X-Git-Url: https://git.saurik.com/apple/icu.git/blobdiff_plain/729e4ab9bc6618bc3d8a898e575df7f4019e29ca..3d1f044b704633e2e541231cd17ae9ecf9ad5c7a:/icuSources/common/rbbiscan.cpp?ds=sidebyside diff --git a/icuSources/common/rbbiscan.cpp b/icuSources/common/rbbiscan.cpp index dacbd63a..f536ab5d 100644 --- a/icuSources/common/rbbiscan.cpp +++ b/icuSources/common/rbbiscan.cpp @@ -1,8 +1,9 @@ - +// © 2016 and later: Unicode, Inc. and others. +// License & terms of use: http://www.unicode.org/copyright.html // // file: rbbiscan.cpp // -// Copyright (C) 2002-2010, International Business Machines Corporation and others. +// Copyright (C) 2002-2016, International Business Machines Corporation and others. // All Rights Reserved. // // This file contains the Rule Based Break Iterator Rule Builder functions for @@ -23,7 +24,6 @@ #include "unicode/uchriter.h" #include "unicode/parsepos.h" #include "unicode/parseerr.h" -#include "util.h" #include "cmemory.h" #include "cstring.h" @@ -36,7 +36,6 @@ #include "uassert.h" - //------------------------------------------------------------------------------ // // Unicode Set init strings for each of the character classes needed for parsing a rule file. @@ -48,6 +47,7 @@ // //------------------------------------------------------------------------------ static const UChar gRuleSet_rule_char_pattern[] = { + // Characters that may appear as literals in patterns without escaping or quoting. // [ ^ [ \ p { Z } \ u 0 0 2 0 0x5b, 0x5e, 0x5b, 0x5c, 0x70, 0x7b, 0x5a, 0x7d, 0x5c, 0x75, 0x30, 0x30, 0x32, 0x30, // - \ u 0 0 7 f ] - [ \ p @@ -72,7 +72,7 @@ static const UChar kAny[] = {0x61, 0x6e, 0x79, 0x00}; // "any" U_CDECL_BEGIN static void U_CALLCONV RBBISetTable_deleter(void *p) { - U_NAMESPACE_QUALIFIER RBBISetTableEl *px = (U_NAMESPACE_QUALIFIER RBBISetTableEl *)p; + icu::RBBISetTableEl *px = (icu::RBBISetTableEl *)p; delete px->key; // Note: px->val is owned by the linked list "fSetsListHead" in scanner. // Don't delete the value nodes here. @@ -90,24 +90,27 @@ U_NAMESPACE_BEGIN RBBIRuleScanner::RBBIRuleScanner(RBBIRuleBuilder *rb) { fRB = rb; + fScanIndex = 0; + fNextIndex = 0; + fQuoteMode = FALSE; + fLineNum = 1; + fCharNum = 0; + fLastChar = 0; + + fStateTable = NULL; + fStack[0] = 0; fStackPtr = 0; - fStack[fStackPtr] = 0; - fNodeStackPtr = 0; - fRuleNum = 0; fNodeStack[0] = NULL; - - fSymbolTable = NULL; - fSetTable = NULL; - - fScanIndex = 0; - fNextIndex = 0; + fNodeStackPtr = 0; fReverseRule = FALSE; fLookAheadRule = FALSE; + fNoChainInRule = FALSE; - fLineNum = 1; - fCharNum = 0; - fQuoteMode = FALSE; + fSymbolTable = NULL; + fSetTable = NULL; + fRuleNum = 0; + fOptionStart = 0; // Do not check status until after all critical fields are sufficiently initialized // that the destructor can run cleanly. @@ -121,19 +124,20 @@ RBBIRuleScanner::RBBIRuleScanner(RBBIRuleBuilder *rb) // all instances of RBBIRuleScanners. BUT this is quite a bit simpler, // and the time to build these few sets should be small compared to a // full break iterator build. - fRuleSets[kRuleSet_rule_char-128] = UnicodeSet(gRuleSet_rule_char_pattern, *rb->fStatus); - UnicodeSet *whitespaceSet = uprv_openRuleWhiteSpaceSet(rb->fStatus); - if (U_FAILURE(*rb->fStatus)) { - return; - } - fRuleSets[kRuleSet_white_space-128] = *whitespaceSet; - delete