1b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// file: regexcmp.cpp 3b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 4fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius// Copyright (C) 2002-2014 International Business Machines Corporation and others. 5b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// All Rights Reserved. 6b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 7b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// This file contains the ICU regular expression compiler, which is responsible 8b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// for processing a regular expression pattern into the compiled form that 9b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// is used by the match finding engine. 10b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 11b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 12b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/utypes.h" 13b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 14b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#if !UCONFIG_NO_REGULAR_EXPRESSIONS 15b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho#include "unicode/ustring.h" 17b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/unistr.h" 18b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/uniset.h" 19b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/uchar.h" 20b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/uchriter.h" 21b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/parsepos.h" 22b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/parseerr.h" 23b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "unicode/regex.h" 24103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius#include "unicode/utf.h" 25103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius#include "unicode/utf16.h" 26b26ce3a7367e4ed2ee7ddddcdc3f3d3377a455c2claireho#include "patternprops.h" 2750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho#include "putilimp.h" 28b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "cmemory.h" 29b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "cstring.h" 30b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "uvectr32.h" 3150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho#include "uvectr64.h" 32b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "uassert.h" 33b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "ucln_in.h" 34c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru#include "uinvchar.h" 35b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 36b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "regeximp.h" 37b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "regexcst.h" // Contains state table for the regex pattern parser. 38b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // generated by a Perl script. 39b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "regexcmp.h" 40b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#include "regexst.h" 4150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho#include "regextxt.h" 42b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 43b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 44b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 45b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruU_NAMESPACE_BEGIN 46b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 47b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 48b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 49b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 50b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Constructor. 51b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 52b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 53c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste QueruRegexCompile::RegexCompile(RegexPattern *rxp, UErrorCode &status) : 54c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fParenStack(status), fSetStack(status), fSetOpStack(status) 55b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru{ 5650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // Lazy init of all shared global sets (needed for init()'s empty text) 5750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho RegexStaticSets::initGlobals(&status); 5850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho 59b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStatus = &status; 60b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 61b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat = rxp; 62b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fScanIndex = 0; 63807b6b36605a2970f69dc767fee84a1b2a31e5e3Elliott Hughes fLastChar = -1; 64b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fPeekChar = -1; 65b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fLineNum = 1; 66b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fCharNum = 0; 67b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fQuoteMode = FALSE; 68b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fInBackslashQuote = FALSE; 69b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fModeFlags = fRXPat->fFlags | 0x80000000; 70b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fEOLComments = TRUE; 71b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 72b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fMatchOpenParen = -1; 73b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fMatchCloseParen = -1; 74b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 75b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_SUCCESS(status) && U_FAILURE(rxp->fDeferredStatus)) { 76b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru status = rxp->fDeferredStatus; 77b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 78b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 79b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 80c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chAmp = 0x26; // '&' 81c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chDash = 0x2d; // '-' 82b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 83b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 84b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 85b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 86b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Destructor 87b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 88b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 89b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruRegexCompile::~RegexCompile() { 90b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 91b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 92c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic inline void addCategory(UnicodeSet *set, int32_t value, UErrorCode& ec) { 93c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->addAll(UnicodeSet().applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, value, ec)); 94c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru} 95c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 96b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 97b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 98b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Compile regex pattern. The state machine for rexexp pattern parsing is here. 99b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The state tables are hand-written in the file regexcst.txt, 100b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// and converted to the form used here by a perl 101b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// script regexcst.pl 102b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 103b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 104b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::compile( 105b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru const UnicodeString &pat, // Source pat to be compiled. 106b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UParseError &pp, // Error position info 107b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UErrorCode &e) // Error Code 108b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru{ 109103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fPatternString = new UnicodeString(pat); 11050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UText patternText = UTEXT_INITIALIZER; 11150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho utext_openConstUnicodeString(&patternText, fRXPat->fPatternString, &e); 112fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 11350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (U_SUCCESS(e)) { 11450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho compile(&patternText, pp, e); 11550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho utext_close(&patternText); 11650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } 11750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho} 11850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho 11950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// 12050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// compile, UText mode 12150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// All the work is actually done here. 12250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// 12350294ead5e5d23f5bbfed76e00e6b510bd41eee1clairehovoid RegexCompile::compile( 12450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UText *pat, // Source pat to be compiled. 12550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UParseError &pp, // Error position info 12650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UErrorCode &e) // Error Code 12750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho{ 128b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStatus = &e; 129b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParseErr = &pp; 130b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStackPtr = 0; 131b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStack[fStackPtr] = 0; 132b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 133b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 134b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 135b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 136b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 137b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // There should be no pattern stuff in the RegexPattern object. They can not be reused. 13850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho U_ASSERT(fRXPat->fPattern == NULL || utext_nativeLength(fRXPat->fPattern) == 0); 139b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 140b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Prepare the RegexPattern object to receive the compiled pattern. 14150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fRXPat->fPattern = utext_clone(fRXPat->fPattern, pat, FALSE, TRUE, fStatus); 142b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStaticSets = RegexStaticSets::gStaticSets->fPropSets; 143b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStaticSets8 = RegexStaticSets::gStaticSets->fPropSets8; 144b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 145b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 146b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Initialize the pattern scanning state machine 14750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fPatternLength = utext_nativeLength(pat); 148b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru uint16_t state = 1; 149b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru const RegexTableEl *tableEl; 150103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 151103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // UREGEX_LITERAL force entire pattern to be treated as a literal string. 152103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (fModeFlags & UREGEX_LITERAL) { 153103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fQuoteMode = TRUE; 154103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 155103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 156b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru nextChar(fC); // Fetch the first char from the pattern string. 157b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 158b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 159b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Main loop for the regex pattern parsing state machine. 160b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Runs once per state transition. 161b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Each time through optionally performs, depending on the state table, 162b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - an advance to the the next pattern char 163b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - an action to be performed. 164b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - pushing or popping a state to/from the local state return stack. 165b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // file regexcst.txt is the source for the state table. The logic behind 166b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // recongizing the pattern syntax is there, not here. 167b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 168b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 169b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Bail out if anything has gone wrong. 170b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Regex pattern parsing stops on the first error encountered. 171b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 172b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 173b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 174b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 175b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(state != 0); 176b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 177b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Find the state table element that matches the input char from the pattern, or the 178b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // class of the input character. Start with the first table row for this 179b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // state, then linearly scan forward until we find a row that matches the 180b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // character. The last row for each state always matches all characters, so 181b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the search will stop there, if not before. 182b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 183b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru tableEl = &gRuleParseStateTable[state]; 184c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru REGEX_SCAN_DEBUG_PRINTF(("char, line, col = (\'%c\', %d, %d) state=%s ", 185b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fC.fChar, fLineNum, fCharNum, RegexStateNames[state])); 186b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 187b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { // loop through table rows belonging to this state, looking for one 188b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // that matches the current input char. 189b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru REGEX_SCAN_DEBUG_PRINTF((".")); 190b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fCharClass < 127 && fC.fQuoted == FALSE && tableEl->fCharClass == fC.fChar) { 191b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Table row specified an individual character, not a set, and 192b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the input character is not quoted, and 193b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the input character matched it. 194b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 195b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 196b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fCharClass == 255) { 197b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Table row specified default, match anything character class. 198b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 199b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 200b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fCharClass == 254 && fC.fQuoted) { 201b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Table row specified "quoted" and the char was quoted. 202b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 203b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 204b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fCharClass == 253 && fC.fChar == (UChar32)-1) { 205b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Table row specified eof and we hit eof on the input. 206b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 207b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 208b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 209b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 && // Table specs a char class && 210b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fC.fQuoted == FALSE && // char is not escaped && 211b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fC.fChar != (UChar32)-1) { // char is not EOF 212103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius U_ASSERT(tableEl->fCharClass <= 137); 213c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (RegexStaticSets::gStaticSets->fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) { 214b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Table row specified a character class, or set of characters, 215b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // and the current char matches it. 216b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 217b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 218b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 219b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 220b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // No match on this row, advance to the next row for this state, 221b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru tableEl++; 222b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 223b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru REGEX_SCAN_DEBUG_PRINTF(("\n")); 224b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 225b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 226b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We've found the row of the state table that matches the current input 227b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // character from the rules string. 228b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Perform any action specified by this row in the state table. 229b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (doParseActions(tableEl->fAction) == FALSE) { 230b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Break out of the state machine loop if the 231b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the action signalled some kind of error, or 232b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the action was to exit, occurs on normal end-of-rules-input. 233b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 234b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 235b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 236b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fPushState != 0) { 237b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStackPtr++; 238b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fStackPtr >= kStackSize) { 239b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_INTERNAL_ERROR); 240b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru REGEX_SCAN_DEBUG_PRINTF(("RegexCompile::parse() - state stack overflow.\n")); 241b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStackPtr--; 242b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 243b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStack[fStackPtr] = tableEl->fPushState; 244b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 245b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 246b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 247b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // NextChar. This is where characters are actually fetched from the pattern. 248b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Happens under control of the 'n' tag in the state table. 249b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 250b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fNextChar) { 251b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru nextChar(fC); 252b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 253b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 254b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Get the next state from the table entry, or from the 255b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // state stack if the next state was specified as "pop". 256b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (tableEl->fNextState != 255) { 257b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru state = tableEl->fNextState; 258b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 259b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru state = fStack[fStackPtr]; 260b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStackPtr--; 261b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fStackPtr < 0) { 262b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // state stack underflow 263b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // This will occur if the user pattern has mis-matched parentheses, 264b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // with extra close parens. 265c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 266b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fStackPtr++; 267b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_MISMATCHED_PAREN); 268b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 269b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 270b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 271b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 272b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 273c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 274c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Bail out if the pattern had errors. 275c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Set stack cleanup: a successful compile would have left it empty, 276c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // but errors can leave temporary sets hanging around. 277c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru while (!fSetStack.empty()) { 278c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete (UnicodeSet *)fSetStack.pop(); 279c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 280c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return; 281c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 282c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 283b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 284b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The pattern has now been read and processed, and the compiled code generated. 285b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 286b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 287b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 288b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compute the number of digits requried for the largest capture group number. 289b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 290b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fMaxCaptureDigits = 1; 291b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t n = 10; 29250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t groupCount = fRXPat->fGroupMap->size(); 29350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho while (n <= groupCount) { 294b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fMaxCaptureDigits++; 295b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru n *= 10; 296b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 297b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 298b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 299b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The pattern's fFrameSize so far has accumulated the requirements for 300b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // storage for capture parentheses, counters, etc. that are encountered 301b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // in the pattern. Add space for the two variables that are always 30250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // present in the saved state: the input string position (int64_t) and 30350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // the position in the compiled pattern. 30450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // 30550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fRXPat->fFrameSize+=RESTACKFRAME_HDRCOUNT; 30650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho 307b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 30850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // Optimization pass 1: NOPs, back-references, and case-folding 30950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // 31050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho stripNOPs(); 311b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 312b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 313b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Get bounds for the minimum and maximum length of a string that this 314b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // pattern can match. Used to avoid looking for matches in strings that 315b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // are too short. 316b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 317b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fMinMatchLen = minMatchLength(3, fRXPat->fCompiledPat->size()-1); 318b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 319b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 32050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // Optimization pass 2: match start type 321c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 322c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru matchStartType(); 323b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 324b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 325b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Set up fast latin-1 range sets 326b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 327b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t numSets = fRXPat->fSets->size(); 328b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fSets8 = new Regex8BitSet[numSets]; 329c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Null pointer check. 330c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fRXPat->fSets8 == NULL) { 331c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru e = *fStatus = U_MEMORY_ALLOCATION_ERROR; 332c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return; 333c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 334b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t i; 335b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (i=0; i<numSets; i++) { 336b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(i); 337b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fSets8[i].init(s); 338b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 339b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 340b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 341b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 342b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 343b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 344b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 345b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 346b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 347b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 348b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// doParseAction Do some action during regex pattern parsing. 349b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Called by the parse state machine. 350b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 351b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Generation of the match engine PCode happens here, or 352b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// in functions called from the parse actions defined here. 353b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 354b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 355b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 356b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruUBool RegexCompile::doParseActions(int32_t action) 357b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru{ 358b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UBool returnVal = TRUE; 359b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 360b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch ((Regex_PatternParseAction)action) { 361b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 362b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPatStart: 363b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Start of pattern compiles to: 364b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru //0 SAVE 2 Fall back to position of FAIL 365b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru //1 jmp 3 366b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru //2 FAIL Stop if we ever reach here. 367b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru //3 NOP Dummy, so start of pattern looks the same as 368b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the start of an ( grouping. 369b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru //4 NOP Resreved, will be replaced by a save if there are 370b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // OR | operators at the top level 371b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_STATE_SAVE, 2), *fStatus); 372b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_JMP, 3), *fStatus); 373b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_FAIL, 0), *fStatus); 374b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 375b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Standard open nonCapture paren action emits the two NOPs and 376b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // sets up the paren stack frame. 377b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru doParseActions(doOpenNonCaptureParen); 378b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 379b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 380b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPatFinish: 381b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We've scanned to the end of the pattern 382b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The end of pattern compiles to: 383b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // URX_END 384b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // which will stop the runtime match engine. 385b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Encountering end of pattern also behaves like a close paren, 386b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // and forces fixups of the State Save at the beginning of the compiled pattern 387b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // and of any OR operations at the top level. 388b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 389b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru handleCloseParen(); 390b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fParenStack.size() > 0) { 391b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Missing close paren in pattern. 392b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_MISMATCHED_PAREN); 393b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 394b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 395b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // add the END operation to the compiled pattern. 396b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_END, 0), *fStatus); 397b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 398b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Terminate the pattern compilation state machine. 399b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru returnVal = FALSE; 400b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 401b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 402b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 403b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 404b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOrOperator: 405b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Scanning a '|', as in (A|B) 406b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 407103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Generate code for any pending literals preceding the '|' 408103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 409103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 410b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert a SAVE operation at the start of the pattern section preceding 411b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // this OR at this level. This SAVE will branch the match forward 412b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // to the right hand side of the OR in the event that the left hand 413b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // side fails to match and backtracks. Locate the position for the 414b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // save from the location on the top of the parentheses stack. 415b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t savePosition = fParenStack.popi(); 41650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(savePosition); 417b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(op) == URX_NOP); // original contents of reserved location 418b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+1); 419b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, savePosition); 420b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 421b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Append an JMP operation into the compiled pattern. The operand for 422b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the JMP will eventually be the location following the ')' for the 423b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // group. This will be patched in later, when the ')' is encountered. 424b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_JMP, 0); 425b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 426b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 427b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Push the position of the newly added JMP op onto the parentheses stack. 428b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // This registers if for fixup when this block's close paren is encountered. 429b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); 430b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 431b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Append a NOP to the compiled pattern. This is the slot reserved 432b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // for a SAVE in the event that there is yet another '|' following 433b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // this one. 434b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 435b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); 436b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 437b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 438b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 439b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 440b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenCaptureParen: 441b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Open Paren. 442b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compile to a 443b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which later may be replaced by a save-state if the 444b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // parenthesized group gets a * quantifier, followed by 445b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - START_CAPTURE n where n is stack frame offset to the capture group variables. 446b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which may later be replaced by a save-state if there 447b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // is an '|' alternation within the parens. 448b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 449b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Each capture group gets three slots in the save stack frame: 45050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // 0: Capture Group start position (in input string being matched.) 45150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // 1: Capture Group end position. 45250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // 2: Start of Match-in-progress. 453b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The first two locations are for a completed capture group, and are 454b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // referred to by back references and the like. 455b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The third location stores the capture start position when an START_CAPTURE is 456b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // encountered. This will be promoted to a completed capture when (and if) the corresponding 45750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // END_CAPTURE is encountered. 458b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 459103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 460b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 461b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t varsLoc = fRXPat->fFrameSize; // Reserve three slots in match stack frame. 462b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize += 3; 463b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t cop = URX_BUILD(URX_START_CAPTURE, varsLoc); 464b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(cop, *fStatus); 465b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 466b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 467b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 468b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the two NOPs. Depending on what follows in the pattern, the 469b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // NOPs may be changed to SAVE_STATE or JMP ops, with a target 470b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // address of the end of the parenthesized group. 471b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 472b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(capturing, *fStatus); // Frame type. 473b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus); // The first NOP location 474b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP loc 475b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 476b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Save the mapping from group number to stack frame variable position. 477b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fGroupMap->addElement(varsLoc, *fStatus); 478b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 479b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 480b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 481b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenNonCaptureParen: 482b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Open non-caputuring (grouping only) Paren. 483b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compile to a 484b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which later may be replaced by a save-state if the 485b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // parenthesized group gets a * quantifier, followed by 486b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which may later be replaced by a save-state if there 487b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // is an '|' alternation within the parens. 488b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 489103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 490b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 491b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 492b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 493b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 494b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the two NOPs. 495b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 496b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(plain, *fStatus); // Begin a new frame. 497b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 498b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP loc 499b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 500b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 501b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 502b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 503b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenAtomicParen: 504b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Open Atomic Paren. (?> 505b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compile to a 506c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // - NOP, which later may be replaced if the parenthesized group 507b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // has a quantifier, followed by 508b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - STO_SP save state stack position, so it can be restored at the ")" 509b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which may later be replaced by a save-state if there 510b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // is an '|' alternation within the parens. 511b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 512103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 513b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 514b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t varLoc = fRXPat->fDataSize; // Reserve a data location for saving the 515b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fDataSize += 1; // state stack ptr. 516b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stoOp = URX_BUILD(URX_STO_SP, varLoc); 517b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(stoOp, *fStatus); 518b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 519b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 520b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 521b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the two NOPs. Depending on what follows in the pattern, the 522b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // NOPs may be changed to SAVE_STATE or JMP ops, with a target 523b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // address of the end of the parenthesized group. 