned just before a new-line that is located at the
1.2962 + // end of input, succeed.
1.2963 + UChar32 c = UTEXT_NEXT32(fInputText);
1.2964 + if (UTEXT_GETNATIVEINDEX(fInputText) >= fAnchorLimit) {
1.2965 + if ((c>=0x0a && c<=0x0d) || c==0x85 || c==0x2028 || c==0x2029) {
1.2966 + // If not in the middle of a CR/LF sequence
1.2967 + if ( !(c==0x0a && fp->fInputIdx>fAnchorStart && ((void)UTEXT_PREVIOUS32(fInputText), UTEXT_PREVIOUS32(fInputText))==0x0d)) {
1.2968 + // At new-line at end of input. Success
1.2969 + fHitEnd = TRUE;
1.2970 + fRequireEnd = TRUE;
1.2971 +
1.2972 + break;
1.2973 + }
1.2974 + }
1.2975 + } else {
1.2976 + UChar32 nextC = UTEXT_NEXT32(fInputText);
1.2977 + if (c == 0x0d && nextC == 0x0a && UTEXT_GETNATIVEINDEX(fInputText) >= fAnchorLimit) {
1.2978 + fHitEnd = TRUE;
1.2979 + fRequireEnd = TRUE;
1.2980 + break; // At CR/LF at end of input. Success
1.2981 + }
1.2982 + }
1.2983 +
1.2984 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.2985 + }
1.2986 + break;
1.2987 +
1.2988 +
1.2989 + case URX_DOLLAR_D: // $, test for End of Line, in UNIX_LINES mode.
1.2990 + if (fp->fInputIdx >= fAnchorLimit) {
1.2991 + // Off the end of input. Success.
1.2992 + fHitEnd = TRUE;
1.2993 + fRequireEnd = TRUE;
1.2994 + break;
1.2995 + } else {
1.2996 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.2997 + UChar32 c = UTEXT_NEXT32(fInputText);
1.2998 + // Either at the last character of input, or off the end.
1.2999 + if (c == 0x0a && UTEXT_GETNATIVEINDEX(fInputText) == fAnchorLimit) {
1.3000 + fHitEnd = TRUE;
1.3001 + fRequireEnd = TRUE;
1.3002 + break;
1.3003 + }
1.3004 + }
1.3005 +
1.3006 + // Not at end of input. Back-track out.
1.3007 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3008 + break;
1.3009 +
1.3010 +
1.3011 + case URX_DOLLAR_M: // $, test for End of line in multi-line mode
1.3012 + {
1.3013 + if (fp->fInputIdx >= fAnchorLimit) {
1.3014 + // We really are at the end of input. Success.
1.3015 + fHitEnd = TRUE;
1.3016 + fRequireEnd = TRUE;
1.3017 + break;
1.3018 + }
1.3019 + // If we are positioned just before a new-line, succeed.
1.3020 + // It makes no difference where the new-line is within the input.
1.3021 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3022 + UChar32 c = UTEXT_CURRENT32(fInputText);
1.3023 + if ((c>=0x0a && c<=0x0d) || c==0x85 ||c==0x2028 || c==0x2029) {
1.3024 + // At a line end, except for the odd chance of being in the middle of a CR/LF sequence
1.3025 + // In multi-line mode, hitting a new-line just before the end of input does not
1.3026 + // set the hitEnd or requireEnd flags
1.3027 + if ( !(c==0x0a && fp->fInputIdx>fAnchorStart && UTEXT_PREVIOUS32(fInputText)==0x0d)) {
1.3028 + break;
1.3029 + }
1.3030 + }
1.3031 + // not at a new line. Fail.
1.3032 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3033 + }
1.3034 + break;
1.3035 +
1.3036 +
1.3037 + case URX_DOLLAR_MD: // $, test for End of line in multi-line and UNIX_LINES mode
1.3038 + {
1.3039 + if (fp->fInputIdx >= fAnchorLimit) {
1.3040 + // We really are at the end of input. Success.
1.3041 + fHitEnd = TRUE;
1.3042 + fRequireEnd = TRUE; // Java set requireEnd in this case, even though
1.3043 + break; // adding a new-line would not lose the match.
1.3044 + }
1.3045 + // If we are not positioned just before a new-line, the test fails; backtrack out.
1.3046 + // It makes no difference where the new-line is within the input.
1.3047 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3048 + if (UTEXT_CURRENT32(fInputText) != 0x0a) {
1.3049 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3050 + }
1.3051 + }
1.3052 + break;
1.3053 +
1.3054 +
1.3055 + case URX_CARET: // ^, test for start of line
1.3056 + if (fp->fInputIdx != fAnchorStart) {
1.3057 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3058 + }
1.3059 + break;
1.3060 +
1.3061 +
1.3062 + case URX_CARET_M: // ^, test for start of line in mulit-line mode
1.3063 + {
1.3064 + if (fp->fInputIdx == fAnchorStart) {
1.3065 + // We are at the start input. Success.
1.3066 + break;
1.3067 + }
1.3068 + // Check whether character just before the current pos is a new-line
1.3069 + // unless we are at the end of input
1.3070 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3071 + UChar32 c = UTEXT_PREVIOUS32(fInputText);
1.3072 + if ((fp->fInputIdx < fAnchorLimit) &&
1.3073 + ((c<=0x0d && c>=0x0a) || c==0x85 ||c==0x2028 || c==0x2029)) {
1.3074 + // It's a new-line. ^ is true. Success.
1.3075 + // TODO: what should be done with positions between a CR and LF?
1.3076 + break;
1.3077 + }
1.3078 + // Not at the start of a line. Fail.
1.3079 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3080 + }
1.3081 + break;
1.3082 +
1.3083 +
1.3084 + case URX_CARET_M_UNIX: // ^, test for start of line in mulit-line + Unix-line mode
1.3085 + {
1.3086 + U_ASSERT(fp->fInputIdx >= fAnchorStart);
1.3087 + if (fp->fInputIdx <= fAnchorStart) {
1.3088 + // We are at the start input. Success.
1.3089 + break;
1.3090 + }
1.3091 + // Check whether character just before the current pos is a new-line
1.3092 + U_ASSERT(fp->fInputIdx <= fAnchorLimit);
1.3093 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3094 + UChar32 c = UTEXT_PREVIOUS32(fInputText);
1.3095 + if (c != 0x0a) {
1.3096 + // Not at the start of a line. Back-track out.
1.3097 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3098 + }
1.3099 + }
1.3100 + break;
1.3101 +
1.3102 + case URX_BACKSLASH_B: // Test for word boundaries
1.3103 + {
1.3104 + UBool success = isWordBoundary(fp->fInputIdx);
1.3105 + success ^= (UBool)(opValue != 0); // flip sense for \B
1.3106 + if (!success) {
1.3107 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3108 + }
1.3109 + }
1.3110 + break;
1.3111 +
1.3112 +
1.3113 + case URX_BACKSLASH_BU: // Test for word boundaries, Unicode-style
1.3114 + {
1.3115 + UBool success = isUWordBoundary(fp->fInputIdx);
1.3116 + success ^= (UBool)(opValue != 0); // flip sense for \B
1.3117 + if (!success) {
1.3118 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3119 + }
1.3120 + }
1.3121 + break;
1.3122 +
1.3123 +
1.3124 + case URX_BACKSLASH_D: // Test for decimal digit
1.3125 + {
1.3126 + if (fp->fInputIdx >= fActiveLimit) {
1.3127 + fHitEnd = TRUE;
1.3128 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3129 + break;
1.3130 + }
1.3131 +
1.3132 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3133 +
1.3134 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3135 + int8_t ctype = u_charType(c); // TODO: make a unicode set for this. Will be faster.
1.3136 + UBool success = (ctype == U_DECIMAL_DIGIT_NUMBER);
1.3137 + success ^= (UBool)(opValue != 0); // flip sense for \D
1.3138 + if (success) {
1.3139 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3140 + } else {
1.3141 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3142 + }
1.3143 + }
1.3144 + break;
1.3145 +
1.3146 +
1.3147 + case URX_BACKSLASH_G: // Test for position at end of previous match
1.3148 + if (!((fMatch && fp->fInputIdx==fMatchEnd) || (fMatch==FALSE && fp->fInputIdx==fActiveStart))) {
1.3149 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3150 + }
1.3151 + break;
1.3152 +
1.3153 +
1.3154 + case URX_BACKSLASH_X:
1.3155 + // Match a Grapheme, as defined by Unicode TR 29.
1.3156 + // Differs slightly from Perl, which consumes combining marks independently
1.3157 + // of context.
1.3158 + {
1.3159 +
1.3160 + // Fail if at end of input
1.3161 + if (fp->fInputIdx >= fActiveLimit) {
1.3162 + fHitEnd = TRUE;
1.3163 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3164 + break;
1.3165 + }
1.3166 +
1.3167 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3168 +
1.3169 + // Examine (and consume) the current char.
1.3170 + // Dispatch into a little state machine, based on the char.
1.3171 + UChar32 c;
1.3172 + c = UTEXT_NEXT32(fInputText);
1.3173 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3174 + UnicodeSet **sets = fPattern->fStaticSets;
1.3175 + if (sets[URX_GC_NORMAL]->contains(c)) goto GC_Extend;
1.3176 + if (sets[URX_GC_CONTROL]->contains(c)) goto GC_Control;
1.3177 + if (sets[URX_GC_L]->contains(c)) goto GC_L;
1.3178 + if (sets[URX_GC_LV]->contains(c)) goto GC_V;
1.3179 + if (sets[URX_GC_LVT]->contains(c)) goto GC_T;
1.3180 + if (sets[URX_GC_V]->contains(c)) goto GC_V;
1.3181 + if (sets[URX_GC_T]->contains(c)) goto GC_T;
1.3182 + goto GC_Extend;
1.3183 +
1.3184 +
1.3185 +
1.3186 +GC_L:
1.3187 + if (fp->fInputIdx >= fActiveLimit) goto GC_Done;
1.3188 + c = UTEXT_NEXT32(fInputText);
1.3189 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3190 + if (sets[URX_GC_L]->contains(c)) goto GC_L;
1.3191 + if (sets[URX_GC_LV]->contains(c)) goto GC_V;
1.3192 + if (sets[URX_GC_LVT]->contains(c)) goto GC_T;
1.3193 + if (sets[URX_GC_V]->contains(c)) goto GC_V;
1.3194 + (void)UTEXT_PREVIOUS32(fInputText);
1.3195 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3196 + goto GC_Extend;
1.3197 +
1.3198 +GC_V:
1.3199 + if (fp->fInputIdx >= fActiveLimit) goto GC_Done;
1.3200 + c = UTEXT_NEXT32(fInputText);
1.3201 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3202 + if (sets[URX_GC_V]->contains(c)) goto GC_V;
1.3203 + if (sets[URX_GC_T]->contains(c)) goto GC_T;
1.3204 + (void)UTEXT_PREVIOUS32(fInputText);
1.3205 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3206 + goto GC_Extend;
1.3207 +
1.3208 +GC_T:
1.3209 + if (fp->fInputIdx >= fActiveLimit) goto GC_Done;
1.3210 + c = UTEXT_NEXT32(fInputText);
1.3211 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3212 + if (sets[URX_GC_T]->contains(c)) goto GC_T;
1.3213 + (void)UTEXT_PREVIOUS32(fInputText);
1.3214 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3215 + goto GC_Extend;
1.3216 +
1.3217 +GC_Extend:
1.3218 + // Combining characters are consumed here
1.3219 + for (;;) {
1.3220 + if (fp->fInputIdx >= fActiveLimit) {
1.3221 + break;
1.3222 + }
1.3223 + c = UTEXT_CURRENT32(fInputText);
1.3224 + if (sets[URX_GC_EXTEND]->contains(c) == FALSE) {
1.3225 + break;
1.3226 + }
1.3227 + (void)UTEXT_NEXT32(fInputText);
1.3228 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3229 + }
1.3230 + goto GC_Done;
1.3231 +
1.3232 +GC_Control:
1.3233 + // Most control chars stand alone (don't combine with combining chars),
1.3234 + // except for that CR/LF sequence is a single grapheme cluster.
1.3235 + if (c == 0x0d && fp->fInputIdx < fActiveLimit && UTEXT_CURRENT32(fInputText) == 0x0a) {
1.3236 + c = UTEXT_NEXT32(fInputText);
1.3237 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3238 + }
1.3239 +
1.3240 +GC_Done:
1.3241 + if (fp->fInputIdx >= fActiveLimit) {
1.3242 + fHitEnd = TRUE;
1.3243 + }
1.3244 + break;
1.3245 + }
1.3246 +
1.3247 +
1.3248 +
1.3249 +
1.3250 + case URX_BACKSLASH_Z: // Test for end of Input
1.3251 + if (fp->fInputIdx < fAnchorLimit) {
1.3252 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3253 + } else {
1.3254 + fHitEnd = TRUE;
1.3255 + fRequireEnd = TRUE;
1.3256 + }
1.3257 + break;
1.3258 +
1.3259 +
1.3260 +
1.3261 + case URX_STATIC_SETREF:
1.3262 + {
1.3263 + // Test input character against one of the predefined sets
1.3264 + // (Word Characters, for example)
1.3265 + // The high bit of the op value is a flag for the match polarity.
1.3266 + // 0: success if input char is in set.
1.3267 + // 1: success if input char is not in set.
1.3268 + if (fp->fInputIdx >= fActiveLimit) {
1.3269 + fHitEnd = TRUE;
1.3270 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3271 + break;
1.3272 + }
1.3273 +
1.3274 + UBool success = ((opValue & URX_NEG_SET) == URX_NEG_SET);
1.3275 + opValue &= ~URX_NEG_SET;
1.3276 + U_ASSERT(opValue > 0 && opValue < URX_LAST_SET);
1.3277 +
1.3278 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3279 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3280 + if (c < 256) {
1.3281 + Regex8BitSet *s8 = &fPattern->fStaticSets8[opValue];
1.3282 + if (s8->contains(c)) {
1.3283 + success = !success;
1.3284 + }
1.3285 + } else {
1.3286 + const UnicodeSet *s = fPattern->fStaticSets[opValue];
1.3287 + if (s->contains(c)) {
1.3288 + success = !success;
1.3289 + }
1.3290 + }
1.3291 + if (success) {
1.3292 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3293 + } else {
1.3294 + // the character wasn't in the set.
1.3295 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3296 + }
1.3297 + }
1.3298 + break;
1.3299 +
1.3300 +
1.3301 + case URX_STAT_SETREF_N:
1.3302 + {
1.3303 + // Test input character for NOT being a member of one of
1.3304 + // the predefined sets (Word Characters, for example)
1.3305 + if (fp->fInputIdx >= fActiveLimit) {
1.3306 + fHitEnd = TRUE;
1.3307 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3308 + break;
1.3309 + }
1.3310 +
1.3311 + U_ASSERT(opValue > 0 && opValue < URX_LAST_SET);
1.3312 +
1.3313 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3314 +
1.3315 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3316 + if (c < 256) {
1.3317 + Regex8BitSet *s8 = &fPattern->fStaticSets8[opValue];
1.3318 + if (s8->contains(c) == FALSE) {
1.3319 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3320 + break;
1.3321 + }
1.3322 + } else {
1.3323 + const UnicodeSet *s = fPattern->fStaticSets[opValue];
1.3324 + if (s->contains(c) == FALSE) {
1.3325 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3326 + break;
1.3327 + }
1.3328 + }
1.3329 + // the character wasn't in the set.