whitespaceSet; - fRuleSets[kRuleSet_name_char-128] = UnicodeSet(gRuleSet_name_char_pattern, *rb->fStatus); - fRuleSets[kRuleSet_name_start_char-128] = UnicodeSet(gRuleSet_name_start_char_pattern, *rb->fStatus); - fRuleSets[kRuleSet_digit_char-128] = UnicodeSet(gRuleSet_digit_char_pattern, *rb->fStatus); + fRuleSets[kRuleSet_rule_char-128] + = UnicodeSet(UnicodeString(gRuleSet_rule_char_pattern), *rb->fStatus); + // fRuleSets[kRuleSet_white_space-128] = [:Pattern_White_Space:] + fRuleSets[kRuleSet_white_space-128]. + add(9, 0xd).add(0x20).add(0x85).add(0x200e, 0x200f).add(0x2028, 0x2029); + fRuleSets[kRuleSet_name_char-128] + = UnicodeSet(UnicodeString(gRuleSet_name_char_pattern), *rb->fStatus); + fRuleSets[kRuleSet_name_start_char-128] + = UnicodeSet(UnicodeString(gRuleSet_name_start_char_pattern), *rb->fStatus); + fRuleSets[kRuleSet_digit_char-128] + = UnicodeSet(UnicodeString(gRuleSet_digit_char_pattern), *rb->fStatus); if (*rb->fStatus == U_ILLEGAL_ARGUMENT_ERROR) { // This case happens if ICU's data is missing. UnicodeSet tries to look up property - // names from the init string, can't find them, and claims an illegal arguement. + // names from the init string, can't find them, and claims an illegal argument. // Change the error so that the actual problem will be clearer to users. *rb->fStatus = U_BRK_INIT_ERROR; } @@ -207,11 +211,20 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) break; + case doNoChain: + // Scanned a '^' while on the rule start state. + fNoChainInRule = TRUE; + break; + + case doExprOrOperator: { fixOpStack(RBBINode::precOpCat); RBBINode *operandNode = fNodeStack[fNodeStackPtr--]; RBBINode *orNode = pushNewNode(RBBINode::opOr); + if (U_FAILURE(*fRB->fStatus)) { + break; + } orNode->fLeftChild = operandNode; operandNode->fParent = orNode; } @@ -226,6 +239,9 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) fixOpStack(RBBINode::precOpCat); RBBINode *operandNode = fNodeStack[fNodeStackPtr--]; RBBINode *catNode = pushNewNode(RBBINode::opCat); + if (U_FAILURE(*fRB->fStatus)) { + break; + } catNode->fLeftChild = operandNode; operandNode->fParent = catNode; } @@ -314,21 +330,40 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "rtree")) {printNodeStack("end of rule");} #endif U_ASSERT(fNodeStackPtr == 1); + RBBINode *thisRule = fNodeStack[fNodeStackPtr]; // If this rule includes a look-ahead '/', add a endMark node to the // expression tree. if (fLookAheadRule) { - RBBINode *thisRule = fNodeStack[fNodeStackPtr]; RBBINode *endNode = pushNewNode(RBBINode::endMark); RBBINode *catNode = pushNewNode(RBBINode::opCat); + if (U_FAILURE(*fRB->fStatus)) { + break; + } fNodeStackPtr -= 2; catNode->fLeftChild = thisRule; catNode->fRightChild = endNode; fNodeStack[fNodeStackPtr] = catNode; endNode->fVal = fRuleNum; endNode->fLookAheadEnd = TRUE; + thisRule = catNode; + + // TODO: Disable chaining out of look-ahead (hard break) rules. + // The break on rule match is forced, so there is no point in building up + // the state table to chain into another rule for a longer match. } + // Mark this node as being the root of a rule. + thisRule->fRuleRoot = TRUE; + + // Flag if chaining into this rule is wanted. + // + if (fRB->fChainRules && // If rule chaining is enabled globally via !!chain + !fNoChainInRule) { // and no '^' chain-in inhibit was on this rule + thisRule->fChainIn = TRUE; + } + + // All rule expressions are ORed together. // The ';' that