524b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 525b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(atomic, *fStatus); // Frame type. 526b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-3, *fStatus); // The first NOP 527b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP 528b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 529b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 530b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 531b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 532b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenLookAhead: 533b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Positive Look-ahead (?= stuff ) 534c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 535c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Note: Addition of transparent input regions, with the need to 536c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // restore the original regions when failing out of a lookahead 537c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // block, complicated this sequence. Some conbined opcodes 538c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // might make sense - or might not, lookahead aren't that common. 539c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 540c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Caution: min match length optimization knows about this 541c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // sequence; don't change without making updates there too. 542c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 543b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 544c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 1 START_LA dataLoc Saves SP, Input Pos 545c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 2. STATE_SAVE 4 on failure of lookahead, goto 4 546c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 3 JMP 6 continue ... 547c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 548c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4. LA_END Look Ahead failed. Restore regions. 549c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 5. BACKTRACK and back track again. 550c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 551c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 6. NOP reserved for use by quantifiers on the block. 552b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Look-ahead can't have quantifiers, but paren stack 553b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // compile time conventions require the slot anyhow. 554c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 7. NOP may be replaced if there is are '|' ops in the block. 555c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 8. code for parenthesized stuff. 556c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 9. LA_END 557c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 558b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Two data slots are reserved, for saving the stack ptr and the input position. 559b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 560103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 561b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = fRXPat->fDataSize; 562c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fDataSize += 2; 563b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LA_START, dataLoc); 564b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 565b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 566c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+ 2); 567c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 568c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 569c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_BUILD(URX_JMP, fRXPat->fCompiledPat->size()+ 3); 570c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 571fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 572c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_BUILD(URX_LA_END, dataLoc); 573c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 574c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 575c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_BUILD(URX_BACKTRACK, 0); 576c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 577fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 578b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_NOP, 0); 579b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 580b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 581b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 582b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 583c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // of the NOPs. 584b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 585b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(lookAhead, *fStatus); // Frame type. 586b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 587b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP location 588b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 589b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 590b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 591b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenLookAheadNeg: 592b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Negated Lookahead. (?! stuff ) 593b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 594b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. START_LA dataloc 595b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. SAVE_STATE 7 // Fail within look-ahead block restores to this state, 596b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // // which continues with the match. 597b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. NOP // Std. Open Paren sequence, for possible '|' 598b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. code for parenthesized stuff. 599b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5. END_LA // Cut back stack, remove saved state from step 2. 600c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 6. BACKTRACK // code in block succeeded, so neg. lookahead fails. 601c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 7. END_LA // Restore match region, in case look-ahead was using 602c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // an alternate (transparent) region. 603b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 604103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 605b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = fRXPat->fDataSize; 606c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fDataSize += 2; 607b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LA_START, dataLoc); 608b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 609b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 610b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_STATE_SAVE, 0); // dest address will be patched later. 611b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 612b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 613b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_NOP, 0); 614b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 615b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 616b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 617c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // of the StateSave and NOP. 618b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 619c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fParenStack.push(negLookAhead, *fStatus); // Frame type 620b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The STATE_SAVE location 621b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP location 622c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 62350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // Instructions #5 - #7 will be added when the ')' is encountered. 624b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 625b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 626b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 627b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenLookBehind: 628b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 629b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compile a (?<= look-behind open paren. 630b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 631b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 632b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 0 URX_LB_START dataLoc 633b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1 URX_LB_CONT dataLoc 634b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2 MinMatchLen 635b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3 MaxMatchLen 636b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4 URX_NOP Standard '(' boilerplate. 637b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5 URX_NOP Reserved slot for use with '|' ops within (block). 638b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 6 <code for LookBehind expression> 639b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 7 URX_LB_END dataLoc # Check match len, restore input len 640b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 8 URX_LA_END dataLoc # Restore stack, input pos 641b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 642b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Allocate a block of matcher data, to contain (when running a match) 643b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 0: Stack ptr on entry 644b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1: Input Index on entry 645b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2: Start index of match current match attempt. 646c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 3: Original Input String len. 647b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 648103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Generate match code for any pending literals. 649103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 650103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 651b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Allocate data space 652b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = fRXPat->fDataSize; 653c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fDataSize += 4; 654c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 655b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit URX_LB_START 656b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LB_START, dataLoc); 657b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 658c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 659b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit URX_LB_CONT 660b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LB_CONT, dataLoc); 661b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 662b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(0, *fStatus); // MinMatchLength. To be filled later. 663b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(0, *fStatus); // MaxMatchLength. To be filled later. 664c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 665b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the NOP 666b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_NOP, 0); 667b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 668b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 669c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 670b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 671c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // of the URX_LB_CONT and the NOP. 672b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 673b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(lookBehind, *fStatus); // Frame type 674b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 675b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The 2nd NOP location 676c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 677b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The final two instructions will be added when the ')' is encountered. 678b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 679b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 680b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 681b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 682b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpenLookBehindNeg: 683b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 684b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compile a (?<! negated look-behind open paren. 685b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 686b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 687b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 0 URX_LB_START dataLoc # Save entry stack, input len 688b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1 URX_LBN_CONT dataLoc # Iterate possible match positions 689b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2 MinMatchLen 690b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3 MaxMatchLen 691b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4 continueLoc (9) 692b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5 URX_NOP Standard '(' boilerplate. 693b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 6 URX_NOP Reserved slot for use with '|' ops within (block). 694b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 7 <code for LookBehind expression> 695b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 8 URX_LBN_END dataLoc # Check match len, cause a FAIL 696b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 9 ... 697b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 698b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Allocate a block of matcher data, to contain (when running a match) 699b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 0: Stack ptr on entry 700b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1: Input Index on entry 701b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2: Start index of match current match attempt. 702c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 3: Original Input String len. 703b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 704103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Generate match code for any pending literals. 705103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 706103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 707b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Allocate data space 708b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = fRXPat->fDataSize; 709c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fDataSize += 4; 710c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 711b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit URX_LB_START 712b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LB_START, dataLoc); 713b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 714c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 715b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit URX_LBN_CONT 716b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LBN_CONT, dataLoc); 717b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 718b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(0, *fStatus); // MinMatchLength. To be filled later. 719b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(0, *fStatus); // MaxMatchLength. To be filled later. 720b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(0, *fStatus); // Continue Loc. To be filled later. 721c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 722b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the NOP 723b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_NOP, 0); 724b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 725b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 726c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 727b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 728c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // of the URX_LB_CONT and the NOP. 729b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); // Match mode state 730b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(lookBehindN, *fStatus); // Frame type 731b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP location 732b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The 2nd NOP location 733c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 734b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The final two instructions will be added when the ')' is encountered. 735b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 736b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 737b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 738b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doConditionalExpr: 739b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Conditionals such as (?(1)a:b) 740b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPerlInline: 741b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Perl inline-condtionals. (?{perl code}a|b) We're not perl, no way to do them. 742b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_UNIMPLEMENTED); 743b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 744b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 745b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 746b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doCloseParen: 747b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru handleCloseParen(); 748b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fParenStack.size() <= 0) { 749b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Extra close paren, or missing open paren. 750b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_MISMATCHED_PAREN); 751b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 752b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 753b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 754b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doNOP: 755b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 756b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 757b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 758b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBadOpenParenType: 759b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doRuleError: 760b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_RULE_SYNTAX); 761b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 762b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 763b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 764b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doMismatchedParenErr: 765b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_MISMATCHED_PAREN); 766b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 767b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 768b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPlus: 769b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Normal '+' compiles to 770b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. stuff to be repeated (already built) 771b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. jmp-sav 1 772b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. ... 773b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 774b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Or, if the item to be repeated can match a zero length string, 775b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. STO_INP_LOC data-loc 776b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. body of stuff to be repeated 777b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. JMP_SAV_X 2 778b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. ... 779b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 780b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 781b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Or, if the item to be repeated is simple 782b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. Item to be repeated. 783b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. LOOP_SR_I set number (assuming repeated item is a set ref) 784b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. LOOP_C stack location 785b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 786b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(FALSE); // location of item #1 787b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t frameLoc; 788b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 789b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Check for simple constructs, which may get special optimized code. 790b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (topLoc == fRXPat->fCompiledPat->size() - 1) { 79150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc); 792b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 793b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(repeatedOp) == URX_SETREF) { 794b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit optimized code for [char set]+ 795b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpI = URX_BUILD(URX_LOOP_SR_I, URX_VAL(repeatedOp)); 796b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(loopOpI, *fStatus); 797b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru frameLoc = fRXPat->fFrameSize; 798b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 799b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpC = URX_BUILD(URX_LOOP_C, frameLoc); 800b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 801b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 802b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 803b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 804b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(repeatedOp) == URX_DOTANY || 805c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru URX_TYPE(repeatedOp) == URX_DOTANY_ALL || 806c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) { 807b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit Optimized code for .+ operations. 808b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpI = URX_BUILD(URX_LOOP_DOT_I, 0); 809b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) { 810c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // URX_LOOP_DOT_I operand is a flag indicating ". matches any" mode. 811b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loopOpI |= 1; 812b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 813c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fModeFlags & UREGEX_UNIX_LINES) { 814c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru loopOpI |= 2; 815c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 816b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(loopOpI, *fStatus); 817b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru frameLoc = fRXPat->fFrameSize; 818b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 819b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpC = URX_BUILD(URX_LOOP_C, frameLoc); 820b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 821b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 822b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 823b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 824b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 825b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 826b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // General case. 827b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 828b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Check for minimum match length of zero, which requires 829b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // extra loop-breaking code. 830b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (minMatchLength(topLoc, fRXPat->fCompiledPat->size()-1) == 0) { 831b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Zero length match is possible. 832b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the code sequence that can handle it. 833b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topLoc); 834b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru frameLoc = fRXPat->fFrameSize; 835b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 836b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 837b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_STO_INP_LOC, frameLoc); 838b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc); 839b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 840b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_JMP_SAV_X, topLoc+1); 841b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 842b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 843b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Simpler code when the repeated body must match something non-empty 844b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpOp = URX_BUILD(URX_JMP_SAV, topLoc); 845b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(jmpOp, *fStatus); 846b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 847b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 848b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 849b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 850b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doNGPlus: 851b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Non-greedy '+?' compiles to 852b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. stuff to be repeated (already built) 853b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. state-save 1 854b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. ... 855b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 856b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(FALSE); 857b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, topLoc); 858b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(saveStateOp, *fStatus); 859b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 860b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 861b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 862b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 863b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doOpt: 864b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Normal (greedy) ? quantifier. 865b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 866b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. state save 3 867b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. body of optional block 868b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. ... 869b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert the state save into the compiled pattern, and we're done. 870b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 871b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveStateLoc = blockTopLoc(TRUE); 872b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()); 873b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc); 874b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 875b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 876b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 877b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doNGOpt: 878b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Non-greedy ?? quantifier 879b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // compiles to 880b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. jmp 4 881b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. body of optional block 882b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3 jmp 5 883b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. state save 2 884b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5 ... 885b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // This code is less than ideal, with two jmps instead of one, because we can only 886b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // insert one instruction at the top of the block being iterated. 887b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 888b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmp1_loc = blockTopLoc(TRUE); 889b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmp2_loc = fRXPat->fCompiledPat->size(); 890b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 891b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmp1_op = URX_BUILD(URX_JMP, jmp2_loc+1); 892b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(jmp1_op, jmp1_loc); 893b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 894b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmp2_op = URX_BUILD(URX_JMP, jmp2_loc+2); 895b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(jmp2_op, *fStatus); 896b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 897b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t save_op = URX_BUILD(URX_STATE_SAVE, jmp1_loc+1); 898b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(save_op, *fStatus); 899b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 900b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 901b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 902b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 903b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doStar: 904b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Normal (greedy) * quantifier. 905b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 906b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. STATE_SAVE 4 907b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. body of stuff being iterated over 908b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. JMP_SAV 2 909b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. ... 910b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 911b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Or, if the body is a simple [Set], 912b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. LOOP_SR_I set number 913b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. LOOP_C stack location 914b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // ... 915b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 916c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Or if this is a .* 917b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. LOOP_DOT_I (. matches all mode flag) 918b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. LOOP_C stack location 919b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 920b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Or, if the body can match a zero-length string, to inhibit infinite loops, 921b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. STATE_SAVE 5 922b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. STO_INP_LOC data-loc 923b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. body of stuff 924b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. JMP_SAV_X 2 925b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5. ... 926b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 927b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // location of item #1, the STATE_SAVE 928b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(FALSE); 929b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = -1; 930b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 931b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Check for simple *, where the construct being repeated 932b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // compiled to single opcode, and might be optimizable. 933b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (topLoc == fRXPat->fCompiledPat->size() - 1) { 93450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t repeatedOp = (int32_t)fRXPat->fCompiledPat->elementAti(topLoc); 935b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 936b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(repeatedOp) == URX_SETREF) { 937c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Emit optimized code for a [char set]* 938b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpI = URX_BUILD(URX_LOOP_SR_I, URX_VAL(repeatedOp)); 939b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc); 940b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru dataLoc = fRXPat->fFrameSize; 941b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 942b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpC = URX_BUILD(URX_LOOP_C, dataLoc); 943b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 944b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 945b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 946b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 947b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(repeatedOp) == URX_DOTANY || 948c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru URX_TYPE(repeatedOp) == URX_DOTANY_ALL || 949c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru URX_TYPE(repeatedOp) == URX_DOTANY_UNIX) { 950b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit Optimized code for .* operations. 951b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpI = URX_BUILD(URX_LOOP_DOT_I, 0); 952b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(repeatedOp) == URX_DOTANY_ALL) { 953b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // URX_LOOP_DOT_I operand is a flag indicating . matches any mode. 954b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loopOpI |= 1; 955b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 956c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_UNIX_LINES) != 0) { 957c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru loopOpI |= 2; 958c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 959b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(loopOpI, topLoc); 960b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru dataLoc = fRXPat->fFrameSize; 961b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 962b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopOpC = URX_BUILD(URX_LOOP_C, dataLoc); 963b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(loopOpC, *fStatus); 964b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 965b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 966b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 967b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 968b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit general case code for this * 969b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The optimizations did not apply. 970b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 971b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveStateLoc = blockTopLoc(TRUE); 972b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpOp = URX_BUILD(URX_JMP_SAV, saveStateLoc+1); 973b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 974b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Check for minimum match length of zero, which requires 975b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // extra loop-breaking code. 976b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (minMatchLength(saveStateLoc, fRXPat->fCompiledPat->size()-1) == 0) { 977b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(saveStateLoc); 978b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru dataLoc = fRXPat->fFrameSize; 979b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 980b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 981b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_STO_INP_LOC, dataLoc); 982b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, saveStateLoc+1); 983b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru jmpOp = URX_BUILD(URX_JMP_SAV_X, saveStateLoc+2); 984b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 985c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 986b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Locate the position in the compiled pattern where the match will continue 987b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // after completing the *. (4 or 5 in the comment above) 988b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t continueLoc = fRXPat->fCompiledPat->size()+1; 989b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 990b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Put together the save state op store it into the compiled code. 991b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveStateOp = URX_BUILD(URX_STATE_SAVE, continueLoc); 992b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(saveStateOp, saveStateLoc); 993b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 994b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Append the URX_JMP_SAV or URX_JMPX operation to the compiled pattern. 995b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(jmpOp, *fStatus); 996b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 997b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 998b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 999b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doNGStar: 1000b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Non-greedy *? quantifier 1001b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // compiles to 1002b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. JMP 3 1003b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. body of stuff being iterated over 1004b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. STATE_SAVE 2 1005b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4 ... 1006b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1007b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpLoc = blockTopLoc(TRUE); // loc 1. 1008b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveLoc = fRXPat->fCompiledPat->size(); // loc 3. 1009b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpOp = URX_BUILD(URX_JMP, saveLoc); 1010b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stateSaveOp = URX_BUILD(URX_STATE_SAVE, jmpLoc+1); 1011b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(jmpOp, jmpLoc); 1012b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(stateSaveOp, *fStatus); 1013b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1014b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1015b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1016b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1017b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doIntervalInit: 1018b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The '{' opening an interval quantifier was just scanned. 1019b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Init the counter varaiables that will accumulate the values as the digits 1020b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // are scanned. 1021b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fIntervalLow = 0; 1022b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fIntervalUpper = -1; 1023b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1024b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1025b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doIntevalLowerDigit: 1026b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Scanned a digit from the lower value of an {lower,upper} interval 1027b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1028b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t digitValue = u_charDigitValue(fC.fChar); 1029b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(digitValue >= 0); 1030b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fIntervalLow = fIntervalLow*10 + digitValue; 1031b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalLow < 0) { 1032b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_NUMBER_TOO_BIG); 1033b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1034b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1035b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1036b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1037b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doIntervalUpperDigit: 1038b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Scanned a digit from the upper value of an {lower,upper} interval 1039b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1040b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalUpper < 0) { 1041b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fIntervalUpper = 0; 1042b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1043b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t digitValue = u_charDigitValue(fC.fChar); 1044b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(digitValue >= 0); 1045b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fIntervalUpper = fIntervalUpper*10 + digitValue; 1046b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalUpper < 0) { 1047b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_NUMBER_TOO_BIG); 1048b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1049b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1050b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1051b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1052b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doIntervalSame: 1053b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Scanned a single value interval like {27}. Upper = Lower. 1054b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fIntervalUpper = fIntervalLow; 1055b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1056b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1057b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doInterval: 1058b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Finished scanning a normal {lower,upper} interval. Generate the code for it. 1059b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (compileInlineInterval() == FALSE) { 1060b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru compileInterval(URX_CTR_INIT, URX_CTR_LOOP); 1061b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1062b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1063b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1064b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPossessiveInterval: 1065b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Finished scanning a Possessive {lower,upper}+ interval. Generate the code for it. 1066b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1067b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Remember the loc for the top of the block being looped over. 1068c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // (Can not reserve a slot in the compiled pattern at this time, because 1069c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // compileInterval needs to reserve also, and blockTopLoc can only reserve 1070b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // once per block.) 1071b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(FALSE); 1072b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1073b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Produce normal looping code. 1074b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru compileInterval(URX_CTR_INIT, URX_CTR_LOOP); 1075b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1076b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Surround the just-emitted normal looping code with a STO_SP ... LD_SP 1077b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // just as if the loop was inclosed in atomic parentheses. 1078b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1079b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // First the STO_SP before the start of the loop 1080b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topLoc); 1081b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t varLoc = fRXPat->fDataSize; // Reserve a data location for saving the 1082b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fDataSize += 1; // state stack ptr. 1083b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_STO_SP, varLoc); 1084b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc); 1085b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 108650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t loopOp = (int32_t)fRXPat->fCompiledPat->popi(); 1087b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(loopOp) == URX_CTR_LOOP && URX_VAL(loopOp) == topLoc); 1088b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loopOp++; // point LoopOp after the just-inserted STO_SP 1089b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->push(loopOp, *fStatus); 1090b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1091b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Then the LD_SP after the end of the loop 1092b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LD_SP, varLoc); 1093b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1094b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1095b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1096b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1097b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1098b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doNGInterval: 1099b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Finished scanning a non-greedy {lower,upper}? interval. Generate the code for it. 1100b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru compileInterval(URX_CTR_INIT_NG, URX_CTR_LOOP_NG); 1101b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1102b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1103b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doIntervalError: 1104b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_BAD_INTERVAL); 1105b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1106b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1107b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doLiteralChar: 1108c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We've just scanned a "normal" character from the pattern, 1109b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru literalChar(fC.fChar); 1110b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1111b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1112b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1113c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doEscapedLiteralChar: 1114c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We've just scanned an backslashed escaped character with no 1115c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // special meaning. It represents itself. 1116c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 && 1117c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) || // in [A-Z] 1118c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) { // in [a-z] 1119c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_BAD_ESCAPE_SEQUENCE); 1120c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1121c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru literalChar(fC.fChar); 1122c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1123c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1124b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1125b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doDotAny: 1126b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // scanned a ".", match any single character. 1127b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1128103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1129b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op; 1130b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fModeFlags & UREGEX_DOTALL) { 1131b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_DOTANY_ALL, 0); 1132c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if (fModeFlags & UREGEX_UNIX_LINES) { 1133c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_BUILD(URX_DOTANY_UNIX, 0); 1134b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 1135b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_DOTANY, 0); 1136b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1137b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1138b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1139b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1140b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1141c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doCaret: 1142b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1143103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1144c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t op = 0; 1145c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ( (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1146c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_CARET; 1147c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1148c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_CARET_M; 1149c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1150fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius op = URX_CARET; // Only testing true start of input. 1151c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1152c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_CARET_M_UNIX; 1153c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1154b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus); 1155b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1156b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1157b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1158c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doDollar: 1159b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1160103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1161c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t op = 0; 1162c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ( (fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1163c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_DOLLAR; 1164c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) == 0) { 1165c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_DOLLAR_M; 1166c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if ((fModeFlags & UREGEX_MULTILINE) == 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1167c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_DOLLAR_D; 1168c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else if ((fModeFlags & UREGEX_MULTILINE) != 0 && (fModeFlags & UREGEX_UNIX_LINES) != 0) { 1169c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_DOLLAR_MD; 1170c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1171b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus); 1172b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1173b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1174b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1175b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashA: 1176103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1177b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_CARET, 0), *fStatus); 1178b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1179b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1180b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashB: 1181b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1182b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru #if UCONFIG_NO_BREAK_ITERATION==1 1183b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fModeFlags & UREGEX_UWORD) { 1184b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_UNSUPPORTED_ERROR); 1185b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1186b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru #endif 1187103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1188b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B; 1189b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(op, 1), *fStatus); 1190b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1191b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1192b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1193b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashb: 1194b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1195b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru #if UCONFIG_NO_BREAK_ITERATION==1 1196b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fModeFlags & UREGEX_UWORD) { 1197b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_UNSUPPORTED_ERROR); 1198b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1199b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru #endif 1200103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1201b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = (fModeFlags & UREGEX_UWORD)? URX_BACKSLASH_BU : URX_BACKSLASH_B; 1202b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(op, 0), *fStatus); 1203b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1204b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1205b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1206b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashD: 1207103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1208b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_D, 1), *fStatus); 1209b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1210b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1211b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashd: 1212103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1213b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_D, 0), *fStatus); 1214b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1215b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1216b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashG: 1217103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1218b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_G, 0), *fStatus); 1219b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1220b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1221b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashS: 1222103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1223b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement( 1224b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru URX_BUILD(URX_STAT_SETREF_N, URX_ISSPACE_SET), *fStatus); 1225b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1226b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1227b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashs: 1228103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1229b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement( 1230b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru URX_BUILD(URX_STATIC_SETREF, URX_ISSPACE_SET), *fStatus); 1231b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1232b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1233b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashW: 1234103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1235b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement( 1236b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru URX_BUILD(URX_STAT_SETREF_N, URX_ISWORD_SET), *fStatus); 1237b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1238b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1239b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashw: 1240103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1241b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement( 1242b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru URX_BUILD(URX_STATIC_SETREF, URX_ISWORD_SET), *fStatus); 1243b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1244b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1245b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashX: 1246103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1247b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_X, 0), *fStatus); 1248b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1249b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1250b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1251b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashZ: 1252103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1253b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_DOLLAR, 0), *fStatus); 1254b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1255b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1256b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackslashz: 1257103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1258b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKSLASH_Z, 0), *fStatus); 1259b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1260b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1261b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doEscapeError: 1262b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_BAD_ESCAPE_SEQUENCE); 1263b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1264b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1265b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doExit: 1266103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1267b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru returnVal = FALSE; 1268b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1269b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1270b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doProperty: 1271b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1272103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1273b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UnicodeSet *theSet = scanProp(); 1274b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru compileSet(theSet); 1275b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1276b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1277b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1278c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doNamedChar: 1279b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1280c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UChar32 c = scanNamedChar(); 1281c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru literalChar(c); 1282b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1283b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1284fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 1285b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1286b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBackRef: 1287b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // BackReference. Somewhat unusual in that the front-end can not completely parse 1288b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the regular expression, because the number of digits to be consumed 1289b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // depends on the number of capture groups that have been defined. So 1290b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // we have to do it here instead. 1291b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1292b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t numCaptureGroups = fRXPat->fGroupMap->size(); 1293b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t groupNum = 0; 1294b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UChar32 c = fC.fChar; 1295b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1296b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 1297b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop once per digit, for max allowed number of digits in a back reference. 1298b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t digit = u_charDigitValue(c); 1299b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru groupNum = groupNum * 10 + digit; 1300b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (groupNum >= numCaptureGroups) { 1301b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1302b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1303b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c = peekCharLL(); 1304b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (RegexStaticSets::gStaticSets->fRuleDigitsAlias->contains(c) == FALSE) { 1305b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1306b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1307b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru nextCharLL(); 1308b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1309b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1310b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Scan of the back reference in the source regexp is complete. Now generate 1311c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // the compiled code for it. 1312b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Because capture groups can be forward-referenced by back-references, 1313b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // we fill the operand with the capture group number. At the end 1314b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of compilation, it will be changed to the variable's location. 1315103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius U_ASSERT(groupNum > 0); // Shouldn't happen. '\0' begins an octal escape sequence, 1316103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // and shouldn't enter this code path at all. 1317103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1318b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op; 1319b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 1320b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_BACKREF_I, groupNum); 1321b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 1322b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_BACKREF, groupNum); 1323b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1324b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1325b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1326b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1327b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1328b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1329b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPossessivePlus: 1330b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Possessive ++ quantifier. 1331b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 1332b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. STO_SP 1333b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. body of stuff being iterated over 1334b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. STATE_SAVE 5 1335b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. JMP 2 1336b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5. LD_SP 1337b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 6. ... 1338b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1339b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Note: TODO: This is pretty inefficient. A mass of saved state is built up 1340c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // then unconditionally discarded. Perhaps introduce a new opcode. Ticket 6056 1341b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1342b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1343b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the STO_SP 1344b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(TRUE); 1345b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stoLoc = fRXPat->fDataSize; 1346b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fDataSize++; // Reserve the data location for storing save stack ptr. 1347b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_STO_SP, stoLoc); 1348b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc); 1349b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1350b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the STATE_SAVE 1351b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_STATE_SAVE, fRXPat->fCompiledPat->size()+2); 1352b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1353c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1354b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the JMP 1355b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_JMP, topLoc+1); 1356b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1357b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1358b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the LD_SP 1359b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LD_SP, stoLoc); 1360b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1361b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1362b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1363b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1364b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPossessiveStar: 1365b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Possessive *+ quantifier. 1366b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 1367b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. STO_SP loc 1368b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. STATE_SAVE 5 1369b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. body of stuff being iterated over 1370b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. JMP 2 1371b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5. LD_SP loc 1372b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 6 ... 1373b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // TODO: do something to cut back the state stack each time through the loop. 1374b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1375b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Reserve two slots at the top of the block. 1376b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(TRUE); 1377b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topLoc); 1378b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1379b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // emit STO_SP loc 1380b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stoLoc = fRXPat->fDataSize; 1381b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fDataSize++; // Reserve the data location for storing save stack ptr. 1382b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_STO_SP, stoLoc); 1383b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc); 1384b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1385b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the SAVE_STATE 5 1386b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t L7 = fRXPat->fCompiledPat->size()+1; 1387b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_STATE_SAVE, L7); 1388b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc+1); 1389b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1390c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Append the JMP operation. 1391b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_JMP, topLoc+1); 1392b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1393b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1394b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the LD_SP loc 1395b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LD_SP, stoLoc); 1396b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1397b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1398b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1399b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1400b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doPossessiveOpt: 1401b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Possessive ?+ quantifier. 1402b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compiles to 1403b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. STO_SP loc 1404b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. SAVE_STATE 5 1405b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. body of optional block 1406b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. LD_SP loc 1407b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5. ... 1408b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1409b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1410b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Reserve two slots at the top of the block. 1411b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topLoc = blockTopLoc(TRUE); 1412b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topLoc); 1413b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1414b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the STO_SP 1415b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stoLoc = fRXPat->fDataSize; 1416b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fDataSize++; // Reserve the data location for storing save stack ptr. 1417b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_STO_SP, stoLoc); 1418b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc); 1419b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1420b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the SAVE_STATE 1421b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t continueLoc = fRXPat->fCompiledPat->size()+1; 1422b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_STATE_SAVE, continueLoc); 1423b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topLoc+1); 1424b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1425b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Emit the LD_SP 1426b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LD_SP, stoLoc); 1427b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 1428b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1429b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1430b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1431b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1432b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBeginMatchMode: 1433b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fNewModeFlags = fModeFlags; 1434b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fSetModeFlag = TRUE; 1435b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1436b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1437b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doMatchMode: // (?i) and similar 1438b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1439b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t bit = 0; 1440b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (fC.fChar) { 1441b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 0x69: /* 'i' */ bit = UREGEX_CASE_INSENSITIVE; break; 1442c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case 0x64: /* 'd' */ bit = UREGEX_UNIX_LINES; break; 1443b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 0x6d: /* 'm' */ bit = UREGEX_MULTILINE; break; 1444b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 0x73: /* 's' */ bit = UREGEX_DOTALL; break; 1445c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case 0x75: /* 'u' */ bit = 0; /* Unicode casing */ break; 1446b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 0x77: /* 'w' */ bit = UREGEX_UWORD; break; 1447b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 0x78: /* 'x' */ bit = UREGEX_COMMENTS; break; 1448b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 0x2d: /* '-' */ fSetModeFlag = FALSE; break; 1449b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 1450b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); // Should never happen. Other chars are filtered out 1451b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // by the scanner. 1452b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1453b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fSetModeFlag) { 1454b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fNewModeFlags |= bit; 1455b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 1456b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fNewModeFlags &= ~bit; 1457b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1458b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1459b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1460b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1461b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doSetMatchMode: 1462103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Emit code to match any pending literals, using the not-yet changed match mode. 1463103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(); 1464103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 1465b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We've got a (?i) or similar. The match mode is being changed, but 1466b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the change is not scoped to a parenthesized block. 1467b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(fNewModeFlags < 0); 1468b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fModeFlags = fNewModeFlags; 1469b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1470b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1471b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1472b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1473b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doMatchModeParen: 1474b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We've got a (?i: or similar. Begin a parenthesized block, save old 1475b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // mode flags so they can be restored at the close of the block. 1476b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1477b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Compile to a 1478b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which later may be replaced by a save-state if the 1479b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // parenthesized group gets a * quantifier, followed by 1480b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // - NOP, which may later be replaced by a save-state if there 1481b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // is an '|' alternation within the parens. 1482b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1483103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1484b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 1485b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_NOP, 0), *fStatus); 1486b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1487b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // On the Parentheses stack, start a new frame and add the postions 1488b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the two NOPs (a normal non-capturing () frame, except for the 1489b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // saving of the orignal mode flags.) 1490b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fModeFlags, *fStatus); 1491b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(flags, *fStatus); // Frame Marker 1492b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-2, *fStatus); // The first NOP 1493b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.push(fRXPat->fCompiledPat->size()-1, *fStatus); // The second NOP 1494b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1495b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Set the current mode flags to the new values. 1496b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(fNewModeFlags < 0); 1497b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fModeFlags = fNewModeFlags; 1498b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1499b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1500b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1501b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doBadModeFlag: 1502b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_INVALID_FLAG); 1503b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1504b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1505b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case doSuppressComments: 1506b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We have just scanned a '(?'. We now need to prevent the character scanner from 1507b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // treating a '#' as a to-the-end-of-line comment. 1508b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // (This Perl compatibility just gets uglier and uglier to do...) 1509b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fEOLComments = FALSE; 1510b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1511b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1512b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1513c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetAddAmp: 1514c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1515c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1516c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(chAmp); 1517c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1518c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1519c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1520c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetAddDash: 1521c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1522c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1523c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(chDash); 1524c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1525c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1526c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1527c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBackslash_s: 1528c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1529c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1530c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]); 1531c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1532c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1533c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1534c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBackslash_S: 1535c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1536c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1537c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISSPACE_SET]); 1538c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru SSet.complement(); 1539c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->addAll(SSet); 1540c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1541c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1542c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1543c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBackslash_d: 1544c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1545c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1546c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // TODO - make a static set, ticket 6058. 1547c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_ND_MASK, *fStatus); 1548c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1549c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1550c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1551c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBackslash_D: 1552c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1553c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1554c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet digits; 1555c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // TODO - make a static set, ticket 6058. 1556c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru digits.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus); 1557c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru digits.complement(); 1558c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->addAll(digits); 1559c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1560c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1561c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1562c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBackslash_w: 1563c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1564c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1565c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->addAll(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]); 1566c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1567c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1568c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1569c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBackslash_W: 1570c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1571c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set = (UnicodeSet *)fSetStack.peek(); 1572c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet SSet(*RegexStaticSets::gStaticSets->fPropSets[URX_ISWORD_SET]); 1573c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru SSet.complement(); 1574c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->addAll(SSet); 1575c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1576c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1577c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1578c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBegin: 1579103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); 1580c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetStack.push(new UnicodeSet(), *fStatus); 1581c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setStart, *fStatus); 1582c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1583c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setCaseClose, *fStatus); 1584c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1585c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1586c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1587c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBeginDifference1: 1588c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We have scanned something like [[abc]-[ 1589c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Set up a new UnicodeSet for the set beginning with the just-scanned '[' 1590c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Push a Difference operator, which will cause the new set to be subtracted from what 1591c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // went before once it is created. 1592c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setPushOp(setDifference1); 1593c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setStart, *fStatus); 1594c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1595c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setCaseClose, *fStatus); 1596c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1597c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1598c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1599c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBeginIntersection1: 1600c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We have scanned something like [[abc]&[ 1601c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Need both the '&' operator and the open '[' operator. 1602c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setPushOp(setIntersection1); 1603c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setStart, *fStatus); 1604c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1605c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setCaseClose, *fStatus); 1606c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1607c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1608c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1609c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetBeginUnion: 1610c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We have scanned something like [[abc][ 1611c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Need to handle the union operation explicitly [[abc] | [ 1612c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setPushOp(setUnion); 1613c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setStart, *fStatus); 1614c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_CASE_INSENSITIVE) != 0) { 1615c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setCaseClose, *fStatus); 1616c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1617c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1618c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1619c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetDifference2: 1620c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We have scanned something like [abc-- 1621c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Consider this to unambiguously be a set difference operator. 1622c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setPushOp(setDifference2); 1623c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1624c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1625c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetEnd: 1626c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Have encountered the ']' that closes a set. 1627c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Force the evaluation of any pending operations within this set, 1628c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // leave the completed set on the top of the set stack. 1629c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setEval(setEnd); 1630b0ac937921a2c196d8b9da665135bf6ba01a1ccfJean-Baptiste Queru U_ASSERT(fSetOpStack.peeki()==setStart); 1631b0ac937921a2c196d8b9da665135bf6ba01a1ccfJean-Baptiste Queru fSetOpStack.popi(); 1632c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1633c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1634c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetFinish: 1635c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1636c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Finished a complete set expression, including all nested sets. 1637c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The close bracket has already triggered clearing out pending set operators, 1638c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // the operator stack should be empty and the operand stack should have just 1639c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // one entry, the result set. 1640c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(fSetOpStack.empty()); 1641c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *theSet = (UnicodeSet *)fSetStack.pop(); 1642c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(fSetStack.empty()); 1643c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru compileSet(theSet); 1644c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1645c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1646fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 1647c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetIntersection2: 1648c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Have scanned something like [abc&& 1649c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setPushOp(setIntersection2); 1650c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1651c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1652c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetLiteral: 1653c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Union the just-scanned literal character into the set being built. 