1.3330 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3331 + }
1.3332 + break;
1.3333 +
1.3334 +
1.3335 + case URX_SETREF:
1.3336 + if (fp->fInputIdx >= fActiveLimit) {
1.3337 + fHitEnd = TRUE;
1.3338 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3339 + break;
1.3340 + } else {
1.3341 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3342 +
1.3343 + // There is input left. Pick up one char and test it for set membership.
1.3344 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3345 + U_ASSERT(opValue > 0 && opValue < sets->size());
1.3346 + if (c<256) {
1.3347 + Regex8BitSet *s8 = &fPattern->fSets8[opValue];
1.3348 + if (s8->contains(c)) {
1.3349 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3350 + break;
1.3351 + }
1.3352 + } else {
1.3353 + UnicodeSet *s = (UnicodeSet *)sets->elementAt(opValue);
1.3354 + if (s->contains(c)) {
1.3355 + // The character is in the set. A Match.
1.3356 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3357 + break;
1.3358 + }
1.3359 + }
1.3360 +
1.3361 + // the character wasn't in the set.
1.3362 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3363 + }
1.3364 + break;
1.3365 +
1.3366 +
1.3367 + case URX_DOTANY:
1.3368 + {
1.3369 + // . matches anything, but stops at end-of-line.
1.3370 + if (fp->fInputIdx >= fActiveLimit) {
1.3371 + // At end of input. Match failed. Backtrack out.
1.3372 + fHitEnd = TRUE;
1.3373 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3374 + break;
1.3375 + }
1.3376 +
1.3377 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3378 +
1.3379 + // There is input left. Advance over one char, unless we've hit end-of-line
1.3380 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3381 + if (((c & 0x7f) <= 0x29) && // First quickly bypass as many chars as possible
1.3382 + ((c<=0x0d && c>=0x0a) || c==0x85 ||c==0x2028 || c==0x2029)) {
1.3383 + // End of line in normal mode. . does not match.
1.3384 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3385 + break;
1.3386 + }
1.3387 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3388 + }
1.3389 + break;
1.3390 +
1.3391 +
1.3392 + case URX_DOTANY_ALL:
1.3393 + {
1.3394 + // ., in dot-matches-all (including new lines) mode
1.3395 + if (fp->fInputIdx >= fActiveLimit) {
1.3396 + // At end of input. Match failed. Backtrack out.
1.3397 + fHitEnd = TRUE;
1.3398 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3399 + break;
1.3400 + }
1.3401 +
1.3402 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3403 +
1.3404 + // There is input left. Advance over one char, except if we are
1.3405 + // at a cr/lf, advance over both of them.
1.3406 + UChar32 c;
1.3407 + c = UTEXT_NEXT32(fInputText);
1.3408 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3409 + if (c==0x0d && fp->fInputIdx < fActiveLimit) {
1.3410 + // In the case of a CR/LF, we need to advance over both.
1.3411 + UChar32 nextc = UTEXT_CURRENT32(fInputText);
1.3412 + if (nextc == 0x0a) {
1.3413 + (void)UTEXT_NEXT32(fInputText);
1.3414 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3415 + }
1.3416 + }
1.3417 + }
1.3418 + break;
1.3419 +
1.3420 +
1.3421 + case URX_DOTANY_UNIX:
1.3422 + {
1.3423 + // '.' operator, matches all, but stops at end-of-line.
1.3424 + // UNIX_LINES mode, so 0x0a is the only recognized line ending.
1.3425 + if (fp->fInputIdx >= fActiveLimit) {
1.3426 + // At end of input. Match failed. Backtrack out.
1.3427 + fHitEnd = TRUE;
1.3428 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3429 + break;
1.3430 + }
1.3431 +
1.3432 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3433 +
1.3434 + // There is input left. Advance over one char, unless we've hit end-of-line
1.3435 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3436 + if (c == 0x0a) {
1.3437 + // End of line in normal mode. '.' does not match the \n
1.3438 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3439 + } else {
1.3440 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3441 + }
1.3442 + }
1.3443 + break;
1.3444 +
1.3445 +
1.3446 + case URX_JMP:
1.3447 + fp->fPatIdx = opValue;
1.3448 + break;
1.3449 +
1.3450 + case URX_FAIL:
1.3451 + isMatch = FALSE;
1.3452 + goto breakFromLoop;
1.3453 +
1.3454 + case URX_JMP_SAV:
1.3455 + U_ASSERT(opValue < fPattern->fCompiledPat->size());
1.3456 + fp = StateSave(fp, fp->fPatIdx, status); // State save to loc following current
1.3457 + fp->fPatIdx = opValue; // Then JMP.
1.3458 + break;
1.3459 +
1.3460 + case URX_JMP_SAV_X:
1.3461 + // This opcode is used with (x)+, when x can match a zero length string.
1.3462 + // Same as JMP_SAV, except conditional on the match having made forward progress.
1.3463 + // Destination of the JMP must be a URX_STO_INP_LOC, from which we get the
1.3464 + // data address of the input position at the start of the loop.
1.3465 + {
1.3466 + U_ASSERT(opValue > 0 && opValue < fPattern->fCompiledPat->size());
1.3467 + int32_t stoOp = (int32_t)pat[opValue-1];
1.3468 + U_ASSERT(URX_TYPE(stoOp) == URX_STO_INP_LOC);
1.3469 + int32_t frameLoc = URX_VAL(stoOp);
1.3470 + U_ASSERT(frameLoc >= 0 && frameLoc < fFrameSize);
1.3471 + int64_t prevInputIdx = fp->fExtra[frameLoc];
1.3472 + U_ASSERT(prevInputIdx <= fp->fInputIdx);
1.3473 + if (prevInputIdx < fp->fInputIdx) {
1.3474 + // The match did make progress. Repeat the loop.
1.3475 + fp = StateSave(fp, fp->fPatIdx, status); // State save to loc following current
1.3476 + fp->fPatIdx = opValue;
1.3477 + fp->fExtra[frameLoc] = fp->fInputIdx;
1.3478 + }
1.3479 + // If the input position did not advance, we do nothing here,
1.3480 + // execution will fall out of the loop.
1.3481 + }
1.3482 + break;
1.3483 +
1.3484 + case URX_CTR_INIT:
1.3485 + {
1.3486 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-2);
1.3487 + fp->fExtra[opValue] = 0; // Set the loop counter variable to zero
1.3488 +
1.3489 + // Pick up the three extra operands that CTR_INIT has, and
1.3490 + // skip the pattern location counter past
1.3491 + int32_t instrOperandLoc = (int32_t)fp->fPatIdx;
1.3492 + fp->fPatIdx += 3;
1.3493 + int32_t loopLoc = URX_VAL(pat[instrOperandLoc]);
1.3494 + int32_t minCount = (int32_t)pat[instrOperandLoc+1];
1.3495 + int32_t maxCount = (int32_t)pat[instrOperandLoc+2];
1.3496 + U_ASSERT(minCount>=0);
1.3497 + U_ASSERT(maxCount>=minCount || maxCount==-1);
1.3498 + U_ASSERT(loopLoc>=fp->fPatIdx);
1.3499 +
1.3500 + if (minCount == 0) {
1.3501 + fp = StateSave(fp, loopLoc+1, status);
1.3502 + }
1.3503 + if (maxCount == -1) {
1.3504 + fp->fExtra[opValue+1] = fp->fInputIdx; // For loop breaking.
1.3505 + } else if (maxCount == 0) {
1.3506 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3507 + }
1.3508 + }
1.3509 + break;
1.3510 +
1.3511 + case URX_CTR_LOOP:
1.3512 + {
1.3513 + U_ASSERT(opValue>0 && opValue < fp->fPatIdx-2);
1.3514 + int32_t initOp = (int32_t)pat[opValue];
1.3515 + U_ASSERT(URX_TYPE(initOp) == URX_CTR_INIT);
1.3516 + int64_t *pCounter = &fp->fExtra[URX_VAL(initOp)];
1.3517 + int32_t minCount = (int32_t)pat[opValue+2];
1.3518 + int32_t maxCount = (int32_t)pat[opValue+3];
1.3519 + (*pCounter)++;
1.3520 + if ((uint64_t)*pCounter >= (uint32_t)maxCount && maxCount != -1) {
1.3521 + U_ASSERT(*pCounter == maxCount);
1.3522 + break;
1.3523 + }
1.3524 + if (*pCounter >= minCount) {
1.3525 + if (maxCount == -1) {
1.3526 + // Loop has no hard upper bound.
1.3527 + // Check that it is progressing through the input, break if it is not.
1.3528 + int64_t *pLastInputIdx = &fp->fExtra[URX_VAL(initOp) + 1];
1.3529 + if (fp->fInputIdx == *pLastInputIdx) {
1.3530 + break;
1.3531 + } else {
1.3532 + *pLastInputIdx = fp->fInputIdx;
1.3533 + }
1.3534 + }
1.3535 + fp = StateSave(fp, fp->fPatIdx, status);
1.3536 + }
1.3537 + fp->fPatIdx = opValue + 4; // Loop back.
1.3538 + }
1.3539 + break;
1.3540 +
1.3541 + case URX_CTR_INIT_NG:
1.3542 + {
1.3543 + // Initialize a non-greedy loop
1.3544 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-2);
1.3545 + fp->fExtra[opValue] = 0; // Set the loop counter variable to zero
1.3546 +
1.3547 + // Pick up the three extra operands that CTR_INIT_NG has, and
1.3548 + // skip the pattern location counter past
1.3549 + int32_t instrOperandLoc = (int32_t)fp->fPatIdx;
1.3550 + fp->fPatIdx += 3;
1.3551 + int32_t loopLoc = URX_VAL(pat[instrOperandLoc]);
1.3552 + int32_t minCount = (int32_t)pat[instrOperandLoc+1];
1.3553 + int32_t maxCount = (int32_t)pat[instrOperandLoc+2];
1.3554 + U_ASSERT(minCount>=0);
1.3555 + U_ASSERT(maxCount>=minCount || maxCount==-1);
1.3556 + U_ASSERT(loopLoc>fp->fPatIdx);
1.3557 + if (maxCount == -1) {
1.3558 + fp->fExtra[opValue+1] = fp->fInputIdx; // Save initial input index for loop breaking.
1.3559 + }
1.3560 +
1.3561 + if (minCount == 0) {
1.3562 + if (maxCount != 0) {
1.3563 + fp = StateSave(fp, fp->fPatIdx, status);
1.3564 + }
1.3565 + fp->fPatIdx = loopLoc+1; // Continue with stuff after repeated block
1.3566 + }
1.3567 + }
1.3568 + break;
1.3569 +
1.3570 + case URX_CTR_LOOP_NG:
1.3571 + {
1.3572 + // Non-greedy {min, max} loops
1.3573 + U_ASSERT(opValue>0 && opValue < fp->fPatIdx-2);
1.3574 + int32_t initOp = (int32_t)pat[opValue];
1.3575 + U_ASSERT(URX_TYPE(initOp) == URX_CTR_INIT_NG);
1.3576 + int64_t *pCounter = &fp->fExtra[URX_VAL(initOp)];
1.3577 + int32_t minCount = (int32_t)pat[opValue+2];
1.3578 + int32_t maxCount = (int32_t)pat[opValue+3];
1.3579 +
1.3580 + (*pCounter)++;
1.3581 + if ((uint64_t)*pCounter >= (uint32_t)maxCount && maxCount != -1) {
1.3582 + // The loop has matched the maximum permitted number of times.
1.3583 + // Break out of here with no action. Matching will
1.3584 + // continue with the following pattern.
1.3585 + U_ASSERT(*pCounter == maxCount);
1.3586 + break;
1.3587 + }
1.3588 +
1.3589 + if (*pCounter < minCount) {
1.3590 + // We haven't met the minimum number of matches yet.
1.3591 + // Loop back for another one.
1.3592 + fp->fPatIdx = opValue + 4; // Loop back.
1.3593 + } else {
1.3594 + // We do have the minimum number of matches.
1.3595 +
1.3596 + // If there is no upper bound on the loop iterations, check that the input index
1.3597 + // is progressing, and stop the loop if it is not.
1.3598 + if (maxCount == -1) {
1.3599 + int64_t *pLastInputIdx = &fp->fExtra[URX_VAL(initOp) + 1];
1.3600 + if (fp->fInputIdx == *pLastInputIdx) {
1.3601 + break;
1.3602 + }
1.3603 + *pLastInputIdx = fp->fInputIdx;
1.3604 + }
1.3605 +
1.3606 + // Loop Continuation: we will fall into the pattern following the loop
1.3607 + // (non-greedy, don't execute loop body first), but first do
1.3608 + // a state save to the top of the loop, so that a match failure
1.3609 + // in the following pattern will try another iteration of the loop.
1.3610 + fp = StateSave(fp, opValue + 4, status);
1.3611 + }
1.3612 + }
1.3613 + break;
1.3614 +
1.3615 + case URX_STO_SP:
1.3616 + U_ASSERT(opValue >= 0 && opValue < fPattern->fDataSize);
1.3617 + fData[opValue] = fStack->size();
1.3618 + break;
1.3619 +
1.3620 + case URX_LD_SP:
1.3621 + {
1.3622 + U_ASSERT(opValue >= 0 && opValue < fPattern->fDataSize);
1.3623 + int32_t newStackSize = (int32_t)fData[opValue];
1.3624 + U_ASSERT(newStackSize <= fStack->size());
1.3625 + int64_t *newFP = fStack->getBuffer() + newStackSize - fFrameSize;
1.3626 + if (newFP == (int64_t *)fp) {
1.3627 + break;
1.3628 + }
1.3629 + int32_t i;
1.3630 + for (i=0; i<fFrameSize; i++) {
1.3631 + newFP[i] = ((int64_t *)fp)[i];
1.3632 + }
1.3633 + fp = (REStackFrame *)newFP;
1.3634 + fStack->setSize(newStackSize);
1.3635 + }
1.3636 + break;
1.3637 +
1.3638 + case URX_BACKREF:
1.3639 + {
1.3640 + U_ASSERT(opValue < fFrameSize);
1.3641 + int64_t groupStartIdx = fp->fExtra[opValue];
1.3642 + int64_t groupEndIdx = fp->fExtra[opValue+1];
1.3643 + U_ASSERT(groupStartIdx <= groupEndIdx);
1.3644 + if (groupStartIdx < 0) {
1.3645 + // This capture group has not participated in the match thus far,
1.3646 + fp = (REStackFrame *)fStack->popFrame(fFrameSize); // FAIL, no match.
1.3647 + break;
1.3648 + }
1.3649 + UTEXT_SETNATIVEINDEX(fAltInputText, groupStartIdx);
1.3650 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3651 +
1.3652 + // Note: if the capture group match was of an empty string the backref
1.3653 + // match succeeds. Verified by testing: Perl matches succeed
1.3654 + // in this case, so we do too.
1.3655 +
1.3656 + UBool success = TRUE;
1.3657 + for (;;) {
1.3658 + if (utext_getNativeIndex(fAltInputText) >= groupEndIdx) {
1.3659 + success = TRUE;
1.3660 + break;
1.3661 + }
1.3662 + if (utext_getNativeIndex(fInputText) >= fActiveLimit) {
1.3663 + success = FALSE;
1.3664 + fHitEnd = TRUE;
1.3665 + break;
1.3666 + }
1.3667 + UChar32 captureGroupChar = utext_next32(fAltInputText);
1.3668 + UChar32 inputChar = utext_next32(fInputText);
1.3669 + if (inputChar != captureGroupChar) {
1.3670 + success = FALSE;
1.3671 + break;
1.3672 + }
1.3673 + }
1.3674 +
1.3675 + if (success) {
1.3676 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3677 + } else {
1.3678 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3679 + }
1.3680 + }
1.3681 + break;
1.3682 +
1.3683 +
1.3684 +
1.3685 + case URX_BACKREF_I:
1.3686 + {
1.3687 + U_ASSERT(opValue < fFrameSize);
1.3688 + int64_t groupStartIdx = fp->fExtra[opValue];
1.3689 + int64_t groupEndIdx = fp->fExtra[opValue+1];
1.3690 + U_ASSERT(groupStartIdx <= groupEndIdx);
1.3691 + if (groupStartIdx < 0) {
1.3692 + // This capture group has not participated in the match thus far,
1.3693 + fp = (REStackFrame *)fStack->popFrame(fFrameSize); // FAIL, no match.
1.3694 + break;
1.3695 + }
1.3696 + utext_setNativeIndex(fAltInputText, groupStartIdx);
1.3697 + utext_setNativeIndex(fInputText, fp->fInputIdx);
1.3698 + CaseFoldingUTextIterator captureGroupItr(*fAltInputText);
1.3699 + CaseFoldingUTextIterator inputItr(*fInputText);
1.3700 +
1.3701 + // Note: if the capture group match was of an empty string the backref
1.3702 + // match succeeds. Verified by testing: Perl matches succeed
1.3703 + // in this case, so we do too.
1.3704 +
1.3705 + UBool success = TRUE;
1.3706 + for (;;) {
1.3707 + if (!captureGroupItr.inExpansion() && utext_getNativeIndex(fAltInputText) >= groupEndIdx) {
1.3708 + success = TRUE;
1.3709 + break;
1.3710 + }
1.3711 + if (!inputItr.inExpansion() && utext_getNativeIndex(fInputText) >= fActiveLimit) {
1.3712 + success = FALSE;
1.3713 + fHitEnd = TRUE;
1.3714 + break;
1.3715 + }
1.3716 + UChar32 captureGroupChar = captureGroupItr.next();
1.3717 + UChar32 inputChar = inputItr.next();
1.3718 + if (inputChar != captureGroupChar) {
1.3719 + success = FALSE;
1.3720 + break;
1.3721 + }
1.3722 + }
1.3723 +
1.3724 + if (success && inputItr.inExpansion()) {
1.3725 + // We otained a match by consuming part of a string obtained from
1.3726 + // case-folding a single code point of the input text.
1.3727 + // This does not count as an overall match.
1.3728 + success = FALSE;
1.3729 + }
1.3730 +
1.3731 + if (success) {
1.3732 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3733 + } else {
1.3734 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3735 + }
1.3736 +
1.3737 + }
1.3738 + break;
1.3739 +
1.3740 + case URX_STO_INP_LOC:
1.3741 + {
1.3742 + U_ASSERT(opValue >= 0 && opValue < fFrameSize);
1.3743 + fp->fExtra[opValue] = fp->fInputIdx;
1.3744 + }
1.3745 + break;
1.3746 +
1.3747 + case URX_JMPX:
1.3748 + {
1.3749 + int32_t instrOperandLoc = (int32_t)fp->fPatIdx;
1.3750 + fp->fPatIdx += 1;
1.3751 + int32_t dataLoc = URX_VAL(pat[instrOperandLoc]);
1.3752 + U_ASSERT(dataLoc >= 0 && dataLoc < fFrameSize);
1.3753 + int64_t savedInputIdx = fp->fExtra[dataLoc];
1.3754 + U_ASSERT(savedInputIdx <= fp->fInputIdx);
1.3755 + if (savedInputIdx < fp->fInputIdx) {
1.3756 + fp->fPatIdx = opValue; // JMP
1.3757 + } else {
1.3758 + fp = (REStackFrame *)fStack->popFrame(fFrameSize); // FAIL, no progress in loop.
1.3759 + }
1.3760 + }
1.3761 + break;
1.3762 +
1.3763 + case URX_LA_START:
1.3764 + {
1.3765 + // Entering a lookahead block.
1.3766 + // Save Stack Ptr, Input Pos.
1.3767 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.3768 + fData[opValue] = fStack->size();
1.3769 + fData[opValue+1] = fp->fInputIdx;
1.3770 + fActiveStart = fLookStart; // Set the match region change for
1.3771 + fActiveLimit = fLookLimit; // transparent bounds.
1.3772 + }
1.3773 + break;
1.3774 +
1.3775 + case URX_LA_END:
1.3776 + {
1.3777 + // Leaving a look-ahead block.
1.3778 + // restore Stack Ptr, Input Pos to positions they had on entry to block.
1.3779 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.3780 + int32_t stackSize = fStack->size();
1.3781 + int32_t newStackSize =(int32_t)fData[opValue];
1.3782 + U_ASSERT(stackSize >= newStackSize);
1.3783 + if (stackSize > newStackSize) {
1.3784 + // Copy the current top frame back to the new (cut back) top frame.
1.3785 + // This makes the capture groups from within the look-ahead
1.3786 + // expression available.
1.3787 + int64_t *newFP = fStack->getBuffer() + newStackSize - fFrameSize;
1.3788 + int32_t i;
1.3789 + for (i=0; i<fFrameSize; i++) {
1.3790 + newFP[i] = ((int64_t *)fp)[i];
1.3791 + }
1.3792 + fp = (REStackFrame *)newFP;
1.3793 + fStack->setSize(newStackSize);
1.3794 + }
1.3795 + fp->fInputIdx = fData[opValue+1];
1.3796 +
1.3797 + // Restore the active region bounds in the input string; they may have
1.3798 + // been changed because of transparent bounds on a Region.
1.3799 + fActiveStart = fRegionStart;
1.3800 + fActiveLimit = fRegionLimit;
1.3801 + }
1.3802 + break;
1.3803 +
1.3804 + case URX_ONECHAR_I:
1.3805 + // Case insensitive one char. The char from the pattern is already case folded.
1.3806 + // Input text is not, but case folding the input can not reduce two or more code
1.3807 + // points to one.
1.3808 + if (fp->fInputIdx < fActiveLimit) {
1.3809 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3810 +
1.3811 + UChar32 c = UTEXT_NEXT32(fInputText);
1.3812 + if (u_foldCase(c, U_FOLD_CASE_DEFAULT) == opValue) {
1.3813 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3814 + break;
1.3815 + }
1.3816 + } else {
1.3817 + fHitEnd = TRUE;
1.3818 + }
1.3819 +
1.3820 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3821 + break;
1.3822 +
1.3823 + case URX_STRING_I:
1.3824 + {
1.3825 + // Case-insensitive test input against a literal string.
1.3826 + // Strings require two slots in the compiled pattern, one for the
1.3827 + // offset to the string text, and one for the length.
1.3828 + // The compiled string has already been case folded.
1.3829 + {
1.3830 + const UChar *patternString = litText + opValue;
1.3831 + int32_t patternStringIdx = 0;
1.3832 +
1.3833 + op = (int32_t)pat[fp->fPatIdx];
1.3834 + fp->fPatIdx++;
1.3835 + opType = URX_TYPE(op);
1.3836 + opValue = URX_VAL(op);
1.3837 + U_ASSERT(opType == URX_STRING_LEN);
1.3838 + int32_t patternStringLen = opValue; // Length of the string from the pattern.
1.3839 +
1.3840 +
1.3841 + UChar32 cPattern;
1.3842 + UChar32 cText;
1.3843 + UBool success = TRUE;
1.3844 +
1.3845 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.3846 + CaseFoldingUTextIterator inputIterator(*fInputText);
1.3847 + while (patternStringIdx < patternStringLen) {
1.3848 + if (!inputIterator.inExpansion() && UTEXT_GETNATIVEINDEX(fInputText) >= fActiveLimit) {
1.3849 + success = FALSE;
1.3850 + fHitEnd = TRUE;
1.3851 + break;
1.3852 + }
1.3853 + U16_NEXT(patternString, patternStringIdx, patternStringLen, cPattern);
1.3854 + cText = inputIterator.next();
1.3855 + if (cText != cPattern) {
1.3856 + success = FALSE;
1.3857 + break;
1.3858 + }
1.3859 + }
1.3860 + if (inputIterator.inExpansion()) {
1.3861 + success = FALSE;
1.3862 + }
1.3863 +
1.3864 + if (success) {
1.3865 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3866 + } else {
1.3867 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3868 + }
1.3869 + }
1.3870 + }
1.3871 + break;
1.3872 +
1.3873 + case URX_LB_START:
1.3874 + {
1.3875 + // Entering a look-behind block.
1.3876 + // Save Stack Ptr, Input Pos.
1.3877 + // TODO: implement transparent bounds. Ticket #6067
1.3878 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.3879 + fData[opValue] = fStack->size();
1.3880 + fData[opValue+1] = fp->fInputIdx;
1.3881 + // Init the variable containing the start index for attempted matches.
1.3882 + fData[opValue+2] = -1;
1.3883 + // Save input string length, then reset to pin any matches to end at
1.3884 + // the current position.
1.3885 + fData[opValue+3] = fActiveLimit;
1.3886 + fActiveLimit = fp->fInputIdx;
1.3887 + }
1.3888 + break;
1.3889 +
1.3890 +
1.3891 + case URX_LB_CONT:
1.3892 + {
1.3893 + // Positive Look-Behind, at top of loop checking for matches of LB expression
1.3894 + // at all possible input starting positions.
1.3895 +
1.3896 + // Fetch the min and max possible match lengths. They are the operands
1.3897 + // of this op in the pattern.
1.3898 + int32_t minML = (int32_t)pat[fp->fPatIdx++];
1.3899 + int32_t maxML = (int32_t)pat[fp->fPatIdx++];
1.3900 + U_ASSERT(minML <= maxML);
1.3901 + U_ASSERT(minML >= 0);
1.3902 +
1.3903 + // Fetch (from data) the last input index where a match was attempted.
1.3904 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.3905 + int64_t *lbStartIdx = &fData[opValue+2];
1.3906 + if (*lbStartIdx < 0) {
1.3907 + // First time through loop.
1.3908 + *lbStartIdx = fp->fInputIdx - minML;
1.3909 + } else {
1.3910 + // 2nd through nth time through the loop.
1.3911 + // Back up start position for match by one.
1.3912 + if (*lbStartIdx == 0) {
1.3913 + (*lbStartIdx)--;
1.3914 + } else {
1.3915 + UTEXT_SETNATIVEINDEX(fInputText, *lbStartIdx);
1.3916 + (void)UTEXT_PREVIOUS32(fInputText);
1.3917 + *lbStartIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3918 + }
1.3919 + }
1.3920 +
1.3921 + if (*lbStartIdx < 0 || *lbStartIdx < fp->fInputIdx - maxML) {
1.3922 + // We have tried all potential match starting points without
1.3923 + // getting a match. Backtrack out, and out of the
1.3924 + // Look Behind altogether.
1.3925 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3926 + int64_t restoreInputLen = fData[opValue+3];
1.3927 + U_ASSERT(restoreInputLen >= fActiveLimit);
1.3928 + U_ASSERT(restoreInputLen <= fInputLength);
1.3929 + fActiveLimit = restoreInputLen;
1.3930 + break;
1.3931 + }
1.3932 +
1.3933 + // Save state to this URX_LB_CONT op, so failure to match will repeat the loop.
1.3934 + // (successful match will fall off the end of the loop.)
1.3935 + fp = StateSave(fp, fp->fPatIdx-3, status);
1.3936 + fp->fInputIdx = *lbStartIdx;
1.3937 + }
1.3938 + break;
1.3939 +
1.3940 + case URX_LB_END:
1.3941 + // End of a look-behind block, after a successful match.
1.3942 + {
1.3943 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.3944 + if (fp->fInputIdx != fActiveLimit) {
1.3945 + // The look-behind expression matched, but the match did not
1.3946 + // extend all the way to the point that we are looking behind from.
1.3947 + // FAIL out of here, which will take us back to the LB_CONT, which
1.3948 + // will retry the match starting at another position or fail
1.3949 + // the look-behind altogether, whichever is appropriate.
1.3950 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.3951 + break;
1.3952 + }
1.3953 +
1.3954 + // Look-behind match is good. Restore the orignal input string length,
1.3955 + // which had been truncated to pin the end of the lookbehind match to the
1.3956 + // position being looked-behind.
1.3957 + int64_t originalInputLen = fData[opValue+3];
1.3958 + U_ASSERT(originalInputLen >= fActiveLimit);
1.3959 + U_ASSERT(originalInputLen <= fInputLength);
1.3960 + fActiveLimit = originalInputLen;
1.3961 + }
1.3962 + break;
1.3963 +
1.3964 +
1.3965 + case URX_LBN_CONT:
1.3966 + {
1.3967 + // Negative Look-Behind, at top of loop checking for matches of LB expression
1.3968 + // at all possible input starting positions.
1.3969 +
1.3970 + // Fetch the extra parameters of this op.
1.3971 + int32_t minML = (int32_t)pat[fp->fPatIdx++];
1.3972 + int32_t maxML = (int32_t)pat[fp->fPatIdx++];
1.3973 + int32_t continueLoc = (int32_t)pat[fp->fPatIdx++];
1.3974 + continueLoc = URX_VAL(continueLoc);
1.3975 + U_ASSERT(minML <= maxML);
1.3976 + U_ASSERT(minML >= 0);
1.3977 + U_ASSERT(continueLoc > fp->fPatIdx);
1.3978 +
1.3979 + // Fetch (from data) the last input index where a match was attempted.
1.3980 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.3981 + int64_t *lbStartIdx = &fData[opValue+2];
1.3982 + if (*lbStartIdx < 0) {
1.3983 + // First time through loop.
1.3984 + *lbStartIdx = fp->fInputIdx - minML;
1.3985 + } else {
1.3986 + // 2nd through nth time through the loop.
1.3987 + // Back up start position for match by one.
1.3988 + if (*lbStartIdx == 0) {
1.3989 + (*lbStartIdx)--;
1.3990 + } else {
1.3991 + UTEXT_SETNATIVEINDEX(fInputText, *lbStartIdx);
1.3992 + (void)UTEXT_PREVIOUS32(fInputText);
1.3993 + *lbStartIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.3994 + }
1.3995 + }
1.3996 +
1.3997 + if (*lbStartIdx < 0 || *lbStartIdx < fp->fInputIdx - maxML) {
1.3998 + // We have tried all potential match starting points without
1.3999 + // getting a match, which means that the negative lookbehind as
1.4000 + // a whole has succeeded. Jump forward to the continue location
1.4001 + int64_t restoreInputLen = fData[opValue+3];
1.4002 + U_ASSERT(restoreInputLen >= fActiveLimit);
1.4003 + U_ASSERT(restoreInputLen <= fInputLength);
1.4004 + fActiveLimit = restoreInputLen;
1.4005 + fp->fPatIdx = continueLoc;
1.4006 + break;
1.4007 + }
1.4008 +
1.4009 + // Save state to this URX_LB_CONT op, so failure to match will repeat the loop.
1.4010 + // (successful match will cause a FAIL out of the loop altogether.)
1.4011 + fp = StateSave(fp, fp->fPatIdx-4, status);
1.4012 + fp->fInputIdx = *lbStartIdx;
1.4013 + }
1.4014 + break;
1.4015 +
1.4016 + case URX_LBN_END:
1.4017 + // End of a negative look-behind block, after a successful match.
1.4018 + {
1.4019 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.4020 + if (fp->fInputIdx != fActiveLimit) {
1.4021 + // The look-behind expression matched, but the match did not
1.4022 + // extend all the way to the point that we are looking behind from.
1.4023 + // FAIL out of here, which will take us back to the LB_CONT, which
1.4024 + // will retry the match starting at another position or succeed
1.4025 + // the look-behind altogether, whichever is appropriate.
1.4026 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4027 + break;
1.4028 + }
1.4029 +
1.4030 + // Look-behind expression matched, which means look-behind test as
1.4031 + // a whole Fails
1.4032 +
1.4033 + // Restore the orignal input string length, which had been truncated
1.4034 + // inorder to pin the end of the lookbehind match
1.4035 + // to the position being looked-behind.
1.4036 + int64_t originalInputLen = fData[opValue+3];
1.4037 + U_ASSERT(originalInputLen >= fActiveLimit);
1.4038 + U_ASSERT(originalInputLen <= fInputLength);
1.4039 + fActiveLimit = originalInputLen;
1.4040 +
1.4041 + // Restore original stack position, discarding any state saved
1.4042 + // by the successful pattern match.
1.4043 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.4044 + int32_t newStackSize = (int32_t)fData[opValue];
1.4045 + U_ASSERT(fStack->size() > newStackSize);
1.4046 + fStack->setSize(newStackSize);
1.4047 +
1.4048 + // FAIL, which will take control back to someplace
1.4049 + // prior to entering the look-behind test.
1.4050 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4051 + }
1.4052 + break;
1.4053 +
1.4054 +
1.4055 + case URX_LOOP_SR_I:
1.4056 + // Loop Initialization for the optimized implementation of
1.4057 + // [some character set]*
1.4058 + // This op scans through all matching input.
1.4059 + // The following LOOP_C op emulates stack unwinding if the following pattern fails.
1.4060 + {
1.4061 + U_ASSERT(opValue > 0 && opValue < sets->size());
1.4062 + Regex8BitSet *s8 = &fPattern->fSets8[opValue];
1.4063 + UnicodeSet *s = (UnicodeSet *)sets->elementAt(opValue);
1.4064 +
1.4065 + // Loop through input, until either the input is exhausted or
1.4066 + // we reach a character that is not a member of the set.
1.4067 + int64_t ix = fp->fInputIdx;
1.4068 + UTEXT_SETNATIVEINDEX(fInputText, ix);
1.4069 + for (;;) {
1.4070 + if (ix >= fActiveLimit) {
1.4071 + fHitEnd = TRUE;
1.4072 + break;
1.4073 + }
1.4074 + UChar32 c = UTEXT_NEXT32(fInputText);
1.4075 + if (c<256) {
1.4076 + if (s8->contains(c) == FALSE) {
1.4077 + break;
1.4078 + }
1.4079 + } else {
1.4080 + if (s->contains(c) == FALSE) {
1.4081 + break;
1.4082 + }
1.4083 + }
1.4084 + ix = UTEXT_GETNATIVEINDEX(fInputText);
1.4085 + }
1.4086 +
1.4087 + // If there were no matching characters, skip over the loop altogether.
1.4088 + // The loop doesn't run at all, a * op always succeeds.
1.4089 + if (ix == fp->fInputIdx) {
1.4090 + fp->fPatIdx++; // skip the URX_LOOP_C op.
1.4091 + break;
1.4092 + }
1.4093 +
1.4094 + // Peek ahead in the compiled pattern, to the URX_LOOP_C that
1.4095 + // must follow. It's operand is the stack location
1.4096 + // that holds the starting input index for the match of this [set]*
1.4097 + int32_t loopcOp = (int32_t)pat[fp->fPatIdx];
1.4098 + U_ASSERT(URX_TYPE(loopcOp) == URX_LOOP_C);
1.4099 + int32_t stackLoc = URX_VAL(loopcOp);
1.4100 + U_ASSERT(stackLoc >= 0 && stackLoc < fFrameSize);
1.4101 + fp->fExtra[stackLoc] = fp->fInputIdx;
1.4102 + fp->fInputIdx = ix;
1.4103 +
1.4104 + // Save State to the URX_LOOP_C op that follows this one,
1.4105 + // so that match failures in the following code will return to there.
1.4106 + // Then bump the pattern idx so the LOOP_C is skipped on the way out of here.
1.4107 + fp = StateSave(fp, fp->fPatIdx, status);
1.4108 + fp->fPatIdx++;
1.4109 + }
1.4110 + break;
1.4111 +
1.4112 +
1.4113 + case URX_LOOP_DOT_I:
1.4114 + // Loop Initialization for the optimized implementation of .*
1.4115 + // This op scans through all remaining input.
1.4116 + // The following LOOP_C op emulates stack unwinding if the following pattern fails.
1.4117 + {
1.4118 + // Loop through input until the input is exhausted (we reach an end-of-line)
1.4119 + // In DOTALL mode, we can just go straight to the end of the input.
1.4120 + int64_t ix;
1.4121 + if ((opValue & 1) == 1) {
1.4122 + // Dot-matches-All mode. Jump straight to the end of the string.
1.4123 + ix = fActiveLimit;
1.4124 + fHitEnd = TRUE;
1.4125 + } else {
1.4126 + // NOT DOT ALL mode. Line endings do not match '.'
1.4127 + // Scan forward until a line ending or end of input.
1.4128 + ix = fp->fInputIdx;
1.4129 + UTEXT_SETNATIVEINDEX(fInputText, ix);
1.4130 + for (;;) {
1.4131 + if (ix >= fActiveLimit) {
1.4132 + fHitEnd = TRUE;
1.4133 + break;
1.4134 + }
1.4135 + UChar32 c = UTEXT_NEXT32(fInputText);
1.4136 + if ((c & 0x7f) <= 0x29) { // Fast filter of non-new-line-s
1.4137 + if ((c == 0x0a) || // 0x0a is newline in both modes.
1.4138 + (((opValue & 2) == 0) && // IF not UNIX_LINES mode
1.4139 + (c<=0x0d && c>=0x0a)) || c==0x85 ||c==0x2028 || c==0x2029) {
1.4140 + // char is a line ending. Exit the scanning loop.
1.4141 + break;
1.4142 + }
1.4143 + }
1.4144 + ix = UTEXT_GETNATIVEINDEX(fInputText);
1.4145 + }
1.4146 + }
1.4147 +
1.4148 + // If there were no matching characters, skip over the loop altogether.
1.4149 + // The loop doesn't run at all, a * op always succeeds.
1.4150 + if (ix == fp->fInputIdx) {
1.4151 + fp->fPatIdx++; // skip the URX_LOOP_C op.
1.4152 + break;
1.4153 + }
1.4154 +
1.4155 + // Peek ahead in the compiled pattern, to the URX_LOOP_C that
1.4156 + // must follow. It's operand is the stack location
1.4157 + // that holds the starting input index for the match of this .*
1.4158 + int32_t loopcOp = (int32_t)pat[fp->fPatIdx];
1.4159 + U_ASSERT(URX_TYPE(loopcOp) == URX_LOOP_C);
1.4160 + int32_t stackLoc = URX_VAL(loopcOp);
1.4161 + U_ASSERT(stackLoc >= 0 && stackLoc < fFrameSize);
1.4162 + fp->fExtra[stackLoc] = fp->fInputIdx;
1.4163 + fp->fInputIdx = ix;
1.4164 +
1.4165 + // Save State to the URX_LOOP_C op that follows this one,
1.4166 + // so that match failures in the following code will return to there.
1.4167 + // Then bump the pattern idx so the LOOP_C is skipped on the way out of here.
1.4168 + fp = StateSave(fp, fp->fPatIdx, status);
1.4169 + fp->fPatIdx++;
1.4170 + }
1.4171 + break;
1.4172 +
1.4173 +
1.4174 + case URX_LOOP_C:
1.4175 + {
1.4176 + U_ASSERT(opValue>=0 && opValue<fFrameSize);
1.4177 + backSearchIndex = fp->fExtra[opValue];
1.4178 + U_ASSERT(backSearchIndex <= fp->fInputIdx);
1.4179 + if (backSearchIndex == fp->fInputIdx) {
1.4180 + // We've backed up the input idx to the point that the loop started.
1.4181 + // The loop is done. Leave here without saving state.
1.4182 + // Subsequent failures won't come back here.
1.4183 + break;
1.4184 + }
1.4185 + // Set up for the next iteration of the loop, with input index
1.4186 + // backed up by one from the last time through,
1.4187 + // and a state save to this instruction in case the following code fails again.
1.4188 + // (We're going backwards because this loop emulates stack unwinding, not
1.4189 + // the initial scan forward.)
1.4190 + U_ASSERT(fp->fInputIdx > 0);
1.4191 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.4192 + UChar32 prevC = UTEXT_PREVIOUS32(fInputText);
1.4193 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.4194 +
1.4195 + UChar32 twoPrevC = UTEXT_PREVIOUS32(fInputText);
1.4196 + if (prevC == 0x0a &&
1.4197 + fp->fInputIdx > backSearchIndex &&
1.4198 + twoPrevC == 0x0d) {
1.4199 + int32_t prevOp = (int32_t)pat[fp->fPatIdx-2];
1.4200 + if (URX_TYPE(prevOp) == URX_LOOP_DOT_I) {
1.4201 + // .*, stepping back over CRLF pair.
1.4202 + fp->fInputIdx = UTEXT_GETNATIVEINDEX(fInputText);
1.4203 + }
1.4204 + }
1.4205 +
1.4206 +
1.4207 + fp = StateSave(fp, fp->fPatIdx-1, status);
1.4208 + }
1.4209 + break;
1.4210 +
1.4211 +
1.4212 +
1.4213 + default:
1.4214 + // Trouble. The compiled pattern contains an entry with an
1.4215 + // unrecognized type tag.
1.4216 + U_ASSERT(FALSE);
1.4217 + }
1.4218 +
1.4219 + if (U_FAILURE(status)) {
1.4220 + isMatch = FALSE;
1.4221 + break;
1.4222 + }
1.4223 + }
1.4224 +
1.4225 +breakFromLoop:
1.4226 + fMatch = isMatch;
1.4227 + if (isMatch) {
1.4228 + fLastMatchEnd = fMatchEnd;
1.4229 + fMatchStart = startIdx;
1.4230 + fMatchEnd = fp->fInputIdx;
1.4231 + if (fTraceDebug) {
1.4232 + REGEX_RUN_DEBUG_PRINTF(("Match. start=%ld end=%ld\n\n", fMatchStart, fMatchEnd));
1.4233 + }
1.4234 + }
1.4235 + else
1.4236 + {
1.4237 + if (fTraceDebug) {
1.4238 + REGEX_RUN_DEBUG_PRINTF(("No match\n\n"));
1.4239 + }
1.4240 + }
1.4241 +
1.4242 + fFrame = fp; // The active stack frame when the engine stopped.
1.4243 + // Contains the capture group results that we need to
1.4244 + // access later.
1.4245 + return;
1.4246 +}
1.4247 +
1.4248 +
1.4249 +//--------------------------------------------------------------------------------
1.4250 +//
1.4251 +// MatchChunkAt This is the actual matching engine. Like MatchAt, but with the
1.4252 +// assumption that the entire string is available in the UText's
1.4253 +// chunk buffer. For now, that means we can use int32_t indexes,
1.4254 +// except for anything that needs to be saved (like group starts
1.4255 +// and ends).
1.4256 +//
1.4257 +// startIdx: begin matching a this index.
1.4258 +// toEnd: if true, match must extend to end of the input region
1.4259 +//
1.4260 +//--------------------------------------------------------------------------------
1.4261 +void RegexMatcher::MatchChunkAt(int32_t startIdx, UBool toEnd, UErrorCode &status) {
1.4262 + UBool isMatch = FALSE; // True if the we have a match.
1.4263 +
1.4264 + int32_t backSearchIndex = INT32_MAX; // used after greedy single-character matches for searching backwards
1.4265 +
1.4266 + int32_t op; // Operation from the compiled pattern, split into
1.4267 + int32_t opType; // the opcode
1.4268 + int32_t opValue; // and the operand value.
1.4269 +
1.4270 +#ifdef REGEX_RUN_DEBUG
1.4271 + if (fTraceDebug)
1.4272 + {
1.4273 + printf("MatchAt(startIdx=%d)\n", startIdx);
1.4274 + printf("Original Pattern: ");
1.4275 + UChar32 c = utext_next32From(fPattern->fPattern, 0);
1.4276 + while (c != U_SENTINEL) {
1.4277 + if (c<32 || c>256) {
1.4278 + c = '.';
1.4279 + }
1.4280 + REGEX_DUMP_DEBUG_PRINTF(("%c", c));
1.4281 +
1.4282 + c = UTEXT_NEXT32(fPattern->fPattern);
1.4283 + }
1.4284 + printf("\n");
1.4285 + printf("Input String: ");
1.4286 + c = utext_next32From(fInputText, 0);
1.4287 + while (c != U_SENTINEL) {
1.4288 + if (c<32 || c>256) {
1.4289 + c = '.';
1.4290 + }
1.4291 + printf("%c", c);
1.4292 +
1.4293 + c = UTEXT_NEXT32(fInputText);
1.4294 + }
1.4295 + printf("\n");
1.4296 + printf("\n");
1.4297 + }
1.4298 +#endif
1.4299 +
1.4300 + if (U_FAILURE(status)) {
1.4301 + return;
1.4302 + }
1.4303 +
1.4304 + // Cache frequently referenced items from the compiled pattern
1.4305 + //
1.4306 + int64_t *pat = fPattern->fCompiledPat->getBuffer();
1.4307 +
1.4308 + const UChar *litText = fPattern->fLiteralText.getBuffer();
1.4309 + UVector *sets = fPattern->fSets;
1.4310 +
1.4311 + const UChar *inputBuf = fInputText->chunkContents;
1.4312 +
1.4313 + fFrameSize = fPattern->fFrameSize;
1.4314 + REStackFrame *fp = resetStack();
1.4315 +
1.4316 + fp->fPatIdx = 0;
1.4317 + fp->fInputIdx = startIdx;
1.4318 +
1.4319 + // Zero out the pattern's static data
1.4320 + int32_t i;
1.4321 + for (i = 0; i<fPattern->fDataSize; i++) {
1.4322 + fData[i] = 0;
1.4323 + }
1.4324 +
1.4325 + //
1.4326 + // Main loop for interpreting the compiled pattern.
1.4327 + // One iteration of the loop per pattern operation performed.
1.4328 + //
1.4329 + for (;;) {
1.4330 +#if 0
1.4331 + if (_heapchk() != _HEAPOK) {
1.4332 + fprintf(stderr, "Heap Trouble\n");
1.4333 + }
1.4334 +#endif
1.4335 +
1.4336 + op = (int32_t)pat[fp->fPatIdx];
1.4337 + opType = URX_TYPE(op);
1.4338 + opValue = URX_VAL(op);
1.4339 +#ifdef REGEX_RUN_DEBUG
1.4340 + if (fTraceDebug) {
1.4341 + UTEXT_SETNATIVEINDEX(fInputText, fp->fInputIdx);
1.4342 + printf("inputIdx=%ld inputChar=%x sp=%3ld activeLimit=%ld ", fp->fInputIdx,
1.4343 + UTEXT_CURRENT32(fInputText), (int64_t *)fp-fStack->getBuffer(), fActiveLimit);
1.4344 + fPattern->dumpOp(fp->fPatIdx);
1.4345 + }
1.4346 +#endif
1.4347 + fp->fPatIdx++;
1.4348 +
1.4349 + switch (opType) {
1.4350 +
1.4351 +
1.4352 + case URX_NOP:
1.4353 + break;
1.4354 +
1.4355 +
1.4356 + case URX_BACKTRACK:
1.4357 + // Force a backtrack. In some circumstances, the pattern compiler
1.4358 + // will notice that the pattern can't possibly match anything, and will
1.4359 + // emit one of these at that point.
1.4360 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4361 + break;
1.4362 +
1.4363 +
1.4364 + case URX_ONECHAR:
1.4365 + if (fp->fInputIdx < fActiveLimit) {
1.4366 + UChar32 c;
1.4367 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4368 + if (c == opValue) {
1.4369 + break;
1.4370 + }
1.4371 + } else {
1.4372 + fHitEnd = TRUE;
1.4373 + }
1.4374 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4375 + break;
1.4376 +
1.4377 +
1.4378 + case URX_STRING:
1.4379 + {
1.4380 + // Test input against a literal string.
1.4381 + // Strings require two slots in the compiled pattern, one for the
1.4382 + // offset to the string text, and one for the length.
1.4383 + int32_t stringStartIdx = opValue;
1.4384 + int32_t stringLen;
1.4385 +
1.4386 + op = (int32_t)pat[fp->fPatIdx]; // Fetch the second operand
1.4387 + fp->fPatIdx++;
1.4388 + opType = URX_TYPE(op);
1.4389 + stringLen = URX_VAL(op);
1.4390 + U_ASSERT(opType == URX_STRING_LEN);
1.4391 + U_ASSERT(stringLen >= 2);
1.4392 +
1.4393 + const UChar * pInp = inputBuf + fp->fInputIdx;
1.4394 + const UChar * pInpLimit = inputBuf + fActiveLimit;
1.4395 + const UChar * pPat = litText+stringStartIdx;
1.4396 + const UChar * pEnd = pInp + stringLen;
1.4397 + UBool success = TRUE;
1.4398 + while (pInp < pEnd) {
1.4399 + if (pInp >= pInpLimit) {
1.4400 + fHitEnd = TRUE;
1.4401 + success = FALSE;
1.4402 + break;
1.4403 + }
1.4404 + if (*pInp++ != *pPat++) {
1.4405 + success = FALSE;
1.4406 + break;
1.4407 + }
1.4408 + }
1.4409 +
1.4410 + if (success) {
1.4411 + fp->fInputIdx += stringLen;
1.4412 + } else {
1.4413 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4414 + }
1.4415 + }
1.4416 + break;
1.4417 +
1.4418 +
1.4419 + case URX_STATE_SAVE:
1.4420 + fp = StateSave(fp, opValue, status);
1.4421 + break;
1.4422 +
1.4423 +
1.4424 + case URX_END:
1.4425 + // The match loop will exit via this path on a successful match,
1.4426 + // when we reach the end of the pattern.
1.4427 + if (toEnd && fp->fInputIdx != fActiveLimit) {
1.4428 + // The pattern matched, but not to the end of input. Try some more.
1.4429 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4430 + break;
1.4431 + }
1.4432 + isMatch = TRUE;
1.4433 + goto breakFromLoop;
1.4434 +
1.4435 + // Start and End Capture stack frame variables are laid out out like this:
1.4436 + // fp->fExtra[opValue] - The start of a completed capture group
1.4437 + // opValue+1 - The end of a completed capture group
1.4438 + // opValue+2 - the start of a capture group whose end
1.4439 + // has not yet been reached (and might not ever be).
1.4440 + case URX_START_CAPTURE:
1.4441 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-3);
1.4442 + fp->fExtra[opValue+2] = fp->fInputIdx;
1.4443 + break;
1.4444 +
1.4445 +
1.4446 + case URX_END_CAPTURE:
1.4447 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-3);
1.4448 + U_ASSERT(fp->fExtra[opValue+2] >= 0); // Start pos for this group must be set.
1.4449 + fp->fExtra[opValue] = fp->fExtra[opValue+2]; // Tentative start becomes real.
1.4450 + fp->fExtra[opValue+1] = fp->fInputIdx; // End position
1.4451 + U_ASSERT(fp->fExtra[opValue] <= fp->fExtra[opValue+1]);
1.4452 + break;
1.4453 +
1.4454 +
1.4455 + case URX_DOLLAR: // $, test for End of line
1.4456 + // or for position before new line at end of input
1.4457 + if (fp->fInputIdx < fAnchorLimit-2) {
1.4458 + // We are no where near the end of input. Fail.
1.4459 + // This is the common case. Keep it first.
1.4460 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4461 + break;
1.4462 + }
1.4463 + if (fp->fInputIdx >= fAnchorLimit) {
1.4464 + // We really are at the end of input. Success.
1.4465 + fHitEnd = TRUE;
1.4466 + fRequireEnd = TRUE;
1.4467 + break;
1.4468 + }
1.4469 +
1.4470 + // If we are positioned just before a new-line that is located at the
1.4471 + // end of input, succeed.
1.4472 + if (fp->fInputIdx == fAnchorLimit-1) {
1.4473 + UChar32 c;
1.4474 + U16_GET(inputBuf, fAnchorStart, fp->fInputIdx, fAnchorLimit, c);
1.4475 +
1.4476 + if ((c>=0x0a && c<=0x0d) || c==0x85 || c==0x2028 || c==0x2029) {
1.4477 + if ( !(c==0x0a && fp->fInputIdx>fAnchorStart && inputBuf[fp->fInputIdx-1]==0x0d)) {
1.4478 + // At new-line at end of input. Success
1.4479 + fHitEnd = TRUE;
1.4480 + fRequireEnd = TRUE;
1.4481 + break;
1.4482 + }
1.4483 + }
1.4484 + } else if (fp->fInputIdx == fAnchorLimit-2 &&
1.4485 + inputBuf[fp->fInputIdx]==0x0d && inputBuf[fp->fInputIdx+1]==0x0a) {
1.4486 + fHitEnd = TRUE;
1.4487 + fRequireEnd = TRUE;
1.4488 + break; // At CR/LF at end of input. Success
1.4489 + }
1.4490 +
1.4491 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4492 +
1.4493 + break;
1.4494 +
1.4495 +
1.4496 + case URX_DOLLAR_D: // $, test for End of Line, in UNIX_LINES mode.
1.4497 + if (fp->fInputIdx >= fAnchorLimit-1) {
1.4498 + // Either at the last character of input, or off the end.
1.4499 + if (fp->fInputIdx == fAnchorLimit-1) {
1.4500 + // At last char of input. Success if it's a new line.
1.4501 + if (inputBuf[fp->fInputIdx] == 0x0a) {
1.4502 + fHitEnd = TRUE;
1.4503 + fRequireEnd = TRUE;
1.4504 + break;
1.4505 + }
1.4506 + } else {
1.4507 + // Off the end of input. Success.
1.4508 + fHitEnd = TRUE;
1.4509 + fRequireEnd = TRUE;
1.4510 + break;
1.4511 + }
1.4512 + }
1.4513 +
1.4514 + // Not at end of input. Back-track out.
1.4515 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4516 + break;
1.4517 +
1.4518 +
1.4519 + case URX_DOLLAR_M: // $, test for End of line in multi-line mode
1.4520 + {
1.4521 + if (fp->fInputIdx >= fAnchorLimit) {
1.4522 + // We really are at the end of input. Success.
1.4523 + fHitEnd = TRUE;
1.4524 + fRequireEnd = TRUE;
1.4525 + break;
1.4526 + }
1.4527 + // If we are positioned just before a new-line, succeed.
1.4528 + // It makes no difference where the new-line is within the input.
1.4529 + UChar32 c = inputBuf[fp->fInputIdx];
1.4530 + if ((c>=0x0a && c<=0x0d) || c==0x85 ||c==0x2028 || c==0x2029) {
1.4531 + // At a line end, except for the odd chance of being in the middle of a CR/LF sequence
1.4532 + // In multi-line mode, hitting a new-line just before the end of input does not
1.4533 + // set the hitEnd or requireEnd flags
1.4534 + if ( !(c==0x0a && fp->fInputIdx>fAnchorStart && inputBuf[fp->fInputIdx-1]==0x0d)) {
1.4535 + break;
1.4536 + }
1.4537 + }
1.4538 + // not at a new line. Fail.
1.4539 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4540 + }
1.4541 + break;
1.4542 +
1.4543 +
1.4544 + case URX_DOLLAR_MD: // $, test for End of line in multi-line and UNIX_LINES mode
1.4545 + {
1.4546 + if (fp->fInputIdx >= fAnchorLimit) {
1.4547 + // We really are at the end of input. Success.
1.4548 + fHitEnd = TRUE;
1.4549 + fRequireEnd = TRUE; // Java set requireEnd in this case, even though
1.4550 + break; // adding a new-line would not lose the match.
1.4551 + }
1.4552 + // If we are not positioned just before a new-line, the test fails; backtrack out.
1.4553 + // It makes no difference where the new-line is within the input.
1.4554 + if (inputBuf[fp->fInputIdx] != 0x0a) {
1.4555 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4556 + }
1.4557 + }
1.4558 + break;
1.4559 +
1.4560 +
1.4561 + case URX_CARET: // ^, test for start of line
1.4562 + if (fp->fInputIdx != fAnchorStart) {
1.4563 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4564 + }
1.4565 + break;
1.4566 +
1.4567 +
1.4568 + case URX_CARET_M: // ^, test for start of line in mulit-line mode
1.4569 + {
1.4570 + if (fp->fInputIdx == fAnchorStart) {
1.4571 + // We are at the start input. Success.
1.4572 + break;
1.4573 + }
1.4574 + // Check whether character just before the current pos is a new-line
1.4575 + // unless we are at the end of input
1.4576 + UChar c = inputBuf[fp->fInputIdx - 1];
1.4577 + if ((fp->fInputIdx < fAnchorLimit) &&
1.4578 + ((c<=0x0d && c>=0x0a) || c==0x85 ||c==0x2028 || c==0x2029)) {
1.4579 + // It's a new-line. ^ is true. Success.
1.4580 + // TODO: what should be done with positions between a CR and LF?
1.4581 + break;
1.4582 + }
1.4583 + // Not at the start of a line. Fail.
1.4584 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4585 + }
1.4586 + break;
1.4587 +
1.4588 +
1.4589 + case URX_CARET_M_UNIX: // ^, test for start of line in mulit-line + Unix-line mode
1.4590 + {
1.4591 + U_ASSERT(fp->fInputIdx >= fAnchorStart);
1.4592 + if (fp->fInputIdx <= fAnchorStart) {
1.4593 + // We are at the start input. Success.
1.4594 + break;
1.4595 + }
1.4596 + // Check whether character just before the current pos is a new-line
1.4597 + U_ASSERT(fp->fInputIdx <= fAnchorLimit);
1.4598 + UChar c = inputBuf[fp->fInputIdx - 1];
1.4599 + if (c != 0x0a) {
1.4600 + // Not at the start of a line. Back-track out.
1.4601 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4602 + }
1.4603 + }
1.4604 + break;
1.4605 +
1.4606 + case URX_BACKSLASH_B: // Test for word boundaries
1.4607 + {
1.4608 + UBool success = isChunkWordBoundary((int32_t)fp->fInputIdx);
1.4609 + success ^= (UBool)(opValue != 0); // flip sense for \B
1.4610 + if (!success) {
1.4611 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4612 + }
1.4613 + }
1.4614 + break;
1.4615 +
1.4616 +
1.4617 + case URX_BACKSLASH_BU: // Test for word boundaries, Unicode-style
1.4618 + {
1.4619 + UBool success = isUWordBoundary(fp->fInputIdx);
1.4620 + success ^= (UBool)(opValue != 0); // flip sense for \B
1.4621 + if (!success) {
1.4622 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4623 + }
1.4624 + }
1.4625 + break;
1.4626 +
1.4627 +
1.4628 + case URX_BACKSLASH_D: // Test for decimal digit
1.4629 + {
1.4630 + if (fp->fInputIdx >= fActiveLimit) {
1.4631 + fHitEnd = TRUE;
1.4632 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4633 + break;
1.4634 + }
1.4635 +
1.4636 + UChar32 c;
1.4637 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4638 + int8_t ctype = u_charType(c); // TODO: make a unicode set for this. Will be faster.
1.4639 + UBool success = (ctype == U_DECIMAL_DIGIT_NUMBER);
1.4640 + success ^= (UBool)(opValue != 0); // flip sense for \D
1.4641 + if (!success) {
1.4642 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4643 + }
1.4644 + }
1.4645 + break;
1.4646 +
1.4647 +
1.4648 + case URX_BACKSLASH_G: // Test for position at end of previous match
1.4649 + if (!((fMatch && fp->fInputIdx==fMatchEnd) || (fMatch==FALSE && fp->fInputIdx==fActiveStart))) {
1.4650 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4651 + }
1.4652 + break;
1.4653 +
1.4654 +
1.4655 + case URX_BACKSLASH_X:
1.4656 + // Match a Grapheme, as defined by Unicode TR 29.
1.4657 + // Differs slightly from Perl, which consumes combining marks independently
1.4658 + // of context.
1.4659 + {
1.4660 +
1.4661 + // Fail if at end of input
1.4662 + if (fp->fInputIdx >= fActiveLimit) {
1.4663 + fHitEnd = TRUE;
1.4664 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4665 + break;
1.4666 + }
1.4667 +
1.4668 + // Examine (and consume) the current char.
1.4669 + // Dispatch into a little state machine, based on the char.
1.4670 + UChar32 c;
1.4671 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4672 + UnicodeSet **sets = fPattern->fStaticSets;
1.4673 + if (sets[URX_GC_NORMAL]->contains(c)) goto GC_Extend;
1.4674 + if (sets[URX_GC_CONTROL]->contains(c)) goto GC_Control;
1.4675 + if (sets[URX_GC_L]->contains(c)) goto GC_L;
1.4676 + if (sets[URX_GC_LV]->contains(c)) goto GC_V;
1.4677 + if (sets[URX_GC_LVT]->contains(c)) goto GC_T;
1.4678 + if (sets[URX_GC_V]->contains(c)) goto GC_V;
1.4679 + if (sets[URX_GC_T]->contains(c)) goto GC_T;
1.4680 + goto GC_Extend;
1.4681 +
1.4682 +
1.4683 +
1.4684 +GC_L:
1.4685 + if (fp->fInputIdx >= fActiveLimit) goto GC_Done;
1.4686 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4687 + if (sets[URX_GC_L]->contains(c)) goto GC_L;
1.4688 + if (sets[URX_GC_LV]->contains(c)) goto GC_V;
1.4689 + if (sets[URX_GC_LVT]->contains(c)) goto GC_T;
1.4690 + if (sets[URX_GC_V]->contains(c)) goto GC_V;
1.4691 + U16_PREV(inputBuf, 0, fp->fInputIdx, c);
1.4692 + goto GC_Extend;
1.4693 +
1.4694 +GC_V:
1.4695 + if (fp->fInputIdx >= fActiveLimit) goto GC_Done;
1.4696 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4697 + if (sets[URX_GC_V]->contains(c)) goto GC_V;
1.4698 + if (sets[URX_GC_T]->contains(c)) goto GC_T;
1.4699 + U16_PREV(inputBuf, 0, fp->fInputIdx, c);
1.4700 + goto GC_Extend;
1.4701 +
1.4702 +GC_T:
1.4703 + if (fp->fInputIdx >= fActiveLimit) goto GC_Done;
1.4704 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4705 + if (sets[URX_GC_T]->contains(c)) goto GC_T;
1.4706 + U16_PREV(inputBuf, 0, fp->fInputIdx, c);
1.4707 + goto GC_Extend;
1.4708 +
1.4709 +GC_Extend:
1.4710 + // Combining characters are consumed here
1.4711 + for (;;) {
1.4712 + if (fp->fInputIdx >= fActiveLimit) {
1.4713 + break;
1.4714 + }
1.4715 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4716 + if (sets[URX_GC_EXTEND]->contains(c) == FALSE) {
1.4717 + U16_BACK_1(inputBuf, 0, fp->fInputIdx);
1.4718 + break;
1.4719 + }
1.4720 + }
1.4721 + goto GC_Done;
1.4722 +
1.4723 +GC_Control:
1.4724 + // Most control chars stand alone (don't combine with combining chars),
1.4725 + // except for that CR/LF sequence is a single grapheme cluster.
1.4726 + if (c == 0x0d && fp->fInputIdx < fActiveLimit && inputBuf[fp->fInputIdx] == 0x0a) {
1.4727 + fp->fInputIdx++;
1.4728 + }
1.4729 +
1.4730 +GC_Done:
1.4731 + if (fp->fInputIdx >= fActiveLimit) {
1.4732 + fHitEnd = TRUE;
1.4733 + }
1.4734 + break;
1.4735 + }
1.4736 +
1.4737 +
1.4738 +
1.4739 +
1.4740 + case URX_BACKSLASH_Z: // Test for end of Input
1.4741 + if (fp->fInputIdx < fAnchorLimit) {
1.4742 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4743 + } else {
1.4744 + fHitEnd = TRUE;
1.4745 + fRequireEnd = TRUE;
1.4746 + }
1.4747 + break;
1.4748 +
1.4749 +
1.4750 +
1.4751 + case URX_STATIC_SETREF:
1.4752 + {
1.4753 + // Test input character against one of the predefined sets
1.4754 + // (Word Characters, for example)
1.4755 + // The high bit of the op value is a flag for the match polarity.
1.4756 + // 0: success if input char is in set.
1.4757 + // 1: success if input char is not in set.
1.4758 + if (fp->fInputIdx >= fActiveLimit) {
1.4759 + fHitEnd = TRUE;
1.4760 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4761 + break;
1.4762 + }
1.4763 +
1.4764 + UBool success = ((opValue & URX_NEG_SET) == URX_NEG_SET);
1.4765 + opValue &= ~URX_NEG_SET;
1.4766 + U_ASSERT(opValue > 0 && opValue < URX_LAST_SET);
1.4767 +
1.4768 + UChar32 c;
1.4769 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4770 + if (c < 256) {
1.4771 + Regex8BitSet *s8 = &fPattern->fStaticSets8[opValue];
1.4772 + if (s8->contains(c)) {
1.4773 + success = !success;
1.4774 + }
1.4775 + } else {
1.4776 + const UnicodeSet *s = fPattern->fStaticSets[opValue];
1.4777 + if (s->contains(c)) {
1.4778 + success = !success;
1.4779 + }
1.4780 + }
1.4781 + if (!success) {
1.4782 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4783 + }
1.4784 + }
1.4785 + break;
1.4786 +
1.4787 +
1.4788 + case URX_STAT_SETREF_N:
1.4789 + {
1.4790 + // Test input character for NOT being a member of one of
1.4791 + // the predefined sets (Word Characters, for example)
1.4792 + if (fp->fInputIdx >= fActiveLimit) {
1.4793 + fHitEnd = TRUE;
1.4794 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4795 + break;
1.4796 + }
1.4797 +
1.4798 + U_ASSERT(opValue > 0 && opValue < URX_LAST_SET);
1.4799 +
1.4800 + UChar32 c;
1.4801 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4802 + if (c < 256) {
1.4803 + Regex8BitSet *s8 = &fPattern->fStaticSets8[opValue];
1.4804 + if (s8->contains(c) == FALSE) {
1.4805 + break;
1.4806 + }
1.4807 + } else {
1.4808 + const UnicodeSet *s = fPattern->fStaticSets[opValue];
1.4809 + if (s->contains(c) == FALSE) {
1.4810 + break;
1.4811 + }
1.4812 + }
1.4813 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4814 + }
1.4815 + break;
1.4816 +
1.4817 +
1.4818 + case URX_SETREF:
1.4819 + {
1.4820 + if (fp->fInputIdx >= fActiveLimit) {
1.4821 + fHitEnd = TRUE;
1.4822 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4823 + break;
1.4824 + }
1.4825 +
1.4826 + U_ASSERT(opValue > 0 && opValue < sets->size());
1.4827 +
1.4828 + // There is input left. Pick up one char and test it for set membership.
1.4829 + UChar32 c;
1.4830 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4831 + if (c<256) {
1.4832 + Regex8BitSet *s8 = &fPattern->fSets8[opValue];
1.4833 + if (s8->contains(c)) {
1.4834 + // The character is in the set. A Match.
1.4835 + break;
1.4836 + }
1.4837 + } else {
1.4838 + UnicodeSet *s = (UnicodeSet *)sets->elementAt(opValue);
1.4839 + if (s->contains(c)) {
1.4840 + // The character is in the set. A Match.
1.4841 + break;
1.4842 + }
1.4843 + }
1.4844 +
1.4845 + // the character wasn't in the set.
1.4846 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4847 + }
1.4848 + break;
1.4849 +
1.4850 +
1.4851 + case URX_DOTANY:
1.4852 + {
1.4853 + // . matches anything, but stops at end-of-line.
1.4854 + if (fp->fInputIdx >= fActiveLimit) {
1.4855 + // At end of input. Match failed. Backtrack out.
1.4856 + fHitEnd = TRUE;
1.4857 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4858 + break;
1.4859 + }
1.4860 +
1.4861 + // There is input left. Advance over one char, unless we've hit end-of-line
1.4862 + UChar32 c;
1.4863 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4864 + if (((c & 0x7f) <= 0x29) && // First quickly bypass as many chars as possible
1.4865 + ((c<=0x0d && c>=0x0a) || c==0x85 ||c==0x2028 || c==0x2029)) {
1.4866 + // End of line in normal mode. . does not match.
1.4867 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4868 + break;
1.4869 + }
1.4870 + }
1.4871 + break;
1.4872 +
1.4873 +
1.4874 + case URX_DOTANY_ALL:
1.4875 + {
1.4876 + // . in dot-matches-all (including new lines) mode
1.4877 + if (fp->fInputIdx >= fActiveLimit) {
1.4878 + // At end of input. Match failed. Backtrack out.
1.4879 + fHitEnd = TRUE;
1.4880 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4881 + break;
1.4882 + }
1.4883 +
1.4884 + // There is input left. Advance over one char, except if we are
1.4885 + // at a cr/lf, advance over both of them.
1.4886 + UChar32 c;
1.4887 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4888 + if (c==0x0d && fp->fInputIdx < fActiveLimit) {
1.4889 + // In the case of a CR/LF, we need to advance over both.
1.4890 + if (inputBuf[fp->fInputIdx] == 0x0a) {
1.4891 + U16_FWD_1(inputBuf, fp->fInputIdx, fActiveLimit);
1.4892 + }
1.4893 + }
1.4894 + }
1.4895 + break;
1.4896 +
1.4897 +
1.4898 + case URX_DOTANY_UNIX:
1.4899 + {
1.4900 + // '.' operator, matches all, but stops at end-of-line.
1.4901 + // UNIX_LINES mode, so 0x0a is the only recognized line ending.
1.4902 + if (fp->fInputIdx >= fActiveLimit) {
1.4903 + // At end of input. Match failed. Backtrack out.
1.4904 + fHitEnd = TRUE;
1.4905 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4906 + break;
1.4907 + }
1.4908 +
1.4909 + // There is input left. Advance over one char, unless we've hit end-of-line
1.4910 + UChar32 c;
1.4911 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.4912 + if (c == 0x0a) {
1.4913 + // End of line in normal mode. '.' does not match the \n
1.4914 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4915 + }
1.4916 + }
1.4917 + break;
1.4918 +
1.4919 +
1.4920 + case URX_JMP:
1.4921 + fp->fPatIdx = opValue;
1.4922 + break;
1.4923 +
1.4924 + case URX_FAIL:
1.4925 + isMatch = FALSE;
1.4926 + goto breakFromLoop;
1.4927 +
1.4928 + case URX_JMP_SAV:
1.4929 + U_ASSERT(opValue < fPattern->fCompiledPat->size());
1.4930 + fp = StateSave(fp, fp->fPatIdx, status); // State save to loc following current
1.4931 + fp->fPatIdx = opValue; // Then JMP.
1.4932 + break;
1.4933 +
1.4934 + case URX_JMP_SAV_X:
1.4935 + // This opcode is used with (x)+, when x can match a zero length string.
1.4936 + // Same as JMP_SAV, except conditional on the match having made forward progress.
1.4937 + // Destination of the JMP must be a URX_STO_INP_LOC, from which we get the
1.4938 + // data address of the input position at the start of the loop.
1.4939 + {
1.4940 + U_ASSERT(opValue > 0 && opValue < fPattern->fCompiledPat->size());
1.4941 + int32_t stoOp = (int32_t)pat[opValue-1];
1.4942 + U_ASSERT(URX_TYPE(stoOp) == URX_STO_INP_LOC);
1.4943 + int32_t frameLoc = URX_VAL(stoOp);
1.4944 + U_ASSERT(frameLoc >= 0 && frameLoc < fFrameSize);
1.4945 + int32_t prevInputIdx = (int32_t)fp->fExtra[frameLoc];
1.4946 + U_ASSERT(prevInputIdx <= fp->fInputIdx);
1.4947 + if (prevInputIdx < fp->fInputIdx) {
1.4948 + // The match did make progress. Repeat the loop.
1.4949 + fp = StateSave(fp, fp->fPatIdx, status); // State save to loc following current
1.4950 + fp->fPatIdx = opValue;
1.4951 + fp->fExtra[frameLoc] = fp->fInputIdx;
1.4952 + }
1.4953 + // If the input position did not advance, we do nothing here,
1.4954 + // execution will fall out of the loop.
1.4955 + }
1.4956 + break;
1.4957 +
1.4958 + case URX_CTR_INIT:
1.4959 + {
1.4960 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-2);
1.4961 + fp->fExtra[opValue] = 0; // Set the loop counter variable to zero
1.4962 +
1.4963 + // Pick up the three extra operands that CTR_INIT has, and
1.4964 + // skip the pattern location counter past
1.4965 + int32_t instrOperandLoc = (int32_t)fp->fPatIdx;
1.4966 + fp->fPatIdx += 3;
1.4967 + int32_t loopLoc = URX_VAL(pat[instrOperandLoc]);
1.4968 + int32_t minCount = (int32_t)pat[instrOperandLoc+1];
1.4969 + int32_t maxCount = (int32_t)pat[instrOperandLoc+2];
1.4970 + U_ASSERT(minCount>=0);
1.4971 + U_ASSERT(maxCount>=minCount || maxCount==-1);
1.4972 + U_ASSERT(loopLoc>=fp->fPatIdx);
1.4973 +
1.4974 + if (minCount == 0) {
1.4975 + fp = StateSave(fp, loopLoc+1, status);
1.4976 + }
1.4977 + if (maxCount == -1) {
1.4978 + fp->fExtra[opValue+1] = fp->fInputIdx; // For loop breaking.
1.4979 + } else if (maxCount == 0) {
1.4980 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.4981 + }
1.4982 + }
1.4983 + break;
1.4984 +
1.4985 + case URX_CTR_LOOP:
1.4986 + {
1.4987 + U_ASSERT(opValue>0 && opValue < fp->fPatIdx-2);
1.4988 + int32_t initOp = (int32_t)pat[opValue];
1.4989 + U_ASSERT(URX_TYPE(initOp) == URX_CTR_INIT);
1.4990 + int64_t *pCounter = &fp->fExtra[URX_VAL(initOp)];
1.4991 + int32_t minCount = (int32_t)pat[opValue+2];
1.4992 + int32_t maxCount = (int32_t)pat[opValue+3];
1.4993 + (*pCounter)++;
1.4994 + if ((uint64_t)*pCounter >= (uint32_t)maxCount && maxCount != -1) {
1.4995 + U_ASSERT(*pCounter == maxCount);
1.4996 + break;
1.4997 + }
1.4998 + if (*pCounter >= minCount) {
1.4999 + if (maxCount == -1) {
1.5000 + // Loop has no hard upper bound.
1.5001 + // Check that it is progressing through the input, break if it is not.
1.5002 + int64_t *pLastInputIdx = &fp->fExtra[URX_VAL(initOp) + 1];
1.5003 + if (fp->fInputIdx == *pLastInputIdx) {
1.5004 + break;
1.5005 + } else {
1.5006 + *pLastInputIdx = fp->fInputIdx;
1.5007 + }
1.5008 + }
1.5009 + fp = StateSave(fp, fp->fPatIdx, status);
1.5010 + }
1.5011 + fp->fPatIdx = opValue + 4; // Loop back.
1.5012 + }
1.5013 + break;
1.5014 +
1.5015 + case URX_CTR_INIT_NG:
1.5016 + {
1.5017 + // Initialize a non-greedy loop
1.5018 + U_ASSERT(opValue >= 0 && opValue < fFrameSize-2);
1.5019 + fp->fExtra[opValue] = 0; // Set the loop counter variable to zero
1.5020 +
1.5021 + // Pick up the three extra operands that CTR_INIT_NG has, and
1.5022 + // skip the pattern location counter past
1.5023 + int32_t instrOperandLoc = (int32_t)fp->fPatIdx;
1.5024 + fp->fPatIdx += 3;
1.5025 + int32_t loopLoc = URX_VAL(pat[instrOperandLoc]);
1.5026 + int32_t minCount = (int32_t)pat[instrOperandLoc+1];
1.5027 + int32_t maxCount = (int32_t)pat[instrOperandLoc+2];
1.5028 + U_ASSERT(minCount>=0);
1.5029 + U_ASSERT(maxCount>=minCount || maxCount==-1);
1.5030 + U_ASSERT(loopLoc>fp->fPatIdx);
1.5031 + if (maxCount == -1) {
1.5032 + fp->fExtra[opValue+1] = fp->fInputIdx; // Save initial input index for loop breaking.
1.5033 + }
1.5034 +
1.5035 + if (minCount == 0) {
1.5036 + if (maxCount != 0) {
1.5037 + fp = StateSave(fp, fp->fPatIdx, status);
1.5038 + }
1.5039 + fp->fPatIdx = loopLoc+1; // Continue with stuff after repeated block
1.5040 + }
1.5041 + }
1.5042 + break;
1.5043 +
1.5044 + case URX_CTR_LOOP_NG:
1.5045 + {
1.5046 + // Non-greedy {min, max} loops
1.5047 + U_ASSERT(opValue>0 && opValue < fp->fPatIdx-2);
1.5048 + int32_t initOp = (int32_t)pat[opValue];
1.5049 + U_ASSERT(URX_TYPE(initOp) == URX_CTR_INIT_NG);
1.5050 + int64_t *pCounter = &fp->fExtra[URX_VAL(initOp)];
1.5051 + int32_t minCount = (int32_t)pat[opValue+2];
1.5052 + int32_t maxCount = (int32_t)pat[opValue+3];
1.5053 +
1.5054 + (*pCounter)++;
1.5055 + if ((uint64_t)*pCounter >= (uint32_t)maxCount && maxCount != -1) {
1.5056 + // The loop has matched the maximum permitted number of times.
1.5057 + // Break out of here with no action. Matching will
1.5058 + // continue with the following pattern.
1.5059 + U_ASSERT(*pCounter == maxCount);
1.5060 + break;
1.5061 + }
1.5062 +
1.5063 + if (*pCounter < minCount) {
1.5064 + // We haven't met the minimum number of matches yet.
1.5065 + // Loop back for another one.
1.5066 + fp->fPatIdx = opValue + 4; // Loop back.
1.5067 + } else {
1.5068 + // We do have the minimum number of matches.
1.5069 +
1.5070 + // If there is no upper bound on the loop iterations, check that the input index
1.5071 + // is progressing, and stop the loop if it is not.
1.5072 + if (maxCount == -1) {
1.5073 + int64_t *pLastInputIdx = &fp->fExtra[URX_VAL(initOp) + 1];
1.5074 + if (fp->fInputIdx == *pLastInputIdx) {
1.5075 + break;
1.5076 + }
1.5077 + *pLastInputIdx = fp->fInputIdx;
1.5078 + }
1.5079 +
1.5080 + // Loop Continuation: we will fall into the pattern following the loop
1.5081 + // (non-greedy, don't execute loop body first), but first do
1.5082 + // a state save to the top of the loop, so that a match failure
1.5083 + // in the following pattern will try another iteration of the loop.
1.5084 + fp = StateSave(fp, opValue + 4, status);
1.5085 + }
1.5086 + }
1.5087 + break;
1.5088 +
1.5089 + case URX_STO_SP:
1.5090 + U_ASSERT(opValue >= 0 && opValue < fPattern->fDataSize);
1.5091 + fData[opValue] = fStack->size();
1.5092 + break;
1.5093 +
1.5094 + case URX_LD_SP:
1.5095 + {
1.5096 + U_ASSERT(opValue >= 0 && opValue < fPattern->fDataSize);
1.5097 + int32_t newStackSize = (int32_t)fData[opValue];
1.5098 + U_ASSERT(newStackSize <= fStack->size());
1.5099 + int64_t *newFP = fStack->getBuffer() + newStackSize - fFrameSize;
1.5100 + if (newFP == (int64_t *)fp) {
1.5101 + break;
1.5102 + }
1.5103 + int32_t i;
1.5104 + for (i=0; i<fFrameSize; i++) {
1.5105 + newFP[i] = ((int64_t *)fp)[i];
1.5106 + }
1.5107 + fp = (REStackFrame *)newFP;
1.5108 + fStack->setSize(newStackSize);
1.5109 + }
1.5110 + break;
1.5111 +
1.5112 + case URX_BACKREF:
1.5113 + {
1.5114 + U_ASSERT(opValue < fFrameSize);
1.5115 + int64_t groupStartIdx = fp->fExtra[opValue];
1.5116 + int64_t groupEndIdx = fp->fExtra[opValue+1];
1.5117 + U_ASSERT(groupStartIdx <= groupEndIdx);
1.5118 + int64_t inputIndex = fp->fInputIdx;
1.5119 + if (groupStartIdx < 0) {
1.5120 + // This capture group has not participated in the match thus far,
1.5121 + fp = (REStackFrame *)fStack->popFrame(fFrameSize); // FAIL, no match.
1.5122 + break;
1.5123 + }
1.5124 + UBool success = TRUE;
1.5125 + for (int64_t groupIndex = groupStartIdx; groupIndex < groupEndIdx; ++groupIndex,++inputIndex) {
1.5126 + if (inputIndex >= fActiveLimit) {
1.5127 + success = FALSE;
1.5128 + fHitEnd = TRUE;
1.5129 + break;
1.5130 + }
1.5131 + if (inputBuf[groupIndex] != inputBuf[inputIndex]) {
1.5132 + success = FALSE;
1.5133 + break;
1.5134 + }
1.5135 + }
1.5136 + if (success) {
1.5137 + fp->fInputIdx = inputIndex;
1.5138 + } else {
1.5139 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5140 + }
1.5141 + }
1.5142 + break;
1.5143 +
1.5144 + case URX_BACKREF_I:
1.5145 + {
1.5146 + U_ASSERT(opValue < fFrameSize);
1.5147 + int64_t groupStartIdx = fp->fExtra[opValue];
1.5148 + int64_t groupEndIdx = fp->fExtra[opValue+1];
1.5149 + U_ASSERT(groupStartIdx <= groupEndIdx);
1.5150 + if (groupStartIdx < 0) {
1.5151 + // This capture group has not participated in the match thus far,
1.5152 + fp = (REStackFrame *)fStack->popFrame(fFrameSize); // FAIL, no match.
1.5153 + break;
1.5154 + }
1.5155 + CaseFoldingUCharIterator captureGroupItr(inputBuf, groupStartIdx, groupEndIdx);
1.5156 + CaseFoldingUCharIterator inputItr(inputBuf, fp->fInputIdx, fActiveLimit);
1.5157 +
1.5158 + // Note: if the capture group match was of an empty string the backref
1.5159 + // match succeeds. Verified by testing: Perl matches succeed
1.5160 + // in this case, so we do too.
1.5161 +
1.5162 + UBool success = TRUE;
1.5163 + for (;;) {
1.5164 + UChar32 captureGroupChar = captureGroupItr.next();
1.5165 + if (captureGroupChar == U_SENTINEL) {
1.5166 + success = TRUE;
1.5167 + break;
1.5168 + }
1.5169 + UChar32 inputChar = inputItr.next();
1.5170 + if (inputChar == U_SENTINEL) {
1.5171 + success = FALSE;
1.5172 + fHitEnd = TRUE;
1.5173 + break;
1.5174 + }
1.5175 + if (inputChar != captureGroupChar) {
1.5176 + success = FALSE;
1.5177 + break;
1.5178 + }
1.5179 + }
1.5180 +
1.5181 + if (success && inputItr.inExpansion()) {
1.5182 + // We otained a match by consuming part of a string obtained from
1.5183 + // case-folding a single code point of the input text.
1.5184 + // This does not count as an overall match.
1.5185 + success = FALSE;
1.5186 + }
1.5187 +
1.5188 + if (success) {
1.5189 + fp->fInputIdx = inputItr.getIndex();
1.5190 + } else {
1.5191 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5192 + }
1.5193 + }
1.5194 + break;
1.5195 +
1.5196 + case URX_STO_INP_LOC:
1.5197 + {
1.5198 + U_ASSERT(opValue >= 0 && opValue < fFrameSize);
1.5199 + fp->fExtra[opValue] = fp->fInputIdx;
1.5200 + }
1.5201 + break;
1.5202 +
1.5203 + case URX_JMPX:
1.5204 + {
1.5205 + int32_t instrOperandLoc = (int32_t)fp->fPatIdx;
1.5206 + fp->fPatIdx += 1;
1.5207 + int32_t dataLoc = URX_VAL(pat[instrOperandLoc]);
1.5208 + U_ASSERT(dataLoc >= 0 && dataLoc < fFrameSize);
1.5209 + int32_t savedInputIdx = (int32_t)fp->fExtra[dataLoc];
1.5210 + U_ASSERT(savedInputIdx <= fp->fInputIdx);
1.5211 + if (savedInputIdx < fp->fInputIdx) {
1.5212 + fp->fPatIdx = opValue; // JMP
1.5213 + } else {
1.5214 + fp = (REStackFrame *)fStack->popFrame(fFrameSize); // FAIL, no progress in loop.
1.5215 + }
1.5216 + }
1.5217 + break;
1.5218 +
1.5219 + case URX_LA_START:
1.5220 + {
1.5221 + // Entering a lookahead block.
1.5222 + // Save Stack Ptr, Input Pos.
1.5223 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5224 + fData[opValue] = fStack->size();
1.5225 + fData[opValue+1] = fp->fInputIdx;
1.5226 + fActiveStart = fLookStart; // Set the match region change for
1.5227 + fActiveLimit = fLookLimit; // transparent bounds.
1.5228 + }
1.5229 + break;
1.5230 +
1.5231 + case URX_LA_END:
1.5232 + {
1.5233 + // Leaving a look-ahead block.
1.5234 + // restore Stack Ptr, Input Pos to positions they had on entry to block.
1.5235 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5236 + int32_t stackSize = fStack->size();
1.5237 + int32_t newStackSize = (int32_t)fData[opValue];
1.5238 + U_ASSERT(stackSize >= newStackSize);
1.5239 + if (stackSize > newStackSize) {
1.5240 + // Copy the current top frame back to the new (cut back) top frame.
1.5241 + // This makes the capture groups from within the look-ahead
1.5242 + // expression available.
1.5243 + int64_t *newFP = fStack->getBuffer() + newStackSize - fFrameSize;
1.5244 + int32_t i;
1.5245 + for (i=0; i<fFrameSize; i++) {
1.5246 + newFP[i] = ((int64_t *)fp)[i];
1.5247 + }
1.5248 + fp = (REStackFrame *)newFP;
1.5249 + fStack->setSize(newStackSize);
1.5250 + }
1.5251 + fp->fInputIdx = fData[opValue+1];
1.5252 +
1.5253 + // Restore the active region bounds in the input string; they may have
1.5254 + // been changed because of transparent bounds on a Region.
1.5255 + fActiveStart = fRegionStart;
1.5256 + fActiveLimit = fRegionLimit;
1.5257 + }
1.5258 + break;
1.5259 +
1.5260 + case URX_ONECHAR_I:
1.5261 + if (fp->fInputIdx < fActiveLimit) {
1.5262 + UChar32 c;
1.5263 + U16_NEXT(inputBuf, fp->fInputIdx, fActiveLimit, c);
1.5264 + if (u_foldCase(c, U_FOLD_CASE_DEFAULT) == opValue) {
1.5265 + break;
1.5266 + }
1.5267 + } else {
1.5268 + fHitEnd = TRUE;
1.5269 + }
1.5270 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5271 + break;
1.5272 +
1.5273 + case URX_STRING_I:
1.5274 + // Case-insensitive test input against a literal string.
1.5275 + // Strings require two slots in the compiled pattern, one for the
1.5276 + // offset to the string text, and one for the length.
1.5277 + // The compiled string has already been case folded.
1.5278 + {
1.5279 + const UChar *patternString = litText + opValue;
1.5280 +
1.5281 + op = (int32_t)pat[fp->fPatIdx];
1.5282 + fp->fPatIdx++;
1.5283 + opType = URX_TYPE(op);
1.5284 + opValue = URX_VAL(op);
1.5285 + U_ASSERT(opType == URX_STRING_LEN);
1.5286 + int32_t patternStringLen = opValue; // Length of the string from the pattern.
1.5287 +
1.5288 + UChar32 cText;
1.5289 + UChar32 cPattern;
1.5290 + UBool success = TRUE;
1.5291 + int32_t patternStringIdx = 0;
1.5292 + CaseFoldingUCharIterator inputIterator(inputBuf, fp->fInputIdx, fActiveLimit);
1.5293 + while (patternStringIdx < patternStringLen) {
1.5294 + U16_NEXT(patternString, patternStringIdx, patternStringLen, cPattern);
1.5295 + cText = inputIterator.next();
1.5296 + if (cText != cPattern) {
1.5297 + success = FALSE;
1.5298 + if (cText == U_SENTINEL) {
1.5299 + fHitEnd = TRUE;
1.5300 + }
1.5301 + break;
1.5302 + }
1.5303 + }
1.5304 + if (inputIterator.inExpansion()) {
1.5305 + success = FALSE;
1.5306 + }
1.5307 +
1.5308 + if (success) {
1.5309 + fp->fInputIdx = inputIterator.getIndex();
1.5310 + } else {
1.5311 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5312 + }
1.5313 + }
1.5314 + break;
1.5315 +
1.5316 + case URX_LB_START:
1.5317 + {
1.5318 + // Entering a look-behind block.
1.5319 + // Save Stack Ptr, Input Pos.
1.5320 + // TODO: implement transparent bounds. Ticket #6067
1.5321 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5322 + fData[opValue] = fStack->size();
1.5323 + fData[opValue+1] = fp->fInputIdx;
1.5324 + // Init the variable containing the start index for attempted matches.
1.5325 + fData[opValue+2] = -1;
1.5326 + // Save input string length, then reset to pin any matches to end at
1.5327 + // the current position.
1.5328 + fData[opValue+3] = fActiveLimit;
1.5329 + fActiveLimit = fp->fInputIdx;
1.5330 + }
1.5331 + break;
1.5332 +
1.5333 +
1.5334 + case URX_LB_CONT:
1.5335 + {
1.5336 + // Positive Look-Behind, at top of loop checking for matches of LB expression
1.5337 + // at all possible input starting positions.
1.5338 +
1.5339 + // Fetch the min and max possible match lengths. They are the operands
1.5340 + // of this op in the pattern.
1.5341 + int32_t minML = (int32_t)pat[fp->fPatIdx++];
1.5342 + int32_t maxML = (int32_t)pat[fp->fPatIdx++];
1.5343 + U_ASSERT(minML <= maxML);
1.5344 + U_ASSERT(minML >= 0);
1.5345 +
1.5346 + // Fetch (from data) the last input index where a match was attempted.
1.5347 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5348 + int64_t *lbStartIdx = &fData[opValue+2];
1.5349 + if (*lbStartIdx < 0) {
1.5350 + // First time through loop.
1.5351 + *lbStartIdx = fp->fInputIdx - minML;
1.5352 + } else {
1.5353 + // 2nd through nth time through the loop.
1.5354 + // Back up start position for match by one.
1.5355 + if (*lbStartIdx == 0) {
1.5356 + (*lbStartIdx)--;
1.5357 + } else {
1.5358 + U16_BACK_1(inputBuf, 0, *lbStartIdx);
1.5359 + }
1.5360 + }
1.5361 +
1.5362 + if (*lbStartIdx < 0 || *lbStartIdx < fp->fInputIdx - maxML) {
1.5363 + // We have tried all potential match starting points without
1.5364 + // getting a match. Backtrack out, and out of the
1.5365 + // Look Behind altogether.
1.5366 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5367 + int64_t restoreInputLen = fData[opValue+3];
1.5368 + U_ASSERT(restoreInputLen >= fActiveLimit);
1.5369 + U_ASSERT(restoreInputLen <= fInputLength);
1.5370 + fActiveLimit = restoreInputLen;
1.5371 + break;
1.5372 + }
1.5373 +
1.5374 + // Save state to this URX_LB_CONT op, so failure to match will repeat the loop.
1.5375 + // (successful match will fall off the end of the loop.)
1.5376 + fp = StateSave(fp, fp->fPatIdx-3, status);
1.5377 + fp->fInputIdx = *lbStartIdx;
1.5378 + }
1.5379 + break;
1.5380 +
1.5381 + case URX_LB_END:
1.5382 + // End of a look-behind block, after a successful match.
1.5383 + {
1.5384 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5385 + if (fp->fInputIdx != fActiveLimit) {
1.5386 + // The look-behind expression matched, but the match did not
1.5387 + // extend all the way to the point that we are looking behind from.
1.5388 + // FAIL out of here, which will take us back to the LB_CONT, which
1.5389 + // will retry the match starting at another position or fail
1.5390 + // the look-behind altogether, whichever is appropriate.
1.5391 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5392 + break;
1.5393 + }
1.5394 +
1.5395 + // Look-behind match is good. Restore the orignal input string length,
1.5396 + // which had been truncated to pin the end of the lookbehind match to the
1.5397 + // position being looked-behind.
1.5398 + int64_t originalInputLen = fData[opValue+3];
1.5399 + U_ASSERT(originalInputLen >= fActiveLimit);
1.5400 + U_ASSERT(originalInputLen <= fInputLength);
1.5401 + fActiveLimit = originalInputLen;
1.5402 + }
1.5403 + break;
1.5404 +
1.5405 +
1.5406 + case URX_LBN_CONT:
1.5407 + {
1.5408 + // Negative Look-Behind, at top of loop checking for matches of LB expression
1.5409 + // at all possible input starting positions.
1.5410 +
1.5411 + // Fetch the extra parameters of this op.
1.5412 + int32_t minML = (int32_t)pat[fp->fPatIdx++];
1.5413 + int32_t maxML = (int32_t)pat[fp->fPatIdx++];
1.5414 + int32_t continueLoc = (int32_t)pat[fp->fPatIdx++];
1.5415 + continueLoc = URX_VAL(continueLoc);
1.5416 + U_ASSERT(minML <= maxML);
1.5417 + U_ASSERT(minML >= 0);
1.5418 + U_ASSERT(continueLoc > fp->fPatIdx);
1.5419 +
1.5420 + // Fetch (from data) the last input index where a match was attempted.
1.5421 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5422 + int64_t *lbStartIdx = &fData[opValue+2];
1.5423 + if (*lbStartIdx < 0) {
1.5424 + // First time through loop.
1.5425 + *lbStartIdx = fp->fInputIdx - minML;
1.5426 + } else {
1.5427 + // 2nd through nth time through the loop.
1.5428 + // Back up start position for match by one.
1.5429 + if (*lbStartIdx == 0) {
1.5430 + (*lbStartIdx)--; // Because U16_BACK is unsafe starting at 0.
1.5431 + } else {
1.5432 + U16_BACK_1(inputBuf, 0, *lbStartIdx);
1.5433 + }
1.5434 + }
1.5435 +
1.5436 + if (*lbStartIdx < 0 || *lbStartIdx < fp->fInputIdx - maxML) {
1.5437 + // We have tried all potential match starting points without
1.5438 + // getting a match, which means that the negative lookbehind as
1.5439 + // a whole has succeeded. Jump forward to the continue location
1.5440 + int64_t restoreInputLen = fData[opValue+3];
1.5441 + U_ASSERT(restoreInputLen >= fActiveLimit);
1.5442 + U_ASSERT(restoreInputLen <= fInputLength);
1.5443 + fActiveLimit = restoreInputLen;
1.5444 + fp->fPatIdx = continueLoc;
1.5445 + break;
1.5446 + }
1.5447 +
1.5448 + // Save state to this URX_LB_CONT op, so failure to match will repeat the loop.
1.5449 + // (successful match will cause a FAIL out of the loop altogether.)
1.5450 + fp = StateSave(fp, fp->fPatIdx-4, status);
1.5451 + fp->fInputIdx = *lbStartIdx;
1.5452 + }
1.5453 + break;
1.5454 +
1.5455 + case URX_LBN_END:
1.5456 + // End of a negative look-behind block, after a successful match.
1.5457 + {
1.5458 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5459 + if (fp->fInputIdx != fActiveLimit) {
1.5460 + // The look-behind expression matched, but the match did not
1.5461 + // extend all the way to the point that we are looking behind from.
1.5462 + // FAIL out of here, which will take us back to the LB_CONT, which
1.5463 + // will retry the match starting at another position or succeed
1.5464 + // the look-behind altogether, whichever is appropriate.
1.5465 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5466 + break;
1.5467 + }
1.5468 +
1.5469 + // Look-behind expression matched, which means look-behind test as
1.5470 + // a whole Fails
1.5471 +
1.5472 + // Restore the orignal input string length, which had been truncated
1.5473 + // inorder to pin the end of the lookbehind match
1.5474 + // to the position being looked-behind.
1.5475 + int64_t originalInputLen = fData[opValue+3];
1.5476 + U_ASSERT(originalInputLen >= fActiveLimit);
1.5477 + U_ASSERT(originalInputLen <= fInputLength);
1.5478 + fActiveLimit = originalInputLen;
1.5479 +
1.5480 + // Restore original stack position, discarding any state saved
1.5481 + // by the successful pattern match.
1.5482 + U_ASSERT(opValue>=0 && opValue+1<fPattern->fDataSize);
1.5483 + int32_t newStackSize = (int32_t)fData[opValue];
1.5484 + U_ASSERT(fStack->size() > newStackSize);
1.5485 + fStack->setSize(newStackSize);
1.5486 +
1.5487 + // FAIL, which will take control back to someplace
1.5488 + // prior to entering the look-behind test.
1.5489 + fp = (REStackFrame *)fStack->popFrame(fFrameSize);
1.5490 + }
1.5491 + break;
1.5492 +
1.5493 +
1.5494 + case URX_LOOP_SR_I:
1.5495 + // Loop Initialization for the optimized implementation of
1.5496 + // [some character set]*
1.5497 + // This op scans through all matching input.
1.5498 + // The following LOOP_C op emulates stack unwinding if the following pattern fails.
1.5499 + {
1.5500 + U_ASSERT(opValue > 0 && opValue < sets->size());
1.5501 + Regex8BitSet *s8 = &fPattern->fSets8[opValue];
1.5502 + UnicodeSet *s = (UnicodeSet *)sets->elementAt(opValue);
1.5503 +
1.5504 + // Loop through input, until either the input is exhausted or
1.5505 + // we reach a character that is not a member of the set.
1.5506 + int32_t ix = (int32_t)fp->fInputIdx;
1.5507 + for (;;) {
1.5508 + if (ix >= fActiveLimit) {
1.5509 + fHitEnd = TRUE;
1.5510 + break;
1.5511 + }
1.5512 + UChar32 c;
1.5513 + U16_NEXT(inputBuf, ix, fActiveLimit, c);
1.5514 + if (c<256) {
1.5515 + if (s8->contains(c) == FALSE) {
1.5516 + U16_BACK_1(inputBuf, 0, ix);
1.5517 + break;
1.5518 + }
1.5519 + } else {
1.5520 + if (s->contains(c) == FALSE) {
1.5521 + U16_BACK_1(inputBuf, 0, ix);
1.5522 + break;
1.5523 + }
1.5524 + }
1.5525 + }
1.5526 +
1.5527 + // If there were no matching characters, skip over the loop altogether.
1.5528 + // The loop doesn't run at all, a * op always succeeds.
1.5529 + if (ix == fp->fInputIdx) {
1.5530 + fp->fPatIdx++; // skip the URX_LOOP_C op.
1.5531 + break;
1.5532 + }
1.5533 +
1.5534 + // Peek ahead in the compiled pattern, to the URX_LOOP_C that
1.5535 + // must follow. It's operand is the stack location
1.5536 + // that holds the starting input index for the match of this [set]*
1.5537 + int32_t loopcOp = (int32_t)pat[fp->fPatIdx];
1.5538 + U_ASSERT(URX_TYPE(loopcOp) == URX_LOOP_C);
1.5539 + int32_t stackLoc = URX_VAL(loopcOp);
1.5540 + U_ASSERT(stackLoc >= 0 && stackLoc < fFrameSize);
1.5541 + fp->fExtra[stackLoc] = fp->fInputIdx;
1.5542 + fp->fInputIdx = ix;
1.5543 +
1.5544 + // Save State to the URX_LOOP_C op that follows this one,
1.5545 + // so that match failures in the following code will return to there.
1.5546 + // Then bump the pattern idx so the LOOP_C is skipped on the way out of here.
1.5547 + fp = StateSave(fp, fp->fPatIdx, status);
1.5548 + fp->fPatIdx++;
1.5549 + }
1.5550 + break;
1.5551 +
1.5552 +
1.5553 + case URX_LOOP_DOT_I:
1.5554 + // Loop Initialization for the optimized implementation of .*
1.5555 + // This op scans through all remaining input.
1.5556 + // The following LOOP_C op emulates stack unwinding if the following pattern fails.
1.5557 + {
1.5558 + // Loop through input until the input is exhausted (we reach an end-of-line)
1.5559 + // In DOTALL mode, we can just go straight to the end of the input.
1.5560 + int32_t ix;
1.5561 + if ((opValue & 1) == 1) {
1.5562 + // Dot-matches-All mode. Jump straight to the end of the string.
1.5563 + ix = (int32_t)fActiveLimit;
1.5564 + fHitEnd = TRUE;
1.5565 + } else {
1.5566 + // NOT DOT ALL mode. Line endings do not match '.'
1.5567 + // Scan forward until a line ending or end of input.
1.5568 + ix = (int32_t)fp->fInputIdx;
1.5569 + for (;;) {
1.5570 + if (ix >= fActiveLimit) {
1.5571 + fHitEnd = TRUE;
1.5572 + break;
1.5573 + }
1.5574 + UChar32 c;
1.5575 + U16_NEXT(inputBuf, ix, fActiveLimit, c); // c = inputBuf[ix++]
1.5576 + if ((c & 0x7f) <= 0x29) { // Fast filter of non-new-line-s
1.5577 + if ((c == 0x0a) || // 0x0a is newline in both modes.
1.5578 + (((opValue & 2) == 0) && // IF not UNIX_LINES mode
1.5579 + ((c<=0x0d && c>=0x0a) || c==0x85 || c==0x2028 || c==0x2029))) {
1.5580 + // char is a line ending. Put the input pos back to the
1.5581 + // line ending char, and exit the scanning loop.
1.5582 + U16_BACK_1(inputBuf, 0, ix);
1.5583 + break;
1.5584 + }
1.5585 + }
1.5586 + }
1.5587 + }
1.5588 +
1.5589 + // If there were no matching characters, skip over the loop altogether.
1.5590 + // The loop doesn't run at all, a * op always succeeds.
1.5591 + if (ix == fp->fInputIdx) {
1.5592 + fp->fPatIdx++; // skip the URX_LOOP_C op.
1.5593 + break;
1.5594 + }
1.5595 +
1.5596 + // Peek ahead in the compiled pattern, to the URX_LOOP_C that
1.5597 + // must follow. It's operand is the stack location
1.5598 + // that holds the starting input index for the match of this .*
1.5599 + int32_t loopcOp = (int32_t)pat[fp->fPatIdx];
1.5600 + U_ASSERT(URX_TYPE(loopcOp) == URX_LOOP_C);
1.5601 + int32_t stackLoc = URX_VAL(loopcOp);
1.5602 + U_ASSERT(stackLoc >= 0 && stackLoc < fFrameSize);
1.5603 + fp->fExtra[stackLoc] = fp->fInputIdx;
1.5604 + fp->fInputIdx = ix;
1.5605 +
1.5606 + // Save State to the URX_LOOP_C op that follows this one,
1.5607 + // so that match failures in the following code will return to there.
1.5608 + // Then bump the pattern idx so the LOOP_C is skipped on the way out of here.
1.5609 + fp = StateSave(fp, fp->fPatIdx, status);
1.5610 + fp->fPatIdx++;
1.5611 + }
1.5612 + break;
1.5613 +
1.5614 +
1.5615 + case URX_LOOP_C:
1.5616 + {
1.5617 + U_ASSERT(opValue>=0 && opValue<fFrameSize);
1.5618 + backSearchIndex = (int32_t)fp->fExtra[opValue];
1.5619 + U_ASSERT(backSearchIndex <= fp->fInputIdx);
1.5620 + if (backSearchIndex == fp->fInputIdx) {
1.5621 + // We've backed up the input idx to the point that the loop started.
1.5622 + // The loop is done. Leave here without saving state.
1.5623 + // Subsequent failures won't come back here.
1.5624 + break;
1.5625 + }
1.5626 + // Set up for the next iteration of the loop, with input index
1.5627 + // backed up by one from the last time through,
1.5628 + // and a state save to this instruction in case the following code fails again.
1.5629 + // (We're going backwards because this loop emulates stack unwinding, not
1.5630 + // the initial scan forward.)
1.5631 + U_ASSERT(fp->fInputIdx > 0);
1.5632 + UChar32 prevC;
1.5633 + U16_PREV(inputBuf, 0, fp->fInputIdx, prevC); // !!!: should this 0 be one of f*Limit?
1.5634 +
1.5635 + if (prevC == 0x0a &&
1.5636 + fp->fInputIdx > backSearchIndex &&
1.5637 + inputBuf[fp->fInputIdx-1] == 0x0d) {
1.5638 + int32_t prevOp = (int32_t)pat[fp->fPatIdx-2];
1.5639 + if (URX_TYPE(prevOp) == URX_LOOP_DOT_I) {
1.5640 + // .*, stepping back over CRLF pair.
1.5641 + U16_BACK_1(inputBuf, 0, fp->fInputIdx);
1.5642 + }
1.5643 + }
1.5644 +
1.5645 +
1.5646 + fp = StateSave(fp, fp->fPatIdx-1, status);
1.5647 + }
1.5648 + break;
1.5649 +
1.5650 +
1.5651 +
1.5652 + default:
1.5653 + // Trouble. The compiled pattern contains an entry with an
1.5654 + // unrecognized type tag.
1.5655 + U_ASSERT(FALSE);
1.5656 + }
1.5657 +
1.5658 + if (U_FAILURE(status)) {
1.5659 + isMatch = FALSE;
1.5660 + break;
1.5661 + }
1.5662 + }
1.5663 +
1.5664 +breakFromLoop:
1.5665 + fMatch = isMatch;
1.5666 + if (isMatch) {
1.5667 + fLastMatchEnd = fMatchEnd;
1.5668 + fMatchStart = startIdx;
1.5669 + fMatchEnd = fp->fInputIdx;
1.5670 + if (fTraceDebug) {
1.5671 + REGEX_RUN_DEBUG_PRINTF(("Match. start=%ld end=%ld\n\n", fMatchStart, fMatchEnd));
1.5672 + }
1.5673 + }
1.5674 + else
1.5675 + {
1.5676 + if (fTraceDebug) {
1.5677 + REGEX_RUN_DEBUG_PRINTF(("No match\n\n"));
1.5678 + }
1.5679 + }
1.5680 +
1.5681 + fFrame = fp; // The active stack frame when the engine stopped.
1.5682 + // Contains the capture group results that we need to
1.5683 + // access later.
1.5684 +
1.5685 + return;
1.5686 +}
1.5687 +
1.5688 +
1.5689 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(RegexMatcher)
1.5690 +
1.5691 +U_NAMESPACE_END
1.5692 +
1.5693 +#endif // !UCONFIG_NO_REGULAR_EXPRESSIONS