terminates an expression really just functions as a '|' with // a low operator prededence. @@ -337,7 +372,7 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) // (forward, reverse, safe_forward, safe_reverse) // OR this rule into the appropriate group of them. // - RBBINode **destRules = (fReverseRule? &fRB->fReverseTree : fRB->fDefaultTree); + RBBINode **destRules = (fReverseRule? &fRB->fSafeRevTree : fRB->fDefaultTree); if (*destRules != NULL) { // This is not the first rule encounted. @@ -345,9 +380,12 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) // with the current rule expression (on the Node Stack) // with the resulting OR expression going to *destRules // - RBBINode *thisRule = fNodeStack[fNodeStackPtr]; + thisRule = fNodeStack[fNodeStackPtr]; RBBINode *prevRules = *destRules; RBBINode *orNode = pushNewNode(RBBINode::opOr); + if (U_FAILURE(*fRB->fStatus)) { + break; + } orNode->fLeftChild = prevRules; prevRules->fParent = orNode; orNode->fRightChild = thisRule; @@ -362,6 +400,7 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) } fReverseRule = FALSE; // in preparation for the next rule. fLookAheadRule = FALSE; + fNoChainInRule = FALSE; fNodeStackPtr = 0; } break; @@ -388,6 +427,9 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) { RBBINode *operandNode = fNodeStack[fNodeStackPtr--]; RBBINode *plusNode = pushNewNode(RBBINode::opPlus); + if (U_FAILURE(*fRB->fStatus)) { + break; + } plusNode->fLeftChild = operandNode; operandNode->fParent = plusNode; } @@ -397,6 +439,9 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) { RBBINode *operandNode = fNodeStack[fNodeStackPtr--]; RBBINode *qNode = pushNewNode(RBBINode::opQuestion); + if (U_FAILURE(*fRB->fStatus)) { + break; + } qNode->fLeftChild = operandNode; operandNode->fParent = qNode; } @@ -406,6 +451,9 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) { RBBINode *operandNode = fNodeStack[fNodeStackPtr--]; RBBINode *starNode = pushNewNode(RBBINode::opStar); + if (U_FAILURE(*fRB->fStatus)) { + break; + } starNode->fLeftChild = operandNode; operandNode->fParent = starNode; } @@ -419,7 +467,10 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) // sets that just happen to contain only one character. { n = pushNewNode(RBBINode::setRef); - findSetFor(fC.fChar, n); + if (U_FAILURE(*fRB->fStatus)) { + break; + } + findSetFor(UnicodeString(fC.fChar), n); n->fFirstPos = fScanIndex; n->fLastPos = fNextIndex; fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText); @@ -430,7 +481,10 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) // scanned a ".", meaning match any single character. { n = pushNewNode(RBBINode::setRef); - findSetFor(kAny, n); + if (U_FAILURE(*fRB->fStatus)) { + break; + } + findSetFor(UnicodeString(TRUE, kAny, 3), n); n->fFirstPos = fScanIndex; n->fLastPos = fNextIndex; fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText); @@ -440,6 +494,9 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) case doSlash: // Scanned a '/', which identifies a look-ahead break position in a rule. n = pushNewNode(RBBINode::lookAhead); + if (U_FAILURE(*fRB->fStatus)) { + break; + } n->fVal = fRuleNum; n->fFirstPos = fScanIndex; n->fLastPos = fNextIndex; @@ -451,6 +508,9 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) case doStartTagValue: // Scanned a '{', the opening delimiter for a tag value within a rule. n = pushNewNode(RBBINode::tag); + if (U_FAILURE(*fRB->fStatus)) { + break; + } n->fVal = 0; n->fFirstPos = fScanIndex; n->fLastPos = fNextIndex; @@ -499,6 +559,10 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) fRB->fDefaultTree = &fRB->fSafeRevTree; } else if (opt == UNICODE_STRING("lookAheadHardBreak", 18)) { fRB->fLookAheadHardBreak = TRUE; + } else if (opt == UNICODE_STRING("quoted_literals_only", 20)) { + fRuleSets[kRuleSet_rule_char-128].clear(); + } else if (opt == UNICODE_STRING("unquoted_literals", 17)) { + fRuleSets[kRuleSet_rule_char-128].applyPattern(UnicodeString(gRuleSet_rule_char_pattern), *fRB->fStatus); } else { error(U_BRK_UNRECOGNIZED_OPTION); } @@ -561,7 +625,7 @@ UBool RBBIRuleScanner::doParseActions(int32_t action) returnVal = FALSE; break; } - return returnVal; + return returnVal && U_SUCCESS(*fRB->fStatus); } @@ -580,7 +644,7 @@ void RBBIRuleScanner::error(UErrorCode e) { fRB->fParseError->line = fLineNum; fRB->fParseError->offset = fCharNum; fRB->fParseError->preContext[0] = 0; - fRB->fParseError->preContext[0] = 0; + fRB->fParseError->postContext[0] = 0; } } } @@ -758,27 +822,24 @@ static const UChar chRParen = 0x29; //------------------------------------------------------------------------------ // -// stripRules Return a rules string without unnecessary -// characters. +// stripRules Return a rules string without extra spaces. +// (Comments are removed separately, during rule parsing.) // //------------------------------------------------------------------------------ UnicodeString RBBIRuleScanner::stripRules(const UnicodeString &rules) { UnicodeString strippedRules; - int rulesLength = rules.length(); - for (int idx = 0; idx < rulesLength; ) { - UChar ch = rules[idx++]; - if (ch == chPound) { - while (idx < rulesLength - && ch != chCR && ch != chLF && ch != chNEL) - { - ch = rules[idx++]; - } - } - if (!u_isISOControl(ch)) { - strippedRules.append(ch); + int32_t rulesLength = rules.length(); + bool skippingSpaces = false; + + for (int32_t idx=0; idxfStrippedRules.setCharAt(i, u' '); + } } if (c.fChar == (UChar32)-1) { return; @@ -963,7 +1028,7 @@ void RBBIRuleScanner::parse() { for (;;) { #ifdef RBBI_DEBUG - if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "scan")) { RBBIDebugPrintf(".");} + if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "scan")) { RBBIDebugPrintf("."); fflush(stdout);} #endif if (tableEl->fCharClass < 127 && fC.fEscaped == FALSE && tableEl->fCharClass == fC.fChar) { // Table row specified an individual character, not a set, and @@ -992,6 +1057,7 @@ void RBBIRuleScanner::parse() { if (tableEl->fCharClass >= 128 && tableEl->fCharClass < 240 && // Table specs a char class && fC.fEscaped == FALSE && // char is not escaped && fC.fChar != (UChar32)-1) { // char is not EOF + U_ASSERT((tableEl->fCharClass-128) < UPRV_LENGTHOF(fRuleSets)); if (fRuleSets[tableEl->fCharClass-128].contains(fC.fChar)) { // Table row specified a character class, or set of characters, // and the current char matches it. @@ -1045,19 +1111,17 @@ void RBBIRuleScanner::parse() { } + if (U_FAILURE(*fRB->fStatus)) { + return; + } + + // If there are no forward rules set an error. // - // If there were NO user specified reverse rules, set up the equivalent of ".