1654c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // This operation is the highest precedence set operation, so we can always do 1655c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // it immediately, without waiting to see what follows. It is necessary to perform 1656c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // any pending '-' or '&' operation first, because these have the same precedence 1657fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius // as union-ing in a literal' 1658c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1659c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setEval(setUnion); 1660c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1661c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru s->add(fC.fChar); 1662c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fLastSetLiteral = fC.fChar; 1663c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1664c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1665c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1666c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetLiteralEscaped: 1667c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // A back-slash escaped literal character was encountered. 1668c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Processing is the same as with setLiteral, above, with the addition of 1669c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // the optional check for errors on escaped ASCII letters. 1670c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1671c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((fModeFlags & UREGEX_ERROR_ON_UNKNOWN_ESCAPES) != 0 && 1672c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru ((fC.fChar >= 0x41 && fC.fChar<= 0x5A) || // in [A-Z] 1673c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru (fC.fChar >= 0x61 && fC.fChar <= 0x7a))) { // in [a-z] 1674c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_BAD_ESCAPE_SEQUENCE); 1675c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1676c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setEval(setUnion); 1677c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1678c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru s->add(fC.fChar); 1679c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fLastSetLiteral = fC.fChar; 1680c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1681c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1682c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1683c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetNamedChar: 1684c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Scanning a \N{UNICODE CHARACTER NAME} 1685c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Aside from the source of the character, the processing is identical to doSetLiteral, 1686c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // above. 1687c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1688c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UChar32 c = scanNamedChar(); 1689c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setEval(setUnion); 1690c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1691c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru s->add(c); 1692c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fLastSetLiteral = c; 1693c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1694c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1695c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1696c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetNamedRange: 1697c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We have scanned literal-\N{CHAR NAME}. Add the range to the set. 1698c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The left character is already in the set, and is saved in fLastSetLiteral. 1699c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The right side needs to be picked up, the scan is at the 'N'. 1700c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Lower Limit > Upper limit being an error matches both Java 1701c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // and ICU UnicodeSet behavior. 1702c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1703c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UChar32 c = scanNamedChar(); 1704c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_SUCCESS(*fStatus) && fLastSetLiteral > c) { 1705c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_INVALID_RANGE); 1706c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1707c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1708c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru s->add(fLastSetLiteral, c); 1709c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fLastSetLiteral = c; 1710c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1711c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1712c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1713c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 171450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho case doSetNegate: 1715c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Scanned a '^' at the start of a set. 1716c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Push the negation operator onto the set op stack. 1717c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // A twist for case-insensitive matching: 1718c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // the case closure operation must happen _before_ negation. 1719c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // But the case closure operation will already be on the stack if it's required. 1720c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // This requires checking for case closure, and swapping the stack order 1721c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // if it is present. 1722c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1723c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t tosOp = fSetOpStack.peeki(); 1724c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (tosOp == setCaseClose) { 1725c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.popi(); 1726c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setNegation, *fStatus); 1727c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setCaseClose, *fStatus); 1728c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else { 1729c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(setNegation, *fStatus); 1730c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1731c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1732c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1733c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1734c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetNoCloseError: 1735c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_MISSING_CLOSE_BRACKET); 1736c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1737c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1738c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetOpError: 1739c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_RULE_SYNTAX); // -- or && at the end of a set. Illegal. 1740c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1741c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1742c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetPosixProp: 1743c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1744c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = scanPosixProp(); 1745c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (s != NULL) { 1746c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *tos = (UnicodeSet *)fSetStack.peek(); 1747c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru tos->addAll(*s); 1748c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete s; 1749c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } // else error. scanProp() reported the error status already. 1750c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1751c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1752fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 1753c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetProp: 1754c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Scanned a \p \P within [brackets]. 1755c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1756c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = scanProp(); 1757c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (s != NULL) { 1758c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *tos = (UnicodeSet *)fSetStack.peek(); 1759c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru tos->addAll(*s); 1760c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete s; 1761c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } // else error. scanProp() reported the error status already. 1762c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1763c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1764c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1765c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1766c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case doSetRange: 1767c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // We have scanned literal-literal. Add the range to the set. 1768c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The left character is already in the set, and is saved in fLastSetLiteral. 1769c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The right side is the current character. 1770c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Lower Limit > Upper limit being an error matches both Java 1771c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // and ICU UnicodeSet behavior. 1772c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 1773c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fLastSetLiteral > fC.fChar) { 1774fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius error(U_REGEX_INVALID_RANGE); 1775c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1776c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *s = (UnicodeSet *)fSetStack.peek(); 1777c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru s->add(fLastSetLiteral, fC.fChar); 1778c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 1779c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 1780c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 1781b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 1782b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); 1783b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_INTERNAL_ERROR); 1784b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 1785b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1786b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1787b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 1788b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru returnVal = FALSE; 1789b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1790b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1791b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return returnVal; 1792b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 1793b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1794b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1795b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1796b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1797b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1798b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// literalChar We've encountered a literal character from the pattern, 1799b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// or an escape sequence that reduces to a character. 1800b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Add it to the string containing all literal chars/strings from 1801b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// the pattern. 1802b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1803b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1804b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::literalChar(UChar32 c) { 1805103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fLiteralChars.append(c); 1806b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 1807b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1808b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1809b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1810b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1811b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// fixLiterals When compiling something that can follow a literal 1812103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius// string in a pattern, emit the code to match the 1813103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius// accumulated literal string. 1814b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1815b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Optionally, split the last char of the string off into 1816b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// a single "ONE_CHAR" operation, so that quantifiers can 1817b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// apply to that char alone. Example: abc* 1818b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The * must apply to the 'c' only. 1819b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1820b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1821b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::fixLiterals(UBool split) { 1822103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius int32_t op = 0; // An op from/for the compiled pattern. 1823103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 1824103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // If no literal characters have been scanned but not yet had code generated 1825103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // for them, nothing needs to be done. 1826103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (fLiteralChars.length() == 0) { 1827b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 1828b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1829b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1830103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius int32_t indexOfLastCodePoint = fLiteralChars.moveIndex32(fLiteralChars.length(), -1); 1831103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius UChar32 lastCodePoint = fLiteralChars.char32At(indexOfLastCodePoint); 1832103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 1833fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius // Split: We need to ensure that the last item in the compiled pattern 1834103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // refers only to the last literal scanned in the pattern, so that 1835103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // quantifiers (*, +, etc.) affect only it, and not a longer string. 1836103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Split before case folding for case insensitive matches. 1837103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 1838103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (split) { 1839103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fLiteralChars.truncate(indexOfLastCodePoint); 1840103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); // Recursive call, emit code to match the first part of the string. 1841103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Note that the truncated literal string may be empty, in which case 1842103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // nothing will be emitted. 1843b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1844103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius literalChar(lastCodePoint); // Re-add the last code point as if it were a new literal. 1845103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(FALSE); // Second recursive call, code for the final code point. 1846b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 1847b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1848b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1849103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // If we are doing case-insensitive matching, case fold the string. This may expand 1850103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // the string, e.g. the German sharp-s turns into "ss" 1851103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 1852103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fLiteralChars.foldCase(); 1853103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius indexOfLastCodePoint = fLiteralChars.moveIndex32(fLiteralChars.length(), -1); 1854103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius lastCodePoint = fLiteralChars.char32At(indexOfLastCodePoint); 1855103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 1856103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 1857103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (indexOfLastCodePoint == 0) { 1858103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Single character, emit a URX_ONECHAR op to match it. 1859fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius if ((fModeFlags & UREGEX_CASE_INSENSITIVE) && 1860103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius u_hasBinaryProperty(lastCodePoint, UCHAR_CASE_SENSITIVE)) { 1861103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius op = URX_BUILD(URX_ONECHAR_I, lastCodePoint); 1862103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } else { 1863103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius op = URX_BUILD(URX_ONECHAR, lastCodePoint); 1864103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 1865103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fCompiledPat->addElement(op, *fStatus); 1866b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 1867103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Two or more chars, emit a URX_STRING to match them. 1868103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 1869103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius op = URX_BUILD(URX_STRING_I, fRXPat->fLiteralText.length()); 1870103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } else { 1871103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // TODO here: add optimization to split case sensitive strings of length two 1872103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // into two single char ops, for efficiency. 1873103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius op = URX_BUILD(URX_STRING, fRXPat->fLiteralText.length()); 1874103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 1875103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fCompiledPat->addElement(op, *fStatus); 1876103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius op = URX_BUILD(URX_STRING_LEN, fLiteralChars.length()); 1877103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fCompiledPat->addElement(op, *fStatus); 1878fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 1879103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Add this string into the accumulated strings of the compiled pattern. 1880103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fLiteralText.append(fLiteralChars); 1881b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1882103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 1883103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fLiteralChars.remove(); 1884b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 1885b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1886b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1887b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1888b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1889b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1890b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1891b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1892b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1893b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// insertOp() Insert a slot for a new opcode into the already 1894b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// compiled pattern code. 1895b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1896b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Fill the slot with a NOP. Our caller will replace it 1897b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// with what they really wanted. 1898b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1899b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1900b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::insertOp(int32_t where) { 190150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UVector64 *code = fRXPat->fCompiledPat; 1902b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(where>0 && where < code->size()); 1903b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1904b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t nop = URX_BUILD(URX_NOP, 0); 1905b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru code->insertElementAt(nop, where, *fStatus); 1906b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1907b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Walk through the pattern, looking for any ops with targets that 1908b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // were moved down by the insert. Fix them. 1909b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loc; 1910b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc=0; loc<code->size(); loc++) { 191150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t op = (int32_t)code->elementAti(loc); 1912b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t opType = URX_TYPE(op); 1913b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t opValue = URX_VAL(op); 1914b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if ((opType == URX_JMP || 1915b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType == URX_JMPX || 1916b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType == URX_STATE_SAVE || 1917b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType == URX_CTR_LOOP || 1918b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType == URX_CTR_LOOP_NG || 1919b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType == URX_JMP_SAV || 1920c53bf83a40a6888f5b246a73f13f6c919de1f5f9claireho opType == URX_JMP_SAV_X || 1921b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType == URX_RELOC_OPRND) && opValue > where) { 1922b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Target location for this opcode is after the insertion point and 1923b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // needs to be incremented to adjust for the insertion. 1924b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opValue++; 1925b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(opType, opValue); 1926b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru code->setElementAt(op, loc); 1927b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1928b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1929b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1930b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Now fix up the parentheses stack. All positive values in it are locations in 1931b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the compiled pattern. (Negative values are frame boundaries, and don't need fixing.) 1932b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc=0; loc<fParenStack.size(); loc++) { 1933b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t x = fParenStack.elementAti(loc); 1934b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(x < code->size()); 1935b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (x>where) { 1936b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru x++; 1937b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fParenStack.setElementAt(x, loc); 1938b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1939b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1940b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1941b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fMatchCloseParen > where) { 1942b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fMatchCloseParen++; 1943b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1944b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fMatchOpenParen > where) { 1945b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fMatchOpenParen++; 1946b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1947b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 1948b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1949b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1950b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1951b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1952b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1953b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// blockTopLoc() Find or create a location in the compiled pattern 1954b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// at the start of the operation or block that has 1955b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// just been compiled. Needed when a quantifier (* or 1956b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// whatever) appears, and we need to add an operation 1957b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// at the start of the thing being quantified. 1958b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1959b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// (Parenthesized Blocks) have a slot with a NOP that 1960b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// is reserved for this purpose. .* or similar don't 1961b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// and a slot needs to be added. 1962b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1963b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// parameter reserveLoc : TRUE - ensure that there is space to add an opcode 1964b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// at the returned location. 1965c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// FALSE - just return the address, 1966b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// do not reserve a location there. 1967b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1968b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 1969b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruint32_t RegexCompile::blockTopLoc(UBool reserveLoc) { 1970b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t theLoc; 1971103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fixLiterals(TRUE); // Emit code for any pending literals. 1972103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // If last item was a string, emit separate op for the its last char. 1973b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fRXPat->fCompiledPat->size() == fMatchCloseParen) 1974b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 1975b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The item just processed is a parenthesized block. 1976b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru theLoc = fMatchOpenParen; // A slot is already reserved for us. 1977b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(theLoc > 0); 1978b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(((uint32_t)fRXPat->fCompiledPat->elementAti(theLoc))) == URX_NOP); 1979b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1980b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru else { 1981103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Item just compiled is a single thing, a ".", or a single char, a string or a set reference. 1982b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // No slot for STATE_SAVE was pre-reserved in the compiled code. 1983b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We need to make space now. 1984b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru theLoc = fRXPat->fCompiledPat->size()-1; 1985103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius int32_t opAtTheLoc = (int32_t)fRXPat->fCompiledPat->elementAti(theLoc); 1986103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (URX_TYPE(opAtTheLoc) == URX_STRING_LEN) { 1987103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Strings take two opcode, we want the position of the first one. 1988103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // We can have a string at this point if a single character case-folded to two. 1989103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius theLoc--; 1990103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 1991b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (reserveLoc) { 1992b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t nop = URX_BUILD(URX_NOP, 0); 1993b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->insertElementAt(nop, theLoc, *fStatus); 1994b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1995b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 1996b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return theLoc; 1997b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 1998b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 1999b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2000b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2001b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2002b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2003b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// handleCloseParen When compiling a close paren, we need to go back 2004b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// and fix up any JMP or SAVE operations within the 2005b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// parenthesized block that need to target the end 2006b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// of the block. The locations of these are kept on 2007b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// the paretheses stack. 2008b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2009b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// This function is called both when encountering a 2010b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// real ) and at the end of the pattern. 2011b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2012b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2013b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::handleCloseParen() { 2014b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t patIdx; 2015b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t patOp; 2016b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fParenStack.size() <= 0) { 2017b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_MISMATCHED_PAREN); 2018b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 2019b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2020b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2021103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Emit code for any pending literals. 2022b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fixLiterals(FALSE); 2023b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2024b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Fixup any operations within the just-closed parenthesized group 2025b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // that need to reference the end of the (block). 2026b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // (The first one popped from the stack is an unused slot for 2027b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // alternation (OR) state save, but applying the fixup to it does no harm.) 2028b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 2029b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru patIdx = fParenStack.popi(); 2030b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (patIdx < 0) { 2031b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // value < 0 flags the start of the frame on the paren stack. 2032b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2033b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2034b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(patIdx>0 && patIdx <= fRXPat->fCompiledPat->size()); 203550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho patOp = (int32_t)fRXPat->fCompiledPat->elementAti(patIdx); 2036b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_VAL(patOp) == 0); // Branch target for JMP should not be set. 2037b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru patOp |= fRXPat->fCompiledPat->size(); // Set it now. 2038b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(patOp, patIdx); 2039b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fMatchOpenParen = patIdx; 2040b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2041b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2042b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // At the close of any parenthesized block, restore the match mode flags to 2043b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the value they had at the open paren. Saved value is 2044c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // at the top of the paren stack. 2045b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fModeFlags = fParenStack.popi(); 2046b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(fModeFlags < 0); 2047c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2048b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // DO any additional fixups, depending on the specific kind of 2049b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // parentesized grouping this is 2050b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2051b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (patIdx) { 2052b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case plain: 2053b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case flags: 2054b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // No additional fixups required. 2055b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // (Grouping-only parentheses) 2056b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2057b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case capturing: 2058b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Capturing Parentheses. 2059b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert a End Capture op into the pattern. 2060b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The frame offset of the variables for this cg is obtained from the 2061b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // start capture op and put it into the end-capture op. 2062b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 206350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t captureOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1); 2064b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(captureOp) == URX_START_CAPTURE); 2065b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2066b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t frameVarLocation = URX_VAL(captureOp); 2067b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t endCaptureOp = URX_BUILD(URX_END_CAPTURE, frameVarLocation); 2068b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(endCaptureOp, *fStatus); 2069b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2070b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2071b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case atomic: 2072b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Atomic Parenthesis. 2073b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert a LD_SP operation to restore the state stack to the position 2074b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // it was when the atomic parens were entered. 2075b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 207650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t stoOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen+1); 2077b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(stoOp) == URX_STO_SP); 2078b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stoLoc = URX_VAL(stoOp); 2079b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t ldOp = URX_BUILD(URX_LD_SP, stoLoc); 2080b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(ldOp, *fStatus); 2081b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2082b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2083b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2084b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case lookAhead: 2085b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 208650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5); 2087b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(startOp) == URX_LA_START); 2088b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = URX_VAL(startOp); 2089b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LA_END, dataLoc); 2090b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2091b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2092b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2093b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2094b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case negLookAhead: 2095b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2096b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // See comment at doOpenLookAheadNeg 209750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-1); 2098b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(startOp) == URX_LA_START); 2099b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = URX_VAL(startOp); 2100b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LA_END, dataLoc); 2101b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2102c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru op = URX_BUILD(URX_BACKTRACK, 0); 2103c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 210450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = URX_BUILD(URX_LA_END, dataLoc); 2105b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2106b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2107b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Patch the URX_SAVE near the top of the block. 2108c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The destination of the SAVE is the final LA_END that was just added. 210950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t saveOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen); 2110b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(saveOp) == URX_STATE_SAVE); 2111c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t dest = fRXPat->fCompiledPat->size()-1; 2112b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru saveOp = URX_BUILD(URX_STATE_SAVE, dest); 2113b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(saveOp, fMatchOpenParen); 2114b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2115b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2116b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2117b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case lookBehind: 2118b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2119b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // See comment at doOpenLookBehind. 2120c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2121b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Append the URX_LB_END and URX_LA_END to the compiled pattern. 212250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-4); 2123b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(startOp) == URX_LB_START); 2124b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = URX_VAL(startOp); 2125b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LB_END, dataLoc); 2126b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2127b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_LA_END, dataLoc); 2128b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2129b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2130b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Determine the min and max bounds for the length of the 2131b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // string that the pattern can match. 2132b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // An unbounded upper limit is an error. 2133b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t patEnd = fRXPat->fCompiledPat->size() - 1; 2134b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t minML = minMatchLength(fMatchOpenParen, patEnd); 2135b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t maxML = maxMatchLength(fMatchOpenParen, patEnd); 2136b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (maxML == INT32_MAX) { 2137b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_LOOK_BEHIND_LIMIT); 2138b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2139b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2140b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(minML <= maxML); 2141b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2142b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert the min and max match len bounds into the URX_LB_CONT op that 2143b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // appears at the top of the look-behind block, at location fMatchOpenParen+1 2144b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(minML, fMatchOpenParen-2); 2145b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(maxML, fMatchOpenParen-1); 2146b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2147b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2148b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2149b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2150b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2151b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2152b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case lookBehindN: 2153b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2154b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // See comment at doOpenLookBehindNeg. 2155c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2156b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Append the URX_LBN_END to the compiled pattern. 215750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t startOp = (int32_t)fRXPat->fCompiledPat->elementAti(fMatchOpenParen-5); 2158b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(startOp) == URX_LB_START); 2159b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dataLoc = URX_VAL(startOp); 2160b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(URX_LBN_END, dataLoc); 2161b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2162b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2163b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Determine the min and max bounds for the length of the 2164b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // string that the pattern can match. 2165b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // An unbounded upper limit is an error. 2166b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t patEnd = fRXPat->fCompiledPat->size() - 1; 2167b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t minML = minMatchLength(fMatchOpenParen, patEnd); 2168b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t maxML = maxMatchLength(fMatchOpenParen, patEnd); 2169b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (maxML == INT32_MAX) { 2170b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_LOOK_BEHIND_LIMIT); 2171b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2172b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2173b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(minML <= maxML); 2174b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2175b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert the min and max match len bounds into the URX_LB_CONT op that 2176b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // appears at the top of the look-behind block, at location fMatchOpenParen+1 2177b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(minML, fMatchOpenParen-3); 2178b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(maxML, fMatchOpenParen-2); 2179b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2180b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Insert the pattern location to continue at after a successful match 2181b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // as the last operand of the URX_LBN_CONT 2182b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_RELOC_OPRND, fRXPat->fCompiledPat->size()); 2183b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, fMatchOpenParen-1); 2184b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2185b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2186b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2187b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2188b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2189b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 2190b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); 2191b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2192b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2193b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // remember the next location in the compiled pattern. 2194b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The compilation of Quantifiers will look at this to see whether its looping 2195b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // over a parenthesized block or a single item 2196b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fMatchCloseParen = fRXPat->fCompiledPat->size(); 2197b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 2198b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2199b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2200b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2201b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2202b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2203b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// compileSet Compile the pattern operations for a reference to a 2204b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// UnicodeSet. 2205b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2206b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2207b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::compileSet(UnicodeSet *theSet) 2208b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru{ 2209b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (theSet == NULL) { 2210b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 2211b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2212c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Remove any strings from the set. 2213c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // There shoudn't be any, but just in case. 2214c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // (Case Closure can add them; if we had a simple case closure avaialble that 2215c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // ignored strings, that would be better.) 2216c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru theSet->removeAllStrings(); 2217b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t setSize = theSet->size(); 2218b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2219b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (setSize) { 2220c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case 0: 2221b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2222c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Set of no elements. Always fails to match. 2223b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(URX_BUILD(URX_BACKTRACK, 0), *fStatus); 2224b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru delete theSet; 2225b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2226b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2227c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2228b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case 1: 2229b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2230b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The set contains only a single code point. Put it into 2231b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the compiled pattern as a single char operation rather 2232b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // than a set, and discard the set itself. 223350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho literalChar(theSet->charAt(0)); 2234b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru delete theSet; 2235b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2236b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2237c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2238c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru default: 2239b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2240b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The set contains two or more chars. (the normal case) 2241b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Put it into the compiled pattern as a set. 2242b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t setNumber = fRXPat->fSets->size(); 2243b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fSets->addElement(theSet, *fStatus); 2244b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t setOp = URX_BUILD(URX_SETREF, setNumber); 2245b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(setOp, *fStatus); 2246b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2247b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2248b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 2249b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2250b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2251b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2252b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2253b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// compileInterval Generate the code for a {min, max} style interval quantifier. 2254b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Except for the specific opcodes used, the code is the same 2255b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// for all three types (greedy, non-greedy, possessive) of 2256b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// intervals. The opcodes are supplied as parameters. 225759d709d503bab6e2b61931737e662dd293b40578ccornelius// (There are two sets of opcodes - greedy & possessive use the 225859d709d503bab6e2b61931737e662dd293b40578ccornelius// same ones, while non-greedy has it's own.) 2259b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2260b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The code for interval loops has this form: 2261b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 0 CTR_INIT counter loc (in stack frame) 2262b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 1 5 patt address of CTR_LOOP at bottom of block 2263b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2 min count 2264b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3 max count (-1 for unbounded) 2265b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 4 ... block to be iterated over 2266c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 5 CTR_LOOP 2267c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 2268c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// In 2269b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2270b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::compileInterval(int32_t InitOp, int32_t LoopOp) 2271b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru{ 2272b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The CTR_INIT op at the top of the block with the {n,m} quantifier takes 2273b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // four slots in the compiled code. Reserve them. 2274b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topOfBlock = blockTopLoc(TRUE); 2275b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topOfBlock); 2276b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topOfBlock); 2277b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topOfBlock); 2278b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2279b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The operands for the CTR_INIT opcode include the index in the matcher data 228059d709d503bab6e2b61931737e662dd293b40578ccornelius // of the counter. Allocate it now. There are two data items 228159d709d503bab6e2b61931737e662dd293b40578ccornelius // counterLoc --> Loop counter 228259d709d503bab6e2b61931737e662dd293b40578ccornelius // +1 --> Input index (for breaking non-progressing loops) 228359d709d503bab6e2b61931737e662dd293b40578ccornelius // (Only present if unbounded upper limit on loop) 2284b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t counterLoc = fRXPat->fFrameSize; 2285b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fFrameSize++; 228659d709d503bab6e2b61931737e662dd293b40578ccornelius if (fIntervalUpper < 0) { 228759d709d503bab6e2b61931737e662dd293b40578ccornelius fRXPat->fFrameSize++; 228859d709d503bab6e2b61931737e662dd293b40578ccornelius } 2289b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2290b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op = URX_BUILD(InitOp, counterLoc); 2291b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topOfBlock); 2292b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2293b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The second operand of CTR_INIT is the location following the end of the loop. 2294b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Must put in as a URX_RELOC_OPRND so that the value will be adjusted if the 2295b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // compilation of something later on causes the code to grow and the target 2296b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // position to move. 2297b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loopEnd = fRXPat->fCompiledPat->size(); 2298b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(URX_RELOC_OPRND, loopEnd); 2299b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, topOfBlock+1); 2300b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2301b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Followed by the min and max counts. 2302b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(fIntervalLow, topOfBlock+2); 2303b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(fIntervalUpper, topOfBlock+3); 2304b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2305b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Apend the CTR_LOOP op. The operand is the location of the CTR_INIT op. 2306b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Goes at end of the block being looped over, so just append to the code so far. 2307b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(LoopOp, topOfBlock); 2308b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2309b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2310b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if ((fIntervalLow & 0xff000000) != 0 || 231127f654740f2a26ad62a5c155af9199af9e69b889claireho (fIntervalUpper > 0 && (fIntervalUpper & 0xff000000) != 0)) { 2312b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_NUMBER_TOO_BIG); 2313b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2314b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2315b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalLow > fIntervalUpper && fIntervalUpper != -1) { 2316b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_MAX_LT_MIN); 2317b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2318b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 2319b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2320b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2321b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2322b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruUBool RegexCompile::compileInlineInterval() { 2323b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalUpper > 10 || fIntervalUpper < fIntervalLow) { 2324b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Too big to inline. Fail, which will cause looping code to be generated. 2325b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // (Upper < Lower picks up unbounded upper and errors, both.) 2326b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return FALSE; 2327b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2328b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2329b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t topOfBlock = blockTopLoc(FALSE); 2330b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalUpper == 0) { 2331b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Pathological case. Attempt no matches, as if the block doesn't exist. 2332b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setSize(topOfBlock); 2333b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return TRUE; 2334b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2335b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2336b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (topOfBlock != fRXPat->fCompiledPat->size()-1 && fIntervalUpper != 1) { 2337b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The thing being repeated is not a single op, but some 2338b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // more complex block. Do it as a loop, not inlines. 2339b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Note that things "repeated" a max of once are handled as inline, because 2340b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the one copy of the code already generated is just fine. 2341b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return FALSE; 2342b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2343b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2344b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Pick up the opcode that is to be repeated 2345b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 234650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(topOfBlock); 2347b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2348c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Compute the pattern location where the inline sequence 2349b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // will end, and set up the state save op that will be needed. 2350c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 2351b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t endOfSequenceLoc = fRXPat->fCompiledPat->size()-1 2352b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru + fIntervalUpper + (fIntervalUpper-fIntervalLow); 2353b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t saveOp = URX_BUILD(URX_STATE_SAVE, endOfSequenceLoc); 2354b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fIntervalLow == 0) { 2355b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru insertOp(topOfBlock); 2356b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(saveOp, topOfBlock); 2357b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2358b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2359b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2360b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2361b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop, emitting the op for the thing being repeated each time. 2362b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop starts at 1 because one instance of the op already exists in the pattern, 2363b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // it was put there when it was originally encountered. 2364b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t i; 2365b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (i=1; i<fIntervalUpper; i++ ) { 2366b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (i == fIntervalLow) { 2367b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(saveOp, *fStatus); 2368b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2369b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (i > fIntervalLow) { 2370b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(saveOp, *fStatus); 2371b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2372b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->addElement(op, *fStatus); 2373b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2374b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return TRUE; 2375b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 2376b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2377b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2378b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2379b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2380b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2381b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// matchStartType Determine how a match can start. 2382b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Used to optimize find() operations. 2383b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2384b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Operation is very similar to minMatchLength(). Walk the compiled 2385b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// pattern, keeping an on-going minimum-match-length. For any 2386b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// op where the min match coming in is zero, add that ops possible 2387b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// starting matches to the possible starts for the overall pattern. 2388b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2389b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2390b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::matchStartType() { 2391b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 2392b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 2393b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2394b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2395b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2396b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loc; // Location in the pattern of the current op being processed. 2397b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op; // The op being processed 2398b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t opType; // The opcode type of the op 2399b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t currentLen = 0; // Minimum length of a match to this point (loc) in the pattern 2400b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t numInitialStrings = 0; // Number of strings encountered that could match at start. 2401b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2402b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UBool atStart = TRUE; // True if no part of the pattern yet encountered 2403b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // could have advanced the position in a match. 2404b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // (Maximum match length so far == 0) 2405b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2406b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // forwardedLength is a vector holding minimum-match-length values that 2407b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // are propagated forward in the pattern by JMP or STATE_SAVE operations. 2408b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // It must be one longer than the pattern being checked because some ops 2409b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // will jmp to a end-of-block+1 location from within a block, and we must 2410b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // count those when checking the block. 2411b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t end = fRXPat->fCompiledPat->size(); 2412b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UVector32 forwardedLength(end+1, *fStatus); 2413b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setSize(end+1); 2414b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc=3; loc<end; loc++) { 2415b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(INT32_MAX, loc); 2416b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2417b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2418b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc = 3; loc<end; loc++) { 241950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 2420b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType = URX_TYPE(op); 2421b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2422b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The loop is advancing linearly through the pattern. 2423b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // If the op we are now at was the destination of a branch in the pattern, 2424b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // and that path has a shorter minimum length than the current accumulated value, 2425b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // replace the current accumulated value. 2426b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(loc) < currentLen) { 2427b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc); 2428b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); 2429b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2430b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2431b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (opType) { 2432b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Ops that don't change the total length matched 2433b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RESERVED_OP: 2434b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END: 2435c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_FAIL: 2436b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING_LEN: 2437b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_NOP: 2438b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_START_CAPTURE: 2439b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END_CAPTURE: 2440b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_B: 2441b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_BU: 2442b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_G: 2443b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_Z: 2444b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOLLAR: 2445c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_M: 2446c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_D: 2447c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_MD: 2448b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RELOC_OPRND: 2449b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_INP_LOC: 2450b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKREF: // BackRef. Must assume that it might be a zero length match 2451b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKREF_I: 2452fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 2453b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. 2454b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LD_SP: 2455b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2456c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2457b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET: 2458b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (atStart) { 2459b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_START; 2460b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2461b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2462b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2463b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET_M: 2464c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_CARET_M_UNIX: 2465b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (atStart) { 2466b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_LINE; 2467b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2468b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2469c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2470b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR: 2471b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2472b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // This character could appear at the start of a match. 2473b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Add it to the set of possible starting characters. 2474b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->add(URX_VAL(op)); 2475b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2476b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2477b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2478b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2479b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2480b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2481c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2482c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_SETREF: 2483b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2484b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t sn = URX_VAL(op); 2485b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(sn > 0 && sn < fRXPat->fSets->size()); 2486b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn); 2487b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->addAll(*s); 2488b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2489b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2490b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2491b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2492b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2493b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2494b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_SR_I: 2495b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // [Set]*, like a SETREF, above, in what it can match, 2496b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // but may not match at all, so currentLen is not incremented. 2497b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2498b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t sn = URX_VAL(op); 2499b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(sn > 0 && sn < fRXPat->fSets->size()); 2500b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru const UnicodeSet *s = (UnicodeSet *)fRXPat->fSets->elementAt(sn); 2501b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->addAll(*s); 2502b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2503b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2504b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2505b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2506b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2507b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_DOT_I: 2508b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2509b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // .* at the start of a pattern. 2510b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Any character can begin the match. 2511b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->clear(); 2512b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->complement(); 2513b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2514b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2515b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2516b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2517b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2518b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2519c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_STATIC_SETREF: 2520b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2521b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t sn = URX_VAL(op); 2522b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(sn>0 && sn<URX_LAST_SET); 2523b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru const UnicodeSet *s = fRXPat->fStaticSets[sn]; 2524b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->addAll(*s); 2525b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2526b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2527b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2528b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2529b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2530b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2531b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2532b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2533c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_STAT_SETREF_N: 2534b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2535b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t sn = URX_VAL(op); 2536b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru const UnicodeSet *s = fRXPat->fStaticSets[sn]; 2537b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UnicodeSet sc(*s); 2538b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru sc.complement(); 2539b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->addAll(sc); 2540b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2541b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2542b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2543b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2544b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2545b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2546b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2547b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2548b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_D: 2549b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Digit Char 2550b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2551c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet s; 2552b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru s.applyIntPropertyValue(UCHAR_GENERAL_CATEGORY_MASK, U_GC_ND_MASK, *fStatus); 2553b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_VAL(op) != 0) { 2554b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru s.complement(); 2555b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2556b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->addAll(s); 2557b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2558b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2559b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2560b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2561b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2562b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2563b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2564b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR_I: 2565b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Case Insensitive Single Character. 2566b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2567b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UChar32 c = URX_VAL(op); 2568b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (u_hasBinaryProperty(c, UCHAR_CASE_SENSITIVE)) { 2569103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 2570103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Disable optimizations on first char of match. 2571103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // TODO: Compute the set of chars that case fold to this char, or to 2572103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // a string that begins with this char. 2573103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // For simple case folding, this code worked: 2574103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // UnicodeSet s(c, c); 2575103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // s.closeOver(USET_CASE_INSENSITIVE); 2576103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // fRXPat->fInitialChars->addAll(s); 2577103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 2578103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fInitialChars->clear(); 2579103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius fRXPat->fInitialChars->complement(); 2580b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 2581b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Char has no case variants. Just add it as-is to the 2582b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // set of possible starting chars. 2583b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->add(c); 2584b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2585b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2586b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2587b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2588b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2589b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2590b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2591b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2592b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_X: // Grahpeme Cluster. Minimum is 1, max unbounded. 2593b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY_ALL: // . matches one or two. 2594b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY: 2595c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOTANY_UNIX: 2596b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2597b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // These constructs are all bad news when they appear at the start 2598b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of a match. Any character can begin the match. 2599b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->clear(); 2600b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->complement(); 2601b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; 2602b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2603b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2604b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2605b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2606b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2607b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2608b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMPX: 2609b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; // Except for extra operand on URX_JMPX, same as URX_JMP. 2610b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP: 2611b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2612b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 2613b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest < loc) { 2614b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop of some kind. Can safely ignore, the worst that will happen 2615b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // is that we understate the true minimum length 2616b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc+1); 2617c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2618b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 2619b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Forward jump. Propagate the current min length to the target loc of the jump. 2620b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(jmpDest <= end+1); 2621b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(jmpDest) > currentLen) { 2622b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 2623b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2624b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2625b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2626b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2627b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2628b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2629b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV: 2630b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV_X: 2631b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Combo of state save to the next loc, + jmp backwards. 2632b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Net effect on min. length computation is nothing. 2633b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2634b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2635b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2636c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_BACKTRACK: 2637b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Fails are kind of like a branch, except that the min length was 2638b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // propagated already, by the state save. 2639b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc+1); 2640b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2641b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2642b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2643b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2644b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATE_SAVE: 2645b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2646b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // State Save, for forward jumps, propagate the current minimum. 2647b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the state save. 2648b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 2649b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest > loc) { 2650b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen < forwardedLength.elementAti(jmpDest)) { 2651b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 2652b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2653c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 2654b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2655b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2656b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2657c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2658b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2659b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2660b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2661b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING: 2662b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2663b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 266450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 2665b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stringLen = URX_VAL(stringLenOp); 2666b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN); 2667b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(stringLenOp >= 2); 2668b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2669b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Add the starting character of this string to the set of possible starting 2670b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // characters for this pattern. 2671b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stringStartIdx = URX_VAL(op); 2672b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UChar32 c = fRXPat->fLiteralText.char32At(stringStartIdx); 2673b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->add(c); 2674b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2675b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Remember this string. After the entire pattern has been checked, 2676b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // if nothing else is identified that can start a match, we'll use it. 2677b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings++; 2678b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialStringIdx = stringStartIdx; 2679b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialStringLen = stringLen; 2680b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2681c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2682b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen += stringLen; 2683b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2684b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2685b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2686b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2687b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING_I: 2688b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2689b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Case-insensitive string. Unlike exact-match strings, we won't 2690b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // attempt a string search for possible match positions. But we 2691b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // do update the set of possible starting characters. 2692b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 269350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 2694b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stringLen = URX_VAL(stringLenOp); 2695b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(URX_TYPE(stringLenOp) == URX_STRING_LEN); 2696b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(stringLenOp >= 2); 2697b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == 0) { 2698b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Add the starting character of this string to the set of possible starting 2699b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // characters for this pattern. 2700b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t stringStartIdx = URX_VAL(op); 2701b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UChar32 c = fRXPat->fLiteralText.char32At(stringStartIdx); 2702b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UnicodeSet s(c, c); 2703103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 2704103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // TODO: compute correct set of starting chars for full case folding. 2705103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // For the moment, say any char can start. 2706103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // s.closeOver(USET_CASE_INSENSITIVE); 2707103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius s.clear(); 2708103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius s.complement(); 2709103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 2710b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars->addAll(s); 2711b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru numInitialStrings += 2; // Matching on an initial string not possible. 2712b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2713b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen += stringLen; 2714b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2715b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2716b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2717b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2718b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT: 2719b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT_NG: 2720b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2721b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop Init Ops. These don't change the min length, but they are 4 word ops 2722b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // so location must be updated accordingly. 2723c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Loop Init Ops. 2724b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // If the min loop count == 0 2725b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // move loc forwards to the end of the loop, skipping over the body. 2726c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // If the min count is > 0, 2727b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // continue normal processing of the body of the loop. 272850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t loopEndLoc = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1); 2729b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loopEndLoc = URX_VAL(loopEndLoc); 273050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2); 2731b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (minLoopCount == 0) { 2732b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Min Loop Count of 0, treat like a forward branch and 2733b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // move the current minimum length up to the target 2734b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // (end of loop) location. 2735b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(loopEndLoc <= end+1); 2736b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(loopEndLoc) > currentLen) { 2737b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, loopEndLoc); 2738b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2739c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 2740b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc+=3; // Skips over operands of CTR_INIT 2741b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2742b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2743b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2744b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2745b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2746b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP: 2747b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP_NG: 2748c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Loop ops. 2749b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The jump is conditional, backwards only. 2750b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2751b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2752c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2753b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_C: 2754b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // More loop ops. These state-save to themselves. 2755b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // don't change the minimum match 2756b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru atStart = FALSE; 2757b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2758c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2759b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2760b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_START: 2761b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_START: 2762b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2763b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Look-around. Scan forward until the matching look-ahead end, 2764b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // without processing the look-around block. This is overly pessimistic. 2765fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 2766c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Keep track of the nesting depth of look-around blocks. Boilerplate code for 2767c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // lookahead contains two LA_END instructions, so count goes up by two 2768c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // for each LA_START. 2769c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t depth = (opType == URX_LA_START? 2: 1); 2770b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 2771b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 277250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 2773c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (URX_TYPE(op) == URX_LA_START) { 2774c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru depth+=2; 2775c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 2776c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (URX_TYPE(op) == URX_LB_START) { 2777b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru depth++; 2778b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2779b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) { 2780c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru depth--; 2781b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (depth == 0) { 2782b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2783b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2784b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2785b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(op) == URX_STATE_SAVE) { 2786b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Need this because neg lookahead blocks will FAIL to outside 2787b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the block. 2788b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 2789b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest > loc) { 2790b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen < forwardedLength.elementAti(jmpDest)) { 2791b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 2792b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2793b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2794b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2795c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(loc <= end); 2796b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2797b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2798b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2799c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2800b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_END: 2801b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_CONT: 2802b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_END: 2803b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_CONT: 2804b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_END: 2805c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(FALSE); // Shouldn't get here. These ops should be 2806b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // consumed by the scan in URX_LA_START and LB_START 2807b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2808b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2809c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2810b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 2811b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); 2812b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2813c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2814b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2815b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2816b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2817b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We have finished walking through the ops. Check whether some forward jump 2818b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // propagated a shorter length to location end+1. 2819b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(end+1) < currentLen) { 2820b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(end+1); 2821b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2822b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2823b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2824b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChars8->init(fRXPat->fInitialChars); 2825b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2826b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2827b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Sort out what we should check for when looking for candidate match start positions. 2828b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // In order of preference, 2829b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 1. Start of input text buffer. 2830b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2. A literal string. 2831b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3. Start of line in multi-line mode. 2832b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 4. A single literal character. 2833b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 5. A character from a set of characters. 2834b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 2835b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fRXPat->fStartType == START_START) { 2836b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Match only at the start of an input text string. 2837b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // start type is already set. We're done. 2838b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else if (numInitialStrings == 1 && fRXPat->fMinMatchLen > 0) { 2839b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Match beginning only with a literal string. 2840b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UChar32 c = fRXPat->fLiteralText.char32At(fRXPat->fInitialStringIdx); 2841b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(fRXPat->fInitialChars->contains(c)); 2842b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_STRING; 2843b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChar = c; 2844b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else if (fRXPat->fStartType == START_LINE) { 2845b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Match at start of line in Multi-Line mode. 2846b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Nothing to do here; everything is already set. 2847b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else if (fRXPat->fMinMatchLen == 0) { 2848b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Zero length match possible. We could start anywhere. 2849b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_NO_INFO; 2850b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else if (fRXPat->fInitialChars->size() == 1) { 2851b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // All matches begin with the same char. 2852b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_CHAR; 2853b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fInitialChar = fRXPat->fInitialChars->charAt(0); 2854b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(fRXPat->fInitialChar != (UChar32)-1); 2855b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else if (fRXPat->fInitialChars->contains((UChar32)0, (UChar32)0x10ffff) == FALSE && 2856b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fMinMatchLen > 0) { 2857b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Matches start with a set of character smaller than the set of all chars. 2858b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_SET; 2859b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 2860b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Matches can start with anything 2861b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fStartType = START_NO_INFO; 2862b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2863b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2864b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 2865b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 2866b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2867b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2868b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2869b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2870b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2871b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// minMatchLength Calculate the length of the shortest string that could 2872c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// match the specified pattern. 2873b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Length is in 16 bit code units, not code points. 2874b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2875b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The calculated length may not be exact. The returned 2876b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// value may be shorter than the actual minimum; it must 2877b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// never be longer. 2878b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2879b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// start and end are the range of p-code operations to be 2880b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// examined. The endpoints are included in the range. 2881b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 2882b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 2883b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruint32_t RegexCompile::minMatchLength(int32_t start, int32_t end) { 2884b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 2885b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return 0; 2886b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2887b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2888b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(start <= end); 2889b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(end < fRXPat->fCompiledPat->size()); 2890b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2891b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2892b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loc; 2893b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op; 2894b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t opType; 2895b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t currentLen = 0; 2896b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2897b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2898b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // forwardedLength is a vector holding minimum-match-length values that 2899b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // are propagated forward in the pattern by JMP or STATE_SAVE operations. 2900b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // It must be one longer than the pattern being checked because some ops 2901b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // will jmp to a end-of-block+1 location from within a block, and we must 2902b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // count those when checking the block. 2903b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UVector32 forwardedLength(end+2, *fStatus); 2904b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setSize(end+2); 2905b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc=start; loc<=end+1; loc++) { 2906b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(INT32_MAX, loc); 2907b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2908b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2909b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc = start; loc<=end; loc++) { 291050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 2911b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType = URX_TYPE(op); 2912b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2913b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The loop is advancing linearly through the pattern. 2914b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // If the op we are now at was the destination of a branch in the pattern, 2915b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // and that path has a shorter minimum length than the current accumulated value, 2916b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // replace the current accumulated value. 2917c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); // MinLength == INT32_MAX for some 2918c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // no-match-possible cases. 2919b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(loc) < currentLen) { 2920b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc); 2921b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); 2922b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2923b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2924b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (opType) { 2925b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Ops that don't change the total length matched 2926b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RESERVED_OP: 2927b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END: 2928b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING_LEN: 2929b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_NOP: 2930b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_START_CAPTURE: 2931b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END_CAPTURE: 2932b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_B: 2933b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_BU: 2934b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_G: 2935b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_Z: 2936b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET: 2937b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOLLAR: 2938c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_M: 2939c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_D: 2940c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_MD: 2941b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RELOC_OPRND: 2942b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_INP_LOC: 2943b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET_M: 2944c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_CARET_M_UNIX: 2945b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKREF: // BackRef. Must assume that it might be a zero length match 2946b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKREF_I: 2947b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2948b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. 2949b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LD_SP: 2950b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2951b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV: 2952b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV_X: 2953b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2954c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 2955b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2956b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Ops that match a minimum of one character (one or two 16 bit code units.) 2957c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 2958b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR: 2959b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATIC_SETREF: 2960b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STAT_SETREF_N: 2961b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_SETREF: 2962b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_D: 2963b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR_I: 2964b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_X: // Grahpeme Cluster. Minimum is 1, max unbounded. 2965b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY_ALL: // . matches one or two. 2966b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY: 2967c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOTANY_UNIX: 2968b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen++; 2969b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2970b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2971b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2972b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMPX: 2973b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; // URX_JMPX has an extra operand, ignored here, 2974b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // otherwise processed identically to URX_JMP. 2975b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP: 2976b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2977b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 2978b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest < loc) { 2979b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop of some kind. Can safely ignore, the worst that will happen 2980b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // is that we understate the true minimum length 2981b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc+1); 2982b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 2983b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Forward jump. Propagate the current min length to the target loc of the jump. 2984b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(jmpDest <= end+1); 2985b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(jmpDest) > currentLen) { 2986b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 2987b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2988b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2989b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2990b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2991b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 2992c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_BACKTRACK: 2993b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 2994c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Back-tracks are kind of like a branch, except that the min length was 2995b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // propagated already, by the state save. 2996b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc+1); 2997b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 2998b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 2999b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3000b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3001b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATE_SAVE: 3002b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3003b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // State Save, for forward jumps, propagate the current minimum. 3004b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the state save. 3005b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 3006b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest > loc) { 3007b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen < forwardedLength.elementAti(jmpDest)) { 3008b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 3009b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3010c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3011b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3012b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3013c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3014b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3015b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING: 3016b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3017b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 301850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3019b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen += URX_VAL(stringLenOp); 3020b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3021b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3022b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3023b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3024103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius case URX_STRING_I: 3025103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius { 3026103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius loc++; 3027103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // TODO: with full case folding, matching input text may be shorter than 3028103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // the string we have here. More smarts could put some bounds on it. 3029103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Assume a min length of one for now. A min length of zero causes 3030103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // optimization failures for a pattern like "string"+ 3031103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // currentLen += URX_VAL(stringLenOp); 3032103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius currentLen += 1; 3033103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 3034103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius break; 3035103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 3036b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT: 3037b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT_NG: 3038b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3039c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Loop Init Ops. 3040b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // If the min loop count == 0 3041b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // move loc forwards to the end of the loop, skipping over the body. 3042c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // If the min count is > 0, 3043b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // continue normal processing of the body of the loop. 304450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t loopEndLoc = (int32_t)fRXPat->fCompiledPat->elementAti(loc+1); 3045b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loopEndLoc = URX_VAL(loopEndLoc); 304650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t minLoopCount = (int32_t)fRXPat->fCompiledPat->elementAti(loc+2); 3047b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (minLoopCount == 0) { 3048b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc = loopEndLoc; 3049b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 3050b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc+=3; // Skips over operands of CTR_INIT 3051b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3052b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3053b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3054b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3055b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3056b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP: 3057b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP_NG: 3058c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Loop ops. 3059b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The jump is conditional, backwards only. 3060b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3061c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3062b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_SR_I: 3063b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_DOT_I: 3064b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_C: 3065b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // More loop ops. These state-save to themselves. 3066b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // don't change the minimum match - could match nothing at all. 3067b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3068c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3069b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3070b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_START: 3071b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_START: 3072b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3073b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Look-around. Scan forward until the matching look-ahead end, 3074c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // without processing the look-around block. This is overly pessimistic for look-ahead, 3075c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // it assumes that the look-ahead match might be zero-length. 3076b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // TODO: Positive lookahead could recursively do the block, then continue 3077c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // with the longer of the block or the value coming in. Ticket 6060 3078c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t depth = (opType == URX_LA_START? 2: 1);; 3079b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 3080b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 308150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3082c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (URX_TYPE(op) == URX_LA_START) { 3083c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The boilerplate for look-ahead includes two LA_END insturctions, 3084c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Depth will be decremented by each one when it is seen. 3085c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru depth += 2; 3086c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3087c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (URX_TYPE(op) == URX_LB_START) { 3088b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru depth++; 3089b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3090c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (URX_TYPE(op) == URX_LA_END) { 3091c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru depth--; 3092b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (depth == 0) { 3093b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3094b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3095c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3096c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (URX_TYPE(op)==URX_LBN_END) { 3097b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru depth--; 3098c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (depth == 0) { 3099c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 3100c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3101b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3102b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(op) == URX_STATE_SAVE) { 3103b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Need this because neg lookahead blocks will FAIL to outside 3104b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the block. 3105b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 3106b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest > loc) { 3107b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen < forwardedLength.elementAti(jmpDest)) { 3108b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 3109b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3110b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3111b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3112c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(loc <= end); 3113b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3114b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3115b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3116c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3117b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_END: 3118b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_CONT: 3119b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_END: 3120b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_CONT: 3121b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_END: 3122b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Only come here if the matching URX_LA_START or URX_LB_START was not in the 3123b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // range being sized, which happens when measuring size of look-behind blocks. 3124b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3125c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3126b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 3127b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); 3128b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3129c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3130b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3131b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3132b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We have finished walking through the ops. Check whether some forward jump 3133b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // propagated a shorter length to location end+1. 3134b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(end+1) < currentLen) { 3135b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(end+1); 3136b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(currentLen>=0 && currentLen < INT32_MAX); 3137b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3138c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3139b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return currentLen; 3140b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3141b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3142103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius// Increment with overflow check. 3143103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius// val and delta will both be positive. 3144103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 3145103e9ffba2cba345d0078eb8b8db33249f81840aCraig Corneliusstatic int32_t safeIncrement(int32_t val, int32_t delta) { 3146103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius if (INT32_MAX - val > delta) { 3147103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius return val + delta; 3148103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } else { 3149103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius return INT32_MAX; 3150103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 3151103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius} 3152b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3153b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3154b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3155b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3156b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// maxMatchLength Calculate the length of the longest string that could 3157c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// match the specified pattern. 3158b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Length is in 16 bit code units, not code points. 3159b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3160b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The calculated length may not be exact. The returned 3161b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// value may be longer than the actual maximum; it must 3162b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// never be shorter. 3163b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3164b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3165b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruint32_t RegexCompile::maxMatchLength(int32_t start, int32_t end) { 3166b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 3167b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return 0; 3168b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3169b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(start <= end); 3170b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(end < fRXPat->fCompiledPat->size()); 3171b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3172b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3173b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loc; 3174b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t op; 3175b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t opType; 3176b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t currentLen = 0; 3177b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UVector32 forwardedLength(end+1, *fStatus); 3178b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setSize(end+1); 3179b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3180b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc=start; loc<=end; loc++) { 3181b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(0, loc); 3182b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3183b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3184b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc = start; loc<=end; loc++) { 318550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3186b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru opType = URX_TYPE(op); 3187b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3188b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The loop is advancing linearly through the pattern. 3189b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // If the op we are now at was the destination of a branch in the pattern, 3190b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // and that path has a longer maximum length than the current accumulated value, 3191b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // replace the current accumulated value. 3192b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(loc) > currentLen) { 3193b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc); 3194b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3195b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3196b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (opType) { 3197b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Ops that don't change the total length matched 3198b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RESERVED_OP: 3199b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END: 3200b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING_LEN: 3201b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_NOP: 3202b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_START_CAPTURE: 3203b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END_CAPTURE: 3204b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_B: 3205b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_BU: 3206b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_G: 3207b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_Z: 3208b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET: 3209b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOLLAR: 3210c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_M: 3211c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_D: 3212c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_MD: 3213b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RELOC_OPRND: 3214b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_INP_LOC: 3215b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET_M: 3216c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_CARET_M_UNIX: 3217b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3218b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_SP: // Setup for atomic or possessive blocks. Doesn't change what can match. 3219b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LD_SP: 3220b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3221b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_END: 3222b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_CONT: 3223b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_CONT: 3224b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_END: 3225b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3226c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3227b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3228b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Ops that increase that cause an unbounded increase in the length 3229b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of a matched string, or that increase it a hard to characterize way. 3230b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Call the max length unbounded, and stop further checking. 3231b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKREF: // BackRef. Must assume that it might be a zero length match 3232b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKREF_I: 3233b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_X: // Grahpeme Cluster. Minimum is 1, max unbounded. 3234b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = INT32_MAX; 3235b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3236b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3237b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3238b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Ops that match a max of one character (possibly two 16 bit code units.) 3239c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 3240b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATIC_SETREF: 3241b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STAT_SETREF_N: 3242b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_SETREF: 3243b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_D: 3244b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR_I: 3245c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOTANY_ALL: 3246b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY: 3247c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOTANY_UNIX: 3248103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius currentLen = safeIncrement(currentLen, 2); 3249b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3250b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3251b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Single literal character. Increase current max length by one or two, 3252b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // depending on whether the char is in the supplementary range. 3253b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR: 3254103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius currentLen = safeIncrement(currentLen, 1); 3255b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_VAL(op) > 0x10000) { 3256103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius currentLen = safeIncrement(currentLen, 1); 3257b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3258b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3259b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3260c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Jumps. 3261b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3262b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP: 3263b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMPX: 3264b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV: 3265b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV_X: 3266b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3267b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 3268b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest < loc) { 3269b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Loop of some kind. Max match length is unbounded. 3270b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = INT32_MAX; 3271b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 3272b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Forward jump. Propagate the current min length to the target loc of the jump. 3273b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (forwardedLength.elementAti(jmpDest) < currentLen) { 3274b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 3275b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3276b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = 0; 3277b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3278b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3279b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3280b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3281c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_BACKTRACK: 3282c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // back-tracks are kind of like a branch, except that the max length was 3283b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // propagated already, by the state save. 3284b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = forwardedLength.elementAti(loc+1); 3285b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3286b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3287b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3288b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATE_SAVE: 3289b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3290b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // State Save, for forward jumps, propagate the current minimum. 3291b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // of the state save. 3292b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // For backwards jumps, they create a loop, maximum 3293b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // match length is unbounded. 3294b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t jmpDest = URX_VAL(op); 3295b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (jmpDest > loc) { 3296b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen > forwardedLength.elementAti(jmpDest)) { 3297b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru forwardedLength.setElementAt(currentLen, jmpDest); 3298b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3299b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } else { 3300b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = INT32_MAX; 3301b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3302b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3303b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3304c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3305b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3306b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3307b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3308b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING: 3309103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius { 3310103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius loc++; 3311103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3312103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp)); 3313103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius break; 3314103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius } 3315103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius 3316b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING_I: 3317103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // TODO: This code assumes that any user string that matches will be no longer 3318103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // than our compiled string, with case insensitive matching. 3319103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Our compiled string has been case-folded already. 3320103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // 3321103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // Any matching user string will have no more code points than our 3322103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // compiled (folded) string. Folding may add code points, but 3323103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // not remove them. 3324103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // 3325fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius // There is a potential problem if a supplemental code point 3326103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // case-folds to a BMP code point. In this case our compiled string 3327103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // could be shorter (in code units) than a matching user string. 3328103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // 3329103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // At this time (Unicode 6.1) there are no such characters, and this case 3330103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // is not being handled. A test, intltest regex/Bug9283, will fail if 3331103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // any problematic characters are added to Unicode. 3332103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // 3333103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // If this happens, we can make a set of the BMP chars that the 3334103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // troublesome supplementals fold to, scan our string, and bump the 3335103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // currentLen one extra for each that is found. 3336103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius // 3337b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3338b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 333950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3340103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp)); 3341b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3342b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3343b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3344b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT: 3345b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT_NG: 334659d709d503bab6e2b61931737e662dd293b40578ccornelius // For Loops, recursively call this function on the pattern for the loop body, 334759d709d503bab6e2b61931737e662dd293b40578ccornelius // then multiply the result by the maximum loop count. 334859d709d503bab6e2b61931737e662dd293b40578ccornelius { 334959d709d503bab6e2b61931737e662dd293b40578ccornelius int32_t loopEndLoc = URX_VAL(fRXPat->fCompiledPat->elementAti(loc+1)); 335059d709d503bab6e2b61931737e662dd293b40578ccornelius if (loopEndLoc == loc+4) { 335159d709d503bab6e2b61931737e662dd293b40578ccornelius // Loop has an empty body. No affect on max match length. 335259d709d503bab6e2b61931737e662dd293b40578ccornelius // Continue processing with code after the loop end. 335359d709d503bab6e2b61931737e662dd293b40578ccornelius loc = loopEndLoc; 335459d709d503bab6e2b61931737e662dd293b40578ccornelius break; 335559d709d503bab6e2b61931737e662dd293b40578ccornelius } 3356fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 335759d709d503bab6e2b61931737e662dd293b40578ccornelius int32_t maxLoopCount = fRXPat->fCompiledPat->elementAti(loc+3); 335859d709d503bab6e2b61931737e662dd293b40578ccornelius if (maxLoopCount == -1) { 335959d709d503bab6e2b61931737e662dd293b40578ccornelius // Unbounded Loop. No upper bound on match length. 336059d709d503bab6e2b61931737e662dd293b40578ccornelius currentLen = INT32_MAX; 336159d709d503bab6e2b61931737e662dd293b40578ccornelius break; 336259d709d503bab6e2b61931737e662dd293b40578ccornelius } 336359d709d503bab6e2b61931737e662dd293b40578ccornelius 336459d709d503bab6e2b61931737e662dd293b40578ccornelius U_ASSERT(loopEndLoc >= loc+4); 336559d709d503bab6e2b61931737e662dd293b40578ccornelius int32_t blockLen = maxMatchLength(loc+4, loopEndLoc-1); // Recursive call. 336659d709d503bab6e2b61931737e662dd293b40578ccornelius if (blockLen == INT32_MAX) { 336759d709d503bab6e2b61931737e662dd293b40578ccornelius currentLen = blockLen; 336859d709d503bab6e2b61931737e662dd293b40578ccornelius break; 336959d709d503bab6e2b61931737e662dd293b40578ccornelius } 337059d709d503bab6e2b61931737e662dd293b40578ccornelius currentLen += blockLen * maxLoopCount; 337159d709d503bab6e2b61931737e662dd293b40578ccornelius loc = loopEndLoc; 337259d709d503bab6e2b61931737e662dd293b40578ccornelius break; 337359d709d503bab6e2b61931737e662dd293b40578ccornelius } 337459d709d503bab6e2b61931737e662dd293b40578ccornelius 3375b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP: 3376b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP_NG: 337759d709d503bab6e2b61931737e662dd293b40578ccornelius // These opcodes will be skipped over by code for URX_CRT_INIT. 337859d709d503bab6e2b61931737e662dd293b40578ccornelius // We shouldn't encounter them here. 337959d709d503bab6e2b61931737e662dd293b40578ccornelius U_ASSERT(FALSE); 338059d709d503bab6e2b61931737e662dd293b40578ccornelius break; 338159d709d503bab6e2b61931737e662dd293b40578ccornelius 3382b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_SR_I: 3383b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_DOT_I: 3384b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_C: 3385b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // For anything to do with loops, make the match length unbounded. 3386b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru currentLen = INT32_MAX; 3387b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3388c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3389c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3390b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3391b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_START: 3392b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_END: 3393b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Look-ahead. Just ignore, treat the look-ahead block as if 3394b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // it were normal pattern. Gives a too-long match length, 3395b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // but good enough for now. 3396b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3397c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3398b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // End of look-ahead ops should always be consumed by the processing at 3399b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the URX_LA_START op. 3400b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // U_ASSERT(FALSE); 3401b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // break; 3402c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3403b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_START: 3404b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3405b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Look-behind. Scan forward until the matching look-around end, 3406c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // without processing the look-behind block. 3407b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t depth = 0; 3408b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 3409b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru loc++; 341050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3411b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(op) == URX_LA_START || URX_TYPE(op) == URX_LB_START) { 3412b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru depth++; 3413b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3414b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(op) == URX_LA_END || URX_TYPE(op)==URX_LBN_END) { 3415b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (depth == 0) { 3416b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3417b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3418b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru depth--; 3419b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3420c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(loc < end); 3421b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3422b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3423b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3424b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3425b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 3426b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); 3427b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3428b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3429c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3430b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (currentLen == INT32_MAX) { 3431b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The maximum length is unbounded. 3432b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Stop further processing of the pattern. 3433b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3434b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3435c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3436b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3437b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return currentLen; 3438c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3439b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3440b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3441b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3442b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3443b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3444b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// stripNOPs Remove any NOP operations from the compiled pattern code. 3445b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Extra NOPs are inserted for some constructs during the initial 3446b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// code generation to provide locations that may be patched later. 3447b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Many end up unneeded, and are removed by this function. 3448b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 344950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// In order to minimize the number of passes through the pattern, 345050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// back-reference fixup is also performed here (adjusting 345150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// back-reference operands to point to the correct frame offsets). 345250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho// 3453b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3454b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::stripNOPs() { 3455b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3456b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 3457b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return; 3458b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3459b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3460b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t end = fRXPat->fCompiledPat->size(); 3461b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UVector32 deltas(end, *fStatus); 3462b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3463b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Make a first pass over the code, computing the amount that things 3464b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // will be offset at each location in the original code. 3465b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t loc; 3466b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t d = 0; 3467b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (loc=0; loc<end; loc++) { 3468b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru deltas.addElement(d, *fStatus); 346950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(loc); 3470b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (URX_TYPE(op) == URX_NOP) { 3471b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru d++; 3472b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3473b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3474fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 347550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UnicodeString caseStringBuffer; 3476b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3477b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Make a second pass over the code, removing the NOPs by moving following 3478b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // code up, and patching operands that refer to code locations that 3479b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // are being moved. The array of offsets from the first step is used 3480b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // to compute the new operand values. 3481b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t src; 3482b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t dst = 0; 3483b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (src=0; src<end; src++) { 348450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t op = (int32_t)fRXPat->fCompiledPat->elementAti(src); 3485b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t opType = URX_TYPE(op); 3486b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru switch (opType) { 3487b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_NOP: 3488b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3489b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3490b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATE_SAVE: 3491b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP: 3492b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP: 3493b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_LOOP_NG: 3494b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RELOC_OPRND: 3495b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMPX: 3496b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV: 3497b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_JMP_SAV_X: 3498b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // These are instructions with operands that refer to code locations. 3499b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3500b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t operandAddress = URX_VAL(op); 3501b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(operandAddress>=0 && operandAddress<deltas.size()); 3502b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru int32_t fixedOperandAddress = operandAddress - deltas.elementAti(operandAddress); 3503b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru op = URX_BUILD(opType, fixedOperandAddress); 3504b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, dst); 3505b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru dst++; 3506b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3507b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3508b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 350950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho case URX_BACKREF: 351050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho case URX_BACKREF_I: 351150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho { 351250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t where = URX_VAL(op); 351350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (where > fRXPat->fGroupMap->size()) { 351450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho error(U_REGEX_INVALID_BACK_REF); 351550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho break; 351650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } 351750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho where = fRXPat->fGroupMap->elementAti(where-1); 351850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho op = URX_BUILD(opType, where); 351950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fRXPat->fCompiledPat->setElementAt(op, dst); 352050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho dst++; 3521fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 352250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fRXPat->fNeedsAltInput = TRUE; 352350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho break; 352450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } 3525b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RESERVED_OP: 3526b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_RESERVED_OP_N: 3527b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKTRACK: 3528b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END: 3529b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_ONECHAR: 3530103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius case URX_STRING: 3531b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STRING_LEN: 3532b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_START_CAPTURE: 3533b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_END_CAPTURE: 3534b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STATIC_SETREF: 3535b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STAT_SETREF_N: 3536b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_SETREF: 3537b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY: 3538b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_FAIL: 3539b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_B: 3540b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_BU: 3541b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_G: 3542b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_X: 3543b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_Z: 3544b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOTANY_ALL: 3545b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_BACKSLASH_D: 3546b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET: 3547b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOLLAR: 3548b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT: 3549b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CTR_INIT_NG: 3550c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOTANY_UNIX: 3551b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_SP: 3552b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LD_SP: 3553b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_STO_INP_LOC: 3554b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_START: 3555b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LA_END: 3556103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius case URX_ONECHAR_I: 3557103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius case URX_STRING_I: 3558b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_DOLLAR_M: 3559b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_CARET_M: 3560c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_CARET_M_UNIX: 3561b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_START: 3562b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_CONT: 3563b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LB_END: 3564b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_CONT: 3565b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LBN_END: 3566b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_SR_I: 3567b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_DOT_I: 3568b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru case URX_LOOP_C: 3569c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_D: 3570c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case URX_DOLLAR_MD: 3571b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // These instructions are unaltered by the relocation. 3572b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setElementAt(op, dst); 3573b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru dst++; 3574b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3575b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3576b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru default: 3577b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Some op is unaccounted for. 3578b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru U_ASSERT(FALSE); 3579b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_INTERNAL_ERROR); 3580b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3581b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3582b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3583b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fRXPat->fCompiledPat->setSize(dst); 3584b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3585b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3586b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3587b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3588b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3589b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3590b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3591b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Error Report a rule parse error. 3592b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Only report it if no previous error has been recorded. 3593b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3594b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3595b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::error(UErrorCode e) { 3596b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_SUCCESS(*fStatus)) { 3597b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru *fStatus = e; 359850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // Hmm. fParseErr (UParseError) line & offset fields are int32_t in public 359950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // API (see common/unicode/parseerr.h), while fLineNum and fCharNum are 360050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // int64_t. If the values of the latter are out of range for the former, 360150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // set them to the appropriate "field not supported" values. 360250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (fLineNum > 0x7FFFFFFF) { 360350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fParseErr->line = 0; 360450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fParseErr->offset = -1; 360550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } else if (fCharNum > 0x7FFFFFFF) { 360650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fParseErr->line = (int32_t)fLineNum; 360750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fParseErr->offset = -1; 360850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } else { 360950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fParseErr->line = (int32_t)fLineNum; 361050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fParseErr->offset = (int32_t)fCharNum; 361150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } 3612fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 361350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UErrorCode status = U_ZERO_ERROR; // throwaway status for extracting context 3614b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3615b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Fill in the context. 3616b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Note: extractBetween() pins supplied indicies to the string bounds. 3617b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru uprv_memset(fParseErr->preContext, 0, sizeof(fParseErr->preContext)); 3618b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru uprv_memset(fParseErr->postContext, 0, sizeof(fParseErr->postContext)); 361950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho utext_extract(fRXPat->fPattern, fScanIndex-U_PARSE_CONTEXT_LEN+1, fScanIndex, fParseErr->preContext, U_PARSE_CONTEXT_LEN, &status); 362050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho utext_extract(fRXPat->fPattern, fScanIndex, fScanIndex+U_PARSE_CONTEXT_LEN-1, fParseErr->postContext, U_PARSE_CONTEXT_LEN, &status); 3621b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3622b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3623b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3624b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3625b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3626b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Assorted Unicode character constants. 3627b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Numeric because there is no portable way to enter them as literals. 3628b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// (Think EBCDIC). 3629b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3630b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Querustatic const UChar chCR = 0x0d; // New lines, for terminating comments. 3631c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chLF = 0x0a; // Line Feed 3632b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Querustatic const UChar chPound = 0x23; // '#', introduces a comment. 3633c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chDigit0 = 0x30; // '0' 3634c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chDigit7 = 0x37; // '9' 3635c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chColon = 0x3A; // ':' 3636b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Querustatic const UChar chE = 0x45; // 'E' 3637c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chQ = 0x51; // 'Q' 363854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius//static const UChar chN = 0x4E; // 'N' 3639c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chP = 0x50; // 'P' 3640b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Querustatic const UChar chBackSlash = 0x5c; // '\' introduces a char escape 364154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius//static const UChar chLBracket = 0x5b; // '[' 3642c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chRBracket = 0x5d; // ']' 3643c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chUp = 0x5e; // '^' 3644c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chLowerP = 0x70; 3645c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chLBrace = 0x7b; // '{' 3646c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chRBrace = 0x7d; // '}' 3647c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chNEL = 0x85; // NEL newline variant 3648c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar chLS = 0x2028; // Unicode Line Separator 3649b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3650b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3651b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3652b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3653b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// nextCharLL Low Level Next Char from the regex pattern. 3654b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Get a char from the string, keep track of input position 3655b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// for error reporting. 3656b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3657b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3658b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruUChar32 RegexCompile::nextCharLL() { 3659b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UChar32 ch; 3660b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3661b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fPeekChar != -1) { 3662b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru ch = fPeekChar; 3663b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fPeekChar = -1; 3664b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return ch; 3665b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3666fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 366750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // assume we're already in the right place 366850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho ch = UTEXT_NEXT32(fRXPat->fPattern); 366950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (ch == U_SENTINEL) { 367050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho return ch; 3671b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3672b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3673b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (ch == chCR || 3674b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru ch == chNEL || 3675b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru ch == chLS || 367627f654740f2a26ad62a5c155af9199af9e69b889claireho (ch == chLF && fLastChar != chCR)) { 3677b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Character is starting a new line. Bump up the line number, and 3678b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // reset the column to 0. 3679b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fLineNum++; 3680b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fCharNum=0; 3681b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3682b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru else { 3683b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Character is not starting a new line. Except in the case of a 3684b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // LF following a CR, increment the column position. 3685b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (ch != chLF) { 3686b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fCharNum++; 3687b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3688b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3689b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fLastChar = ch; 3690b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return ch; 3691b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3692b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3693b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3694b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3695b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// peekCharLL Low Level Character Scanning, sneak a peek at the next 3696b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// character without actually getting it. 3697b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3698b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3699b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruUChar32 RegexCompile::peekCharLL() { 3700b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fPeekChar == -1) { 3701b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fPeekChar = nextCharLL(); 3702b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3703b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return fPeekChar; 3704b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3705b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3706b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3707b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3708b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3709b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// nextChar for pattern scanning. At this level, we handle stripping 3710b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// out comments and processing some backslash character escapes. 3711b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The rest of the pattern grammar is handled at the next level up. 3712b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3713b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3714b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queruvoid RegexCompile::nextChar(RegexPatternChar &c) { 3715b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 371650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fScanIndex = UTEXT_GETNATIVEINDEX(fRXPat->fPattern); 3717b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar = nextCharLL(); 3718b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fQuoted = FALSE; 3719b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3720b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fQuoteMode) { 3721b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fQuoted = TRUE; 3722fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius if ((c.fChar==chBackSlash && peekCharLL()==chE && ((fModeFlags & UREGEX_LITERAL) == 0)) || 3723103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius c.fChar == (UChar32)-1) { 3724b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fQuoteMode = FALSE; // Exit quote mode, 3725103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius nextCharLL(); // discard the E 3726103e9ffba2cba345d0078eb8b8db33249f81840aCraig Cornelius nextChar(c); // recurse to get the real next char 3727b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3728b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3729b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru else if (fInBackslashQuote) { 3730b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // The current character immediately follows a '\' 3731b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Don't check for any further escapes, just return it as-is. 3732b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Don't set c.fQuoted, because that would prevent the state machine from 3733b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // dispatching on the character. 3734b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fInBackslashQuote = FALSE; 3735b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3736b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru else 3737b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3738b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We are not in a \Q quoted region \E of the source. 3739b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3740b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fModeFlags & UREGEX_COMMENTS) { 3741b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3742b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We are in free-spacing and comments mode. 3743c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Scan through any white space and comments, until we 3744b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // reach a significant character or the end of inut. 3745b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 3746b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (c.fChar == (UChar32)-1) { 3747b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; // End of Input 3748b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3749b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (c.fChar == chPound && fEOLComments == TRUE) { 3750b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Start of a comment. Consume the rest of it, until EOF or a new line 3751b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 3752b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar = nextCharLL(); 3753b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (c.fChar == (UChar32)-1 || // EOF 3754b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar == chCR || 3755b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar == chLF || 3756b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar == chNEL || 3757b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar == chLS) { 3758b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3759b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3760b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3761b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3762c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // TODO: check what Java & Perl do with non-ASCII white spaces. Ticket 6061. 3763b26ce3a7367e4ed2ee7ddddcdc3f3d3377a455c2claireho if (PatternProps::isWhiteSpace(c.fChar) == FALSE) { 3764b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3765b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3766b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fChar = nextCharLL(); 3767b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3768b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3769b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3770b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3771b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // check for backslash escaped characters. 3772b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3773b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (c.fChar == chBackSlash) { 377450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int64_t pos = UTEXT_GETNATIVEINDEX(fRXPat->fPattern); 3775b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (RegexStaticSets::gStaticSets->fUnescapeCharSet.contains(peekCharLL())) { 3776b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3777b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // A '\' sequence that is handled by ICU's standard unescapeAt function. 3778b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Includes \uxxxx, \n, \r, many others. 3779b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Return the single equivalent character. 3780b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // 3781b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru nextCharLL(); // get & discard the peeked char. 3782b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru c.fQuoted = TRUE; 3783fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 378450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (UTEXT_FULL_TEXT_IN_CHUNK(fRXPat->fPattern, fPatternLength)) { 378550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t endIndex = (int32_t)pos; 378650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho c.fChar = u_unescapeAt(uregex_ucstr_unescape_charAt, &endIndex, (int32_t)fPatternLength, (void *)fRXPat->fPattern->chunkContents); 3787fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 378850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (endIndex == pos) { 378950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho error(U_REGEX_BAD_ESCAPE_SEQUENCE); 379050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } 379150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fCharNum += endIndex - pos; 379250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UTEXT_SETNATIVEINDEX(fRXPat->fPattern, endIndex); 379350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } else { 379450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int32_t offset = 0; 379550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho struct URegexUTextUnescapeCharContext context = U_REGEX_UTEXT_UNESCAPE_CONTEXT(fRXPat->fPattern); 3796fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 379750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UTEXT_SETNATIVEINDEX(fRXPat->fPattern, pos); 379850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho c.fChar = u_unescapeAt(uregex_utext_unescape_charAt, &offset, INT32_MAX, &context); 379950294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho 380050294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho if (offset == 0) { 380150294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho error(U_REGEX_BAD_ESCAPE_SEQUENCE); 380250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } else if (context.lastOffset == offset) { 380350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UTEXT_PREVIOUS32(fRXPat->fPattern); 380450294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } else if (context.lastOffset != offset-1) { 380550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho utext_moveIndex32(fRXPat->fPattern, offset - context.lastOffset - 1); 380650294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho } 380750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho fCharNum += offset; 3808b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3809b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3810c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (peekCharLL() == chDigit0) { 3811c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Octal Escape, using Java Regexp Conventions 3812c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // which are \0 followed by 1-3 octal digits. 3813c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Different from ICU Unescape handling of Octal, which does not 3814c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // require the leading 0. 381550294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho // Java also has the convention of only consuming 2 octal digits if 3816c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // the three digit number would be > 0xff 3817c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 3818c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru c.fChar = 0; 3819c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextCharLL(); // Consume the initial 0. 3820c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int index; 3821c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru for (index=0; index<3; index++) { 3822c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t ch = peekCharLL(); 3823c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (ch<chDigit0 || ch>chDigit7) { 3824c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (index==0) { 3825c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // \0 is not followed by any octal digits. 3826c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_BAD_ESCAPE_SEQUENCE); 3827c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3828c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 3829c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3830c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru c.fChar <<= 3; 3831c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru c.fChar += ch&7; 3832c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (c.fChar <= 255) { 3833c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextCharLL(); 3834c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else { 3835c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The last digit made the number too big. Forget we saw it. 3836c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru c.fChar >>= 3; 3837c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3838c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3839fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius c.fQuoted = TRUE; 3840fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius } 3841c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (peekCharLL() == chQ) { 3842c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // "\Q" enter quote mode, which will continue until "\E" 3843c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fQuoteMode = TRUE; 3844c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextCharLL(); // discard the 'Q'. 3845c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(c); // recurse to get the real next char. 3846c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3847b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru else 3848b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru { 3849b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // We are in a '\' escape that will be handled by the state table scanner. 3850b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Just return the backslash, but remember that the following char is to 3851c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // be taken literally. 3852b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fInBackslashQuote = TRUE; 3853b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3854b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3855b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3856b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3857b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // re-enable # to end-of-line comments, in case they were disabled. 3858b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // They are disabled by the parser upon seeing '(?', but this lasts for 3859b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // the fetching of the next character only. 3860b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru fEOLComments = TRUE; 3861b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3862b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // putc(c.fChar, stdout); 3863b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 3864b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3865b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3866b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3867b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3868b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3869c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// scanNamedChar 3870c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Get a UChar32 from a \N{UNICODE CHARACTER NAME} in the pattern. 3871c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 3872c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// The scan position will be at the 'N'. On return 3873c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// the scan position should be just after the '}' 3874b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3875c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// Return the UChar32 3876b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3877b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3878c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste QueruUChar32 RegexCompile::scanNamedChar() { 3879b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 3880c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return 0; 3881b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3882b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3883c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 3884c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar != chLBrace) { 3885c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_PROPERTY_SYNTAX); 3886c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return 0; 3887b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3888fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 3889c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeString charName; 3890b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 3891c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 3892c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar == chRBrace) { 3893b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3894b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3895c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar == -1) { 3896c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_PROPERTY_SYNTAX); 3897c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return 0; 3898c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3899c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru charName.append(fC.fChar); 3900b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3901fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 3902c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru char name[100]; 3903c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (!uprv_isInvariantUString(charName.getBuffer(), charName.length()) || 3904c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru (uint32_t)charName.length()>=sizeof(name)) { 3905c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // All Unicode character names have only invariant characters. 3906c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The API to get a character, given a name, accepts only char *, forcing us to convert, 3907c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // which requires this error check 3908c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_PROPERTY_SYNTAX); 3909c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return 0; 3910c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3911c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru charName.extract(0, charName.length(), name, sizeof(name), US_INV); 3912b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3913c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UChar32 theChar = u_charFromName(U_UNICODE_CHAR_NAME, name, fStatus); 3914c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 3915c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_PROPERTY_SYNTAX); 3916c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3917b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3918c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); // Continue overall regex pattern processing with char after the '}' 3919c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return theChar; 3920c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru} 3921b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3922b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3923b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3924b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// scanProp Construct a UnicodeSet from the text at the current scan 3925b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// position, which will be of the form \p{whaterver} 3926b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3927b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// The scan position will be at the 'p' or 'P'. On return 3928b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// the scan position should be just after the '}' 3929b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3930b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// Return a UnicodeSet, constructed from the \P pattern, 3931b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// or NULL if the pattern is invalid. 3932b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru// 3933b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru//------------------------------------------------------------------------------ 3934b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruUnicodeSet *RegexCompile::scanProp() { 3935b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru UnicodeSet *uset = NULL; 3936b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3937b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 3938b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return NULL; 3939b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3940fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius (void)chLowerP; // Suppress compiler unused variable warning. 3941c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(fC.fChar == chLowerP || fC.fChar == chP); 3942c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UBool negated = (fC.fChar == chP); 3943b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3944c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeString propertyName; 3945c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 3946c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar != chLBrace) { 3947c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(U_REGEX_PROPERTY_SYNTAX); 3948c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return NULL; 3949c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3950b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru for (;;) { 3951c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 3952b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fC.fChar == chRBrace) { 3953b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru break; 3954b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3955b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fC.fChar == -1) { 3956b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru // Hit the end of the input string without finding the closing '}' 3957b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru error(U_REGEX_PROPERTY_SYNTAX); 3958b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru return NULL; 3959b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3960c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru propertyName.append(fC.fChar); 3961c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 3962c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru uset = createSetForProperty(propertyName, negated); 3963c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); // Move input scan to position following the closing '}' 3964c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return uset; 3965c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru} 3966c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3967c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru//------------------------------------------------------------------------------ 3968c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 3969c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// scanPosixProp Construct a UnicodeSet from the text at the current scan 3970c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// position, which is expected be of the form [:property expression:] 3971c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 3972c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// The scan position will be at the opening ':'. On return 3973c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// the scan position must be on the closing ']' 3974c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 3975c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// Return a UnicodeSet constructed from the pattern, 3976c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// or NULL if this is not a valid POSIX-style set expression. 3977c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// If not a property expression, restore the initial scan position 3978c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// (to the opening ':') 3979c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 3980c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// Note: the opening '[:' is not sufficient to guarantee that 3981c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// this is a [:property:] expression. 3982c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// [:'+=,] is a perfectly good ordinary set expression that 3983c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// happens to include ':' as one of its characters. 3984c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 3985c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru//------------------------------------------------------------------------------ 3986c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste QueruUnicodeSet *RegexCompile::scanPosixProp() { 3987c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *uset = NULL; 3988c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3989c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 3990c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return NULL; 3991b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 3992b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 3993c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(fC.fChar == chColon); 3994c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 3995c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Save the scanner state. 3996c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // TODO: move this into the scanner, with the state encapsulated in some way. Ticket 6062 399750294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int64_t savedScanIndex = fScanIndex; 399850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int64_t savedNextIndex = UTEXT_GETNATIVEINDEX(fRXPat->fPattern); 3999c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UBool savedQuoteMode = fQuoteMode; 4000c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UBool savedInBackslashQuote = fInBackslashQuote; 4001c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UBool savedEOLComments = fEOLComments; 400250294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int64_t savedLineNum = fLineNum; 400350294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho int64_t savedCharNum = fCharNum; 4004c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UChar32 savedLastChar = fLastChar; 4005c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UChar32 savedPeekChar = fPeekChar; 4006c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru RegexPatternChar savedfC = fC; 4007c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4008c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Scan for a closing ]. A little tricky because there are some perverse 4009c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // edge cases possible. "[:abc\Qdef:] \E]" is a valid non-property expression, 4010fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius // ending on the second closing ]. 4011c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4012c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeString propName; 4013c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UBool negated = FALSE; 4014c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4015c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Check for and consume the '^' in a negated POSIX property, e.g. [:^Letter:] 4016c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 4017c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar == chUp) { 4018c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru negated = TRUE; 4019c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 4020c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4021fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 4022c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Scan for the closing ":]", collecting the property name along the way. 4023c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UBool sawPropSetTerminator = FALSE; 4024c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru for (;;) { 4025c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru propName.append(fC.fChar); 4026c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 4027c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fQuoted || fC.fChar == -1) { 4028c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Escaped characters or end of input - either says this isn't a [:Property:] 4029c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4030c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4031c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar == chColon) { 4032c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru nextChar(fC); 4033c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (fC.fChar == chRBracket) { 4034c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru sawPropSetTerminator = TRUE; 4035c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4036c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4037c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4038c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4039fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 4040c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (sawPropSetTerminator) { 4041c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru uset = createSetForProperty(propName, negated); 4042c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4043c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else 4044c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 4045c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // No closing ":]". 4046c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Restore the original scan position. 4047c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The main scanner will retry the input as a normal set expression, 4048c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // not a [:Property:] expression. 4049c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fScanIndex = savedScanIndex; 4050c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fQuoteMode = savedQuoteMode; 4051c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fInBackslashQuote = savedInBackslashQuote; 4052c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fEOLComments = savedEOLComments; 4053c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fLineNum = savedLineNum; 4054c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fCharNum = savedCharNum; 4055c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fLastChar = savedLastChar; 4056c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fPeekChar = savedPeekChar; 4057c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fC = savedfC; 405850294ead5e5d23f5bbfed76e00e6b510bd41eee1claireho UTEXT_SETNATIVEINDEX(fRXPat->fPattern, savedNextIndex); 4059c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4060c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return uset; 4061c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru} 4062c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4063c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic inline void addIdentifierIgnorable(UnicodeSet *set, UErrorCode& ec) { 4064c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(0, 8).add(0x0e, 0x1b).add(0x7f, 0x9f); 4065c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_CF_MASK, ec); 4066c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru} 4067c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4068c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 4069c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// Create a Unicode Set from a Unicode Property expression. 4070c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// This is common code underlying both \p{...} ane [:...:] expressions. 4071c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// Includes trying the Java "properties" that aren't supported as 4072fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius// normal ICU UnicodeSet properties 4073c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 4074c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar posSetPrefix[] = {0x5b, 0x5c, 0x70, 0x7b, 0}; // "[\p{" 4075c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Querustatic const UChar negSetPrefix[] = {0x5b, 0x5c, 0x50, 0x7b, 0}; // "[\P{" 4076c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste QueruUnicodeSet *RegexCompile::createSetForProperty(const UnicodeString &propName, UBool negated) { 4077c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeString setExpr; 4078c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *set; 4079c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru uint32_t usetFlags = 0; 4080fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 4081c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_FAILURE(*fStatus)) { 4082c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return NULL; 4083c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4084c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4085c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4086c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // First try the property as we received it 4087c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4088c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (negated) { 4089c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(negSetPrefix, -1); 4090c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } else { 4091c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(posSetPrefix, -1); 4092c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4093c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(propName); 4094c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(chRBrace); 4095c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(chRBracket); 4096b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru if (fModeFlags & UREGEX_CASE_INSENSITIVE) { 4097b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru usetFlags |= USET_CASE_INSENSITIVE; 4098b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 4099c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus); 4100c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_SUCCESS(*fStatus)) { 4101c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return set; 4102c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4103c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete set; 4104c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set = NULL; 4105fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 4106c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4107c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // The property as it was didn't work. 410827f654740f2a26ad62a5c155af9199af9e69b889claireho 4109fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius // Do [:word:]. It is not recognized as a property by UnicodeSet. "word" not standard POSIX 411027f654740f2a26ad62a5c155af9199af9e69b889claireho // or standard Java, but many other regular expression packages do recognize it. 4111fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 411227f654740f2a26ad62a5c155af9199af9e69b889claireho if (propName.caseCompare(UNICODE_STRING_SIMPLE("word"), 0) == 0) { 411327f654740f2a26ad62a5c155af9199af9e69b889claireho *fStatus = U_ZERO_ERROR; 411427f654740f2a26ad62a5c155af9199af9e69b889claireho set = new UnicodeSet(*(fRXPat->fStaticSets[URX_ISWORD_SET])); 411527f654740f2a26ad62a5c155af9199af9e69b889claireho if (set == NULL) { 411627f654740f2a26ad62a5c155af9199af9e69b889claireho *fStatus = U_MEMORY_ALLOCATION_ERROR; 411727f654740f2a26ad62a5c155af9199af9e69b889claireho return set; 411827f654740f2a26ad62a5c155af9199af9e69b889claireho } 411927f654740f2a26ad62a5c155af9199af9e69b889claireho if (negated) { 412027f654740f2a26ad62a5c155af9199af9e69b889claireho set->complement(); 412127f654740f2a26ad62a5c155af9199af9e69b889claireho } 412227f654740f2a26ad62a5c155af9199af9e69b889claireho return set; 412327f654740f2a26ad62a5c155af9199af9e69b889claireho } 412427f654740f2a26ad62a5c155af9199af9e69b889claireho 412527f654740f2a26ad62a5c155af9199af9e69b889claireho 412627f654740f2a26ad62a5c155af9199af9e69b889claireho // Do Java fixes - 4127c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // InGreek -> InGreek or Coptic, that being the official Unicode name for that block. 4128c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // InCombiningMarksforSymbols -> InCombiningDiacriticalMarksforSymbols. 4129c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4130c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Note on Spaces: either "InCombiningMarksForSymbols" or "InCombining Marks for Symbols" 4131fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius // is accepted by Java. The property part of the name is compared 4132c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // case-insenstively. The spaces must be exactly as shown, either 4133c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // all there, or all omitted, with exactly one at each position 4134c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // if they are present. From checking against JDK 1.6 4135c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4136c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // This code should be removed when ICU properties support the Java compatibility names 4137c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // (ICU 4.0?) 4138c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4139c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeString mPropName = propName; 4140c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InGreek"), 0) == 0) { 4141c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru mPropName = UNICODE_STRING_SIMPLE("InGreek and Coptic"); 4142c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4143c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombining Marks for Symbols"), 0) == 0 || 4144c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru mPropName.caseCompare(UNICODE_STRING_SIMPLE("InCombiningMarksforSymbols"), 0) == 0) { 4145c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru mPropName = UNICODE_STRING_SIMPLE("InCombining Diacritical Marks for Symbols"); 4146c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4147c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) { 4148c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru mPropName = UNICODE_STRING_SIMPLE("javaValidCodePoint"); 4149c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4150fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius 4151c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // See if the property looks like a Java "InBlockName", which 4152c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // we will recast as "Block=BlockName" 4153c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4154c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru static const UChar IN[] = {0x49, 0x6E, 0}; // "In" 4155c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru static const UChar BLOCK[] = {0x42, 0x6C, 0x6f, 0x63, 0x6b, 0x3d, 00}; // "Block=" 4156c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (mPropName.startsWith(IN, 2) && propName.length()>=3) { 4157c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.truncate(4); // Leaves "[\p{", or "[\P{" 4158c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(BLOCK, -1); 4159c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(UnicodeString(mPropName, 2)); // Property with the leading "In" removed. 4160c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(chRBrace); 4161c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setExpr.append(chRBracket); 4162c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru *fStatus = U_ZERO_ERROR; 4163c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set = new UnicodeSet(setExpr, usetFlags, NULL, *fStatus); 4164c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_SUCCESS(*fStatus)) { 4165c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return set; 4166c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4167c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete set; 4168c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set = NULL; 4169b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 4170b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 4171c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (propName.startsWith(UNICODE_STRING_SIMPLE("java")) || 4172c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru propName.compare(UNICODE_STRING_SIMPLE("all")) == 0) 4173c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru { 4174c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UErrorCode localStatus = U_ZERO_ERROR; 4175c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru //setExpr.remove(); 4176c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set = new UnicodeSet(); 4177c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4178c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // Try the various Java specific properties. 4179c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // These all begin with "java" 4180c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // 4181c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDefined")) == 0) { 4182c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_CN_MASK, localStatus); 4183c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->complement(); 4184c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4185c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaDigit")) == 0) { 4186c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_ND_MASK, localStatus); 4187c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4188c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaIdentifierIgnorable")) == 0) { 4189c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addIdentifierIgnorable(set, localStatus); 4190c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4191c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaISOControl")) == 0) { 4192c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(0, 0x1F).add(0x7F, 0x9F); 4193c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4194c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierPart")) == 0) { 4195c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_L_MASK, localStatus); 4196c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_SC_MASK, localStatus); 4197c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_PC_MASK, localStatus); 4198c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_ND_MASK, localStatus); 4199c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_NL_MASK, localStatus); 4200c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_MC_MASK, localStatus); 4201c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_MN_MASK, localStatus); 4202c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addIdentifierIgnorable(set, localStatus); 4203c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4204c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaJavaIdentifierStart")) == 0) { 4205c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_L_MASK, localStatus); 4206c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_NL_MASK, localStatus); 4207c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_SC_MASK, localStatus); 4208c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_PC_MASK, localStatus); 4209c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4210c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetter")) == 0) { 4211c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_L_MASK, localStatus); 4212c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4213c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLetterOrDigit")) == 0) { 4214c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_L_MASK, localStatus); 4215c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_ND_MASK, localStatus); 4216c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4217c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaLowerCase")) == 0) { 4218c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_LL_MASK, localStatus); 4219c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4220c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaMirrored")) == 0) { 4221c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->applyIntPropertyValue(UCHAR_BIDI_MIRRORED, 1, localStatus); 4222c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4223c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSpaceChar")) == 0) { 4224c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_Z_MASK, localStatus); 4225c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4226c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaSupplementaryCodePoint")) == 0) { 4227c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(0x10000, UnicodeSet::MAX_VALUE); 4228c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4229c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaTitleCase")) == 0) { 4230c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_LT_MASK, localStatus); 4231c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4232c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierStart")) == 0) { 4233c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_L_MASK, localStatus); 4234c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_NL_MASK, localStatus); 4235c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4236c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUnicodeIdentifierPart")) == 0) { 4237c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_L_MASK, localStatus); 4238c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_PC_MASK, localStatus); 4239c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_ND_MASK, localStatus); 4240c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_NL_MASK, localStatus); 4241c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_MC_MASK, localStatus); 4242c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_MN_MASK, localStatus); 4243c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addIdentifierIgnorable(set, localStatus); 4244c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4245c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaUpperCase")) == 0) { 4246c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_LU_MASK, localStatus); 4247c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4248c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaValidCodePoint")) == 0) { 4249c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(0, UnicodeSet::MAX_VALUE); 4250c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4251c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("javaWhitespace")) == 0) { 4252c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru addCategory(set, U_GC_Z_MASK, localStatus); 4253c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->removeAll(UnicodeSet().add(0xa0).add(0x2007).add(0x202f)); 4254c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(9, 0x0d).add(0x1c, 0x1f); 4255c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4256c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru else if (mPropName.compare(UNICODE_STRING_SIMPLE("all")) == 0) { 4257c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->add(0, UnicodeSet::MAX_VALUE); 4258c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4259c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4260c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (U_SUCCESS(localStatus) && !set->isEmpty()) { 4261c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru *fStatus = U_ZERO_ERROR; 4262c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (usetFlags & USET_CASE_INSENSITIVE) { 4263c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->closeOver(USET_CASE_INSENSITIVE); 4264c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4265c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if (negated) { 4266c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set->complement(); 4267c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4268c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru return set; 4269c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4270c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete set; 4271c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru set = NULL; 4272b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru } 4273c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru error(*fStatus); 4274fceb39872958b9fa2505e63f8b8699a9e0f882f4ccornelius return NULL; 4275c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru} 4276b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 4277c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4278c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4279c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 4280c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// SetEval Part of the evaluation of [set expressions]. 4281c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// Perform any pending (stacked) operations with precedence 4282c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// equal or greater to that of the next operator encountered 4283c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// in the expression. 4284c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru// 4285c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queruvoid RegexCompile::setEval(int32_t nextOp) { 4286c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *rightOperand = NULL; 4287c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru UnicodeSet *leftOperand = NULL; 4288c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru for (;;) { 4289c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(fSetOpStack.empty()==FALSE); 4290c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru int32_t pendingSetOperation = fSetOpStack.peeki(); 4291c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru if ((pendingSetOperation&0xffff0000) < (nextOp&0xffff0000)) { 4292c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4293c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4294c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.popi(); 4295c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(fSetStack.empty() == FALSE); 4296c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru rightOperand = (UnicodeSet *)fSetStack.peek(); 4297c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru switch (pendingSetOperation) { 4298c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setNegation: 4299c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru rightOperand->complement(); 4300c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4301c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setCaseClose: 4302c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru // TODO: need a simple close function. Ticket 6065 4303c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru rightOperand->closeOver(USET_CASE_INSENSITIVE); 4304c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru rightOperand->removeAllStrings(); 4305c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4306c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setDifference1: 4307c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setDifference2: 4308c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetStack.pop(); 4309c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru leftOperand = (UnicodeSet *)fSetStack.peek(); 4310c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru leftOperand->removeAll(*rightOperand); 4311c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete rightOperand; 4312c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4313c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setIntersection1: 4314c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setIntersection2: 4315c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetStack.pop(); 4316c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru leftOperand = (UnicodeSet *)fSetStack.peek(); 4317c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru leftOperand->retainAll(*rightOperand); 4318c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete rightOperand; 4319c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4320c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru case setUnion: 4321c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetStack.pop(); 4322c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru leftOperand = (UnicodeSet *)fSetStack.peek(); 4323c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru leftOperand->addAll(*rightOperand); 4324c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru delete rightOperand; 4325c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4326c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru default: 4327c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru U_ASSERT(FALSE); 4328c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru break; 4329c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4330c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4331c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru } 4332c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru 4333c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queruvoid RegexCompile::setPushOp(int32_t op) { 4334c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru setEval(op); 4335c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetOpStack.push(op, *fStatus); 4336c69afcec261fc345fda8daf46f0ea6b4351dc777Jean-Baptiste Queru fSetStack.push(new UnicodeSet(), *fStatus); 4337b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru} 4338b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 4339b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste QueruU_NAMESPACE_END 4340b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru#endif // !UCONFIG_NO_REGULAR_EXPRESSIONS 4341b13da9df870a61b11249bf741347908dbea0edd8Jean-Baptiste Queru 4342