*;" - // - if (fRB->fReverseTree == NULL) { - fRB->fReverseTree = pushNewNode(RBBINode::opStar); - RBBINode *operand = pushNewNode(RBBINode::setRef); - findSetFor(kAny, operand); - fRB->fReverseTree->fLeftChild = operand; - operand->fParent = fRB->fReverseTree; - fNodeStackPtr -= 2; + if (fRB->fForwardTree == NULL) { + error(U_BRK_RULE_SYNTAX); + return; } - // // Parsing of the input RBBI rules is complete. // We now have a parse tree for the rule expressions @@ -1065,16 +1129,15 @@ void RBBIRuleScanner::parse() { // #ifdef RBBI_DEBUG if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "symbols")) {fSymbolTable->rbbiSymtablePrint();} - if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ptree")) - { + if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ptree")) { RBBIDebugPrintf("Completed Forward Rules Parse Tree...\n"); - fRB->fForwardTree->printTree(TRUE); + RBBINode::printTree(fRB->fForwardTree, TRUE); RBBIDebugPrintf("\nCompleted Reverse Rules Parse Tree...\n"); - fRB->fReverseTree->printTree(TRUE); + RBBINode::printTree(fRB->fReverseTree, TRUE); RBBIDebugPrintf("\nCompleted Safe Point Forward Rules Parse Tree...\n"); - fRB->fSafeFwdTree->printTree(TRUE); + RBBINode::printTree(fRB->fSafeFwdTree, TRUE); RBBIDebugPrintf("\nCompleted Safe Point Reverse Rules Parse Tree...\n"); - fRB->fSafeRevTree->printTree(TRUE); + RBBINode::printTree(fRB->fSafeRevTree, TRUE); } #endif } @@ -1089,7 +1152,7 @@ void RBBIRuleScanner::parse() { void RBBIRuleScanner::printNodeStack(const char *title) { int i; RBBIDebugPrintf("%s. Dumping node stack...\n", title); - for (i=fNodeStackPtr; i>0; i--) {fNodeStack[i]->printTree(TRUE);} + for (i=fNodeStackPtr; i>0; i--) {RBBINode::printTree(fNodeStack[i], TRUE);} } #endif @@ -1103,13 +1166,15 @@ void RBBIRuleScanner::printNodeStack(const char *title) { // //------------------------------------------------------------------------------ RBBINode *RBBIRuleScanner::pushNewNode(RBBINode::NodeType t) { - fNodeStackPtr++; - if (fNodeStackPtr >= kStackSize) { - error(U_BRK_INTERNAL_ERROR); + if (U_FAILURE(*fRB->fStatus)) { + return NULL; + } + if (fNodeStackPtr >= kStackSize - 1) { + error(U_BRK_RULE_SYNTAX); RBBIDebugPuts("RBBIRuleScanner::pushNewNode - stack overflow."); - *fRB->fStatus = U_BRK_INTERNAL_ERROR; return NULL; } + fNodeStackPtr++; fNodeStack[fNodeStackPtr] = new RBBINode(t); if (fNodeStack[fNodeStackPtr] == NULL) { *fRB->fStatus = U_MEMORY_ALLOCATION_ERROR; @@ -1146,11 +1211,11 @@ void RBBIRuleScanner::scanSet() { pos.setIndex(fScanIndex); startPos = fScanIndex; UErrorCode localStatus = U_ZERO_ERROR; - uset = new UnicodeSet(fRB->fRules, pos, USET_IGNORE_SPACE, - fSymbolTable, - localStatus); + uset = new UnicodeSet(); if (uset == NULL) { localStatus = U_MEMORY_ALLOCATION_ERROR; + } else { + uset->applyPatternIgnoreSpace(fRB->fRules, pos, fSymbolTable, localStatus); } if (U_FAILURE(localStatus)) { // TODO: Get more accurate position of the error from UnicodeSet's return info. @@ -1165,6 +1230,7 @@ void RBBIRuleScanner::scanSet() { // Verify that the set contains at least one code point. // + U_ASSERT(uset!=NULL); if (uset->isEmpty()) { // This set is empty. // Make it an error, because it almost certainly is not what the user wanted. @@ -1191,6 +1257,9 @@ void RBBIRuleScanner::scanSet() { RBBINode *n; n = pushNewNode(RBBINode::setRef); + if (U_FAILURE(*fRB->fStatus)) { + return; + } n->fFirstPos = startPos; n->fLastPos = fNextIndex; fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText);