michael@0: /*
michael@0: *******************************************************************************
michael@0: *   Copyright (C) 2011-2012, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *******************************************************************************
michael@0: *   file name:  messagepattern.cpp
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2011mar14
michael@0: *   created by: Markus W. Scherer
michael@0: */
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: 
michael@0: #if !UCONFIG_NO_FORMATTING
michael@0: 
michael@0: #include "unicode/messagepattern.h"
michael@0: #include "unicode/unistr.h"
michael@0: #include "unicode/utf16.h"
michael@0: #include "cmemory.h"
michael@0: #include "cstring.h"
michael@0: #include "messageimpl.h"
michael@0: #include "patternprops.h"
michael@0: #include "putilimp.h"
michael@0: #include "uassert.h"
michael@0: 
michael@0: U_NAMESPACE_BEGIN
michael@0: 
michael@0: // Unicode character/code point constants ---------------------------------- ***
michael@0: 
michael@0: static const UChar u_pound=0x23;
michael@0: static const UChar u_apos=0x27;
michael@0: static const UChar u_plus=0x2B;
michael@0: static const UChar u_comma=0x2C;
michael@0: static const UChar u_minus=0x2D;
michael@0: static const UChar u_dot=0x2E;
michael@0: static const UChar u_colon=0x3A;
michael@0: static const UChar u_lessThan=0x3C;
michael@0: static const UChar u_equal=0x3D;
michael@0: static const UChar u_A=0x41;
michael@0: static const UChar u_C=0x43;
michael@0: static const UChar u_D=0x44;
michael@0: static const UChar u_E=0x45;
michael@0: static const UChar u_H=0x48;
michael@0: static const UChar u_I=0x49;
michael@0: static const UChar u_L=0x4C;
michael@0: static const UChar u_N=0x4E;
michael@0: static const UChar u_O=0x4F;
michael@0: static const UChar u_P=0x50;
michael@0: static const UChar u_R=0x52;
michael@0: static const UChar u_S=0x53;
michael@0: static const UChar u_T=0x54;
michael@0: static const UChar u_U=0x55;
michael@0: static const UChar u_Z=0x5A;
michael@0: static const UChar u_a=0x61;
michael@0: static const UChar u_c=0x63;
michael@0: static const UChar u_d=0x64;
michael@0: static const UChar u_e=0x65;
michael@0: static const UChar u_f=0x66;
michael@0: static const UChar u_h=0x68;
michael@0: static const UChar u_i=0x69;
michael@0: static const UChar u_l=0x6C;
michael@0: static const UChar u_n=0x6E;
michael@0: static const UChar u_o=0x6F;
michael@0: static const UChar u_p=0x70;
michael@0: static const UChar u_r=0x72;
michael@0: static const UChar u_s=0x73;
michael@0: static const UChar u_t=0x74;
michael@0: static const UChar u_u=0x75;
michael@0: static const UChar u_z=0x7A;
michael@0: static const UChar u_leftCurlyBrace=0x7B;
michael@0: static const UChar u_pipe=0x7C;
michael@0: static const UChar u_rightCurlyBrace=0x7D;
michael@0: static const UChar u_lessOrEqual=0x2264;  // U+2264 is <=
michael@0: 
michael@0: static const UChar kOffsetColon[]={  // "offset:"
michael@0:     u_o, u_f, u_f, u_s, u_e, u_t, u_colon
michael@0: };
michael@0: 
michael@0: static const UChar kOther[]={  // "other"
michael@0:     u_o, u_t, u_h, u_e, u_r
michael@0: };
michael@0: 
michael@0: // MessagePatternList ------------------------------------------------------ ***
michael@0: 
michael@0: template<typename T, int32_t stackCapacity>
michael@0: class MessagePatternList : public UMemory {
michael@0: public:
michael@0:     MessagePatternList() {}
michael@0:     void copyFrom(const MessagePatternList<T, stackCapacity> &other,
michael@0:                   int32_t length,
michael@0:                   UErrorCode &errorCode);
michael@0:     UBool ensureCapacityForOneMore(int32_t oldLength, UErrorCode &errorCode);
michael@0:     UBool equals(const MessagePatternList<T, stackCapacity> &other, int32_t length) const {
michael@0:         for(int32_t i=0; i<length; ++i) {
michael@0:             if(a[i]!=other.a[i]) { return FALSE; }
michael@0:         }
michael@0:         return TRUE;
michael@0:     }
michael@0: 
michael@0:     MaybeStackArray<T, stackCapacity> a;
michael@0: };
michael@0: 
michael@0: template<typename T, int32_t stackCapacity>
michael@0: void
michael@0: MessagePatternList<T, stackCapacity>::copyFrom(
michael@0:         const MessagePatternList<T, stackCapacity> &other,
michael@0:         int32_t length,
michael@0:         UErrorCode &errorCode) {
michael@0:     if(U_SUCCESS(errorCode) && length>0) {
michael@0:         if(length>a.getCapacity() && NULL==a.resize(length)) {
michael@0:             errorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:             return;
michael@0:         }
michael@0:         uprv_memcpy(a.getAlias(), other.a.getAlias(), length*sizeof(T));
michael@0:     }
michael@0: }
michael@0: 
michael@0: template<typename T, int32_t stackCapacity>
michael@0: UBool
michael@0: MessagePatternList<T, stackCapacity>::ensureCapacityForOneMore(int32_t oldLength, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return FALSE;
michael@0:     }
michael@0:     if(a.getCapacity()>oldLength || a.resize(2*oldLength, oldLength)!=NULL) {
michael@0:         return TRUE;
michael@0:     }
michael@0:     errorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:     return FALSE;
michael@0: }
michael@0: 
michael@0: // MessagePatternList specializations -------------------------------------- ***
michael@0: 
michael@0: class MessagePatternDoubleList : public MessagePatternList<double, 8> {
michael@0: };
michael@0: 
michael@0: class MessagePatternPartsList : public MessagePatternList<MessagePattern::Part, 32> {
michael@0: };
michael@0: 
michael@0: // MessagePattern constructors etc. ---------------------------------------- ***
michael@0: 
michael@0: MessagePattern::MessagePattern(UErrorCode &errorCode)
michael@0:         : aposMode(UCONFIG_MSGPAT_DEFAULT_APOSTROPHE_MODE),
michael@0:           partsList(NULL), parts(NULL), partsLength(0),
michael@0:           numericValuesList(NULL), numericValues(NULL), numericValuesLength(0),
michael@0:           hasArgNames(FALSE), hasArgNumbers(FALSE), needsAutoQuoting(FALSE) {
michael@0:     init(errorCode);
michael@0: }
michael@0: 
michael@0: MessagePattern::MessagePattern(UMessagePatternApostropheMode mode, UErrorCode &errorCode)
michael@0:         : aposMode(mode),
michael@0:           partsList(NULL), parts(NULL), partsLength(0),
michael@0:           numericValuesList(NULL), numericValues(NULL), numericValuesLength(0),
michael@0:           hasArgNames(FALSE), hasArgNumbers(FALSE), needsAutoQuoting(FALSE) {
michael@0:     init(errorCode);
michael@0: }
michael@0: 
michael@0: MessagePattern::MessagePattern(const UnicodeString &pattern, UParseError *parseError, UErrorCode &errorCode)
michael@0:         : aposMode(UCONFIG_MSGPAT_DEFAULT_APOSTROPHE_MODE),
michael@0:           partsList(NULL), parts(NULL), partsLength(0),
michael@0:           numericValuesList(NULL), numericValues(NULL), numericValuesLength(0),
michael@0:           hasArgNames(FALSE), hasArgNumbers(FALSE), needsAutoQuoting(FALSE) {
michael@0:     if(init(errorCode)) {
michael@0:         parse(pattern, parseError, errorCode);
michael@0:     }
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::init(UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return FALSE;
michael@0:     }
michael@0:     partsList=new MessagePatternPartsList();
michael@0:     if(partsList==NULL) {
michael@0:         errorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:         return FALSE;
michael@0:     }
michael@0:     parts=partsList->a.getAlias();
michael@0:     return TRUE;
michael@0: }
michael@0: 
michael@0: MessagePattern::MessagePattern(const MessagePattern &other)
michael@0:         : UObject(other), aposMode(other.aposMode), msg(other.msg),
michael@0:           partsList(NULL), parts(NULL), partsLength(0),
michael@0:           numericValuesList(NULL), numericValues(NULL), numericValuesLength(0),
michael@0:           hasArgNames(other.hasArgNames), hasArgNumbers(other.hasArgNumbers),
michael@0:           needsAutoQuoting(other.needsAutoQuoting) {
michael@0:     UErrorCode errorCode=U_ZERO_ERROR;
michael@0:     if(!copyStorage(other, errorCode)) {
michael@0:         clear();
michael@0:     }
michael@0: }
michael@0: 
michael@0: MessagePattern &
michael@0: MessagePattern::operator=(const MessagePattern &other) {
michael@0:     if(this==&other) {
michael@0:         return *this;
michael@0:     }
michael@0:     aposMode=other.aposMode;
michael@0:     msg=other.msg;
michael@0:     hasArgNames=other.hasArgNames;
michael@0:     hasArgNumbers=other.hasArgNumbers;
michael@0:     needsAutoQuoting=other.needsAutoQuoting;
michael@0:     UErrorCode errorCode=U_ZERO_ERROR;
michael@0:     if(!copyStorage(other, errorCode)) {
michael@0:         clear();
michael@0:     }
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::copyStorage(const MessagePattern &other, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return FALSE;
michael@0:     }
michael@0:     parts=NULL;
michael@0:     partsLength=0;
michael@0:     numericValues=NULL;
michael@0:     numericValuesLength=0;
michael@0:     if(partsList==NULL) {
michael@0:         partsList=new MessagePatternPartsList();
michael@0:         if(partsList==NULL) {
michael@0:             errorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:             return FALSE;
michael@0:         }
michael@0:         parts=partsList->a.getAlias();
michael@0:     }
michael@0:     if(other.partsLength>0) {
michael@0:         partsList->copyFrom(*other.partsList, other.partsLength, errorCode);
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return FALSE;
michael@0:         }
michael@0:         parts=partsList->a.getAlias();
michael@0:         partsLength=other.partsLength;
michael@0:     }
michael@0:     if(other.numericValuesLength>0) {
michael@0:         if(numericValuesList==NULL) {
michael@0:             numericValuesList=new MessagePatternDoubleList();
michael@0:             if(numericValuesList==NULL) {
michael@0:                 errorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:                 return FALSE;
michael@0:             }
michael@0:             numericValues=numericValuesList->a.getAlias();
michael@0:         }
michael@0:         numericValuesList->copyFrom(
michael@0:             *other.numericValuesList, other.numericValuesLength, errorCode);
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return FALSE;
michael@0:         }
michael@0:         numericValues=numericValuesList->a.getAlias();
michael@0:         numericValuesLength=other.numericValuesLength;
michael@0:     }
michael@0:     return TRUE;
michael@0: }
michael@0: 
michael@0: MessagePattern::~MessagePattern() {
michael@0:     delete partsList;
michael@0:     delete numericValuesList;
michael@0: }
michael@0: 
michael@0: // MessagePattern API ------------------------------------------------------ ***
michael@0: 
michael@0: MessagePattern &
michael@0: MessagePattern::parse(const UnicodeString &pattern, UParseError *parseError, UErrorCode &errorCode) {
michael@0:     preParse(pattern, parseError, errorCode);
michael@0:     parseMessage(0, 0, 0, UMSGPAT_ARG_TYPE_NONE, parseError, errorCode);
michael@0:     postParse();
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: MessagePattern &
michael@0: MessagePattern::parseChoiceStyle(const UnicodeString &pattern,
michael@0:                                  UParseError *parseError, UErrorCode &errorCode) {
michael@0:     preParse(pattern, parseError, errorCode);
michael@0:     parseChoiceStyle(0, 0, parseError, errorCode);
michael@0:     postParse();
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: MessagePattern &
michael@0: MessagePattern::parsePluralStyle(const UnicodeString &pattern,
michael@0:                                  UParseError *parseError, UErrorCode &errorCode) {
michael@0:     preParse(pattern, parseError, errorCode);
michael@0:     parsePluralOrSelectStyle(UMSGPAT_ARG_TYPE_PLURAL, 0, 0, parseError, errorCode);
michael@0:     postParse();
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: MessagePattern &
michael@0: MessagePattern::parseSelectStyle(const UnicodeString &pattern,
michael@0:                                  UParseError *parseError, UErrorCode &errorCode) {
michael@0:     preParse(pattern, parseError, errorCode);
michael@0:     parsePluralOrSelectStyle(UMSGPAT_ARG_TYPE_SELECT, 0, 0, parseError, errorCode);
michael@0:     postParse();
michael@0:     return *this;
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::clear() {
michael@0:     // Mostly the same as preParse().
michael@0:     msg.remove();
michael@0:     hasArgNames=hasArgNumbers=FALSE;
michael@0:     needsAutoQuoting=FALSE;
michael@0:     partsLength=0;
michael@0:     numericValuesLength=0;
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::operator==(const MessagePattern &other) const {
michael@0:     if(this==&other) {
michael@0:         return TRUE;
michael@0:     }
michael@0:     return
michael@0:         aposMode==other.aposMode &&
michael@0:         msg==other.msg &&
michael@0:         // parts.equals(o.parts)
michael@0:         partsLength==other.partsLength &&
michael@0:         (partsLength==0 || partsList->equals(*other.partsList, partsLength));
michael@0:     // No need to compare numericValues if msg and parts are the same.
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::hashCode() const {
michael@0:     int32_t hash=(aposMode*37+msg.hashCode())*37+partsLength;
michael@0:     for(int32_t i=0; i<partsLength; ++i) {
michael@0:         hash=hash*37+parts[i].hashCode();
michael@0:     }
michael@0:     return hash;
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::validateArgumentName(const UnicodeString &name) {
michael@0:     if(!PatternProps::isIdentifier(name.getBuffer(), name.length())) {
michael@0:         return UMSGPAT_ARG_NAME_NOT_VALID;
michael@0:     }
michael@0:     return parseArgNumber(name, 0, name.length());
michael@0: }
michael@0: 
michael@0: UnicodeString
michael@0: MessagePattern::autoQuoteApostropheDeep() const {
michael@0:     if(!needsAutoQuoting) {
michael@0:         return msg;
michael@0:     }
michael@0:     UnicodeString modified(msg);
michael@0:     // Iterate backward so that the insertion indexes do not change.
michael@0:     int32_t count=countParts();
michael@0:     for(int32_t i=count; i>0;) {
michael@0:         const Part &part=getPart(--i);
michael@0:         if(part.getType()==UMSGPAT_PART_TYPE_INSERT_CHAR) {
michael@0:            modified.insert(part.index, (UChar)part.value);
michael@0:         }
michael@0:     }
michael@0:     return modified;
michael@0: }
michael@0: 
michael@0: double
michael@0: MessagePattern::getNumericValue(const Part &part) const {
michael@0:     UMessagePatternPartType type=part.type;
michael@0:     if(type==UMSGPAT_PART_TYPE_ARG_INT) {
michael@0:         return part.value;
michael@0:     } else if(type==UMSGPAT_PART_TYPE_ARG_DOUBLE) {
michael@0:         return numericValues[part.value];
michael@0:     } else {
michael@0:         return UMSGPAT_NO_NUMERIC_VALUE;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /**
michael@0:   * Returns the "offset:" value of a PluralFormat argument, or 0 if none is specified.
michael@0:   * @param pluralStart the index of the first PluralFormat argument style part. (0..countParts()-1)
michael@0:   * @return the "offset:" value.
michael@0:   * @draft ICU 4.8
michael@0:   */
michael@0: double
michael@0: MessagePattern::getPluralOffset(int32_t pluralStart) const {
michael@0:     const Part &part=getPart(pluralStart);
michael@0:     if(Part::hasNumericValue(part.type)) {
michael@0:         return getNumericValue(part);
michael@0:     } else {
michael@0:         return 0;
michael@0:     }
michael@0: }
michael@0: 
michael@0: // MessagePattern::Part ---------------------------------------------------- ***
michael@0: 
michael@0: UBool
michael@0: MessagePattern::Part::operator==(const Part &other) const {
michael@0:     if(this==&other) {
michael@0:         return TRUE;
michael@0:     }
michael@0:     return
michael@0:         type==other.type &&
michael@0:         index==other.index &&
michael@0:         length==other.length &&
michael@0:         value==other.value &&
michael@0:         limitPartIndex==other.limitPartIndex;
michael@0: }
michael@0: 
michael@0: // MessagePattern parser --------------------------------------------------- ***
michael@0: 
michael@0: void
michael@0: MessagePattern::preParse(const UnicodeString &pattern, UParseError *parseError, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return;
michael@0:     }
michael@0:     if(parseError!=NULL) {
michael@0:         parseError->line=0;
michael@0:         parseError->offset=0;
michael@0:         parseError->preContext[0]=0;
michael@0:         parseError->postContext[0]=0;
michael@0:     }
michael@0:     msg=pattern;
michael@0:     hasArgNames=hasArgNumbers=FALSE;
michael@0:     needsAutoQuoting=FALSE;
michael@0:     partsLength=0;
michael@0:     numericValuesLength=0;
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::postParse() {
michael@0:     if(partsList!=NULL) {
michael@0:         parts=partsList->a.getAlias();
michael@0:     }
michael@0:     if(numericValuesList!=NULL) {
michael@0:         numericValues=numericValuesList->a.getAlias();
michael@0:     }
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::parseMessage(int32_t index, int32_t msgStartLength,
michael@0:                              int32_t nestingLevel, UMessagePatternArgType parentType,
michael@0:                              UParseError *parseError, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(nestingLevel>Part::MAX_VALUE) {
michael@0:         errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t msgStart=partsLength;
michael@0:     addPart(UMSGPAT_PART_TYPE_MSG_START, index, msgStartLength, nestingLevel, errorCode);
michael@0:     index+=msgStartLength;
michael@0:     for(;;) {  // while(index<msg.length()) with U_FAILURE(errorCode) check
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return 0;
michael@0:         }
michael@0:         if(index>=msg.length()) {
michael@0:             break;
michael@0:         }
michael@0:         UChar c=msg.charAt(index++);
michael@0:         if(c==u_apos) {
michael@0:             if(index==msg.length()) {
michael@0:                 // The apostrophe is the last character in the pattern. 
michael@0:                 // Add a Part for auto-quoting.
michael@0:                 addPart(UMSGPAT_PART_TYPE_INSERT_CHAR, index, 0,
michael@0:                         u_apos, errorCode);  // value=char to be inserted
michael@0:                 needsAutoQuoting=TRUE;
michael@0:             } else {
michael@0:                 c=msg.charAt(index);
michael@0:                 if(c==u_apos) {
michael@0:                     // double apostrophe, skip the second one
michael@0:                     addPart(UMSGPAT_PART_TYPE_SKIP_SYNTAX, index++, 1, 0, errorCode);
michael@0:                 } else if(
michael@0:                     aposMode==UMSGPAT_APOS_DOUBLE_REQUIRED ||
michael@0:                     c==u_leftCurlyBrace || c==u_rightCurlyBrace ||
michael@0:                     (parentType==UMSGPAT_ARG_TYPE_CHOICE && c==u_pipe) ||
michael@0:                     (UMSGPAT_ARG_TYPE_HAS_PLURAL_STYLE(parentType) && c==u_pound)
michael@0:                 ) {
michael@0:                     // skip the quote-starting apostrophe
michael@0:                     addPart(UMSGPAT_PART_TYPE_SKIP_SYNTAX, index-1, 1, 0, errorCode);
michael@0:                     // find the end of the quoted literal text
michael@0:                     for(;;) {
michael@0:                         index=msg.indexOf(u_apos, index+1);
michael@0:                         if(index>=0) {
michael@0:                             if(/*(index+1)<msg.length() &&*/ msg.charAt(index+1)==u_apos) {
michael@0:                                 // double apostrophe inside quoted literal text
michael@0:                                 // still encodes a single apostrophe, skip the second one
michael@0:                                 addPart(UMSGPAT_PART_TYPE_SKIP_SYNTAX, ++index, 1, 0, errorCode);
michael@0:                             } else {
michael@0:                                 // skip the quote-ending apostrophe
michael@0:                                 addPart(UMSGPAT_PART_TYPE_SKIP_SYNTAX, index++, 1, 0, errorCode);
michael@0:                                 break;
michael@0:                             }
michael@0:                         } else {
michael@0:                             // The quoted text reaches to the end of the of the message.
michael@0:                             index=msg.length();
michael@0:                             // Add a Part for auto-quoting.
michael@0:                             addPart(UMSGPAT_PART_TYPE_INSERT_CHAR, index, 0,
michael@0:                                     u_apos, errorCode);  // value=char to be inserted
michael@0:                             needsAutoQuoting=TRUE;
michael@0:                             break;
michael@0:                         }
michael@0:                     }
michael@0:                 } else {
michael@0:                     // Interpret the apostrophe as literal text.
michael@0:                     // Add a Part for auto-quoting.
michael@0:                     addPart(UMSGPAT_PART_TYPE_INSERT_CHAR, index, 0,
michael@0:                             u_apos, errorCode);  // value=char to be inserted
michael@0:                     needsAutoQuoting=TRUE;
michael@0:                 }
michael@0:             }
michael@0:         } else if(UMSGPAT_ARG_TYPE_HAS_PLURAL_STYLE(parentType) && c==u_pound) {
michael@0:             // The unquoted # in a plural message fragment will be replaced
michael@0:             // with the (number-offset).
michael@0:             addPart(UMSGPAT_PART_TYPE_REPLACE_NUMBER, index-1, 1, 0, errorCode);
michael@0:         } else if(c==u_leftCurlyBrace) {
michael@0:             index=parseArg(index-1, 1, nestingLevel, parseError, errorCode);
michael@0:         } else if((nestingLevel>0 && c==u_rightCurlyBrace) ||
michael@0:                   (parentType==UMSGPAT_ARG_TYPE_CHOICE && c==u_pipe)) {
michael@0:             // Finish the message before the terminator.
michael@0:             // In a choice style, report the "}" substring only for the following ARG_LIMIT,
michael@0:             // not for this MSG_LIMIT.
michael@0:             int32_t limitLength=(parentType==UMSGPAT_ARG_TYPE_CHOICE && c==u_rightCurlyBrace) ? 0 : 1;
michael@0:             addLimitPart(msgStart, UMSGPAT_PART_TYPE_MSG_LIMIT, index-1, limitLength,
michael@0:                          nestingLevel, errorCode);
michael@0:             if(parentType==UMSGPAT_ARG_TYPE_CHOICE) {
michael@0:                 // Let the choice style parser see the '}' or '|'.
michael@0:                 return index-1;
michael@0:             } else {
michael@0:                 // continue parsing after the '}'
michael@0:                 return index;
michael@0:             }
michael@0:         }  // else: c is part of literal text
michael@0:     }
michael@0:     if(nestingLevel>0 && !inTopLevelChoiceMessage(nestingLevel, parentType)) {
michael@0:         setParseError(parseError, 0);  // Unmatched '{' braces in message.
michael@0:         errorCode=U_UNMATCHED_BRACES;
michael@0:         return 0;
michael@0:     }
michael@0:     addLimitPart(msgStart, UMSGPAT_PART_TYPE_MSG_LIMIT, index, 0, nestingLevel, errorCode);
michael@0:     return index;
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::parseArg(int32_t index, int32_t argStartLength, int32_t nestingLevel,
michael@0:                          UParseError *parseError, UErrorCode &errorCode) {
michael@0:     int32_t argStart=partsLength;
michael@0:     UMessagePatternArgType argType=UMSGPAT_ARG_TYPE_NONE;
michael@0:     addPart(UMSGPAT_PART_TYPE_ARG_START, index, argStartLength, argType, errorCode);
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t nameIndex=index=skipWhiteSpace(index+argStartLength);
michael@0:     if(index==msg.length()) {
michael@0:         setParseError(parseError, 0);  // Unmatched '{' braces in message.
michael@0:         errorCode=U_UNMATCHED_BRACES;
michael@0:         return 0;
michael@0:     }
michael@0:     // parse argument name or number
michael@0:     index=skipIdentifier(index);
michael@0:     int32_t number=parseArgNumber(nameIndex, index);
michael@0:     if(number>=0) {
michael@0:         int32_t length=index-nameIndex;
michael@0:         if(length>Part::MAX_LENGTH || number>Part::MAX_VALUE) {
michael@0:             setParseError(parseError, nameIndex);  // Argument number too large.
michael@0:             errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         hasArgNumbers=TRUE;
michael@0:         addPart(UMSGPAT_PART_TYPE_ARG_NUMBER, nameIndex, length, number, errorCode);
michael@0:     } else if(number==UMSGPAT_ARG_NAME_NOT_NUMBER) {
michael@0:         int32_t length=index-nameIndex;
michael@0:         if(length>Part::MAX_LENGTH) {
michael@0:             setParseError(parseError, nameIndex);  // Argument name too long.
michael@0:             errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         hasArgNames=TRUE;
michael@0:         addPart(UMSGPAT_PART_TYPE_ARG_NAME, nameIndex, length, 0, errorCode);
michael@0:     } else {  // number<-1 (ARG_NAME_NOT_VALID)
michael@0:         setParseError(parseError, nameIndex);  // Bad argument syntax.
michael@0:         errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     index=skipWhiteSpace(index);
michael@0:     if(index==msg.length()) {
michael@0:         setParseError(parseError, 0);  // Unmatched '{' braces in message.
michael@0:         errorCode=U_UNMATCHED_BRACES;
michael@0:         return 0;
michael@0:     }
michael@0:     UChar c=msg.charAt(index);
michael@0:     if(c==u_rightCurlyBrace) {
michael@0:         // all done
michael@0:     } else if(c!=u_comma) {
michael@0:         setParseError(parseError, nameIndex);  // Bad argument syntax.
michael@0:         errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:         return 0;
michael@0:     } else /* ',' */ {
michael@0:         // parse argument type: case-sensitive a-zA-Z
michael@0:         int32_t typeIndex=index=skipWhiteSpace(index+1);
michael@0:         while(index<msg.length() && isArgTypeChar(msg.charAt(index))) {
michael@0:             ++index;
michael@0:         }
michael@0:         int32_t length=index-typeIndex;
michael@0:         index=skipWhiteSpace(index);
michael@0:         if(index==msg.length()) {
michael@0:             setParseError(parseError, 0);  // Unmatched '{' braces in message.
michael@0:             errorCode=U_UNMATCHED_BRACES;
michael@0:             return 0;
michael@0:         }
michael@0:         if(length==0 || ((c=msg.charAt(index))!=u_comma && c!=u_rightCurlyBrace)) {
michael@0:             setParseError(parseError, nameIndex);  // Bad argument syntax.
michael@0:             errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         if(length>Part::MAX_LENGTH) {
michael@0:             setParseError(parseError, nameIndex);  // Argument type name too long.
michael@0:             errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         argType=UMSGPAT_ARG_TYPE_SIMPLE;
michael@0:         if(length==6) {
michael@0:             // case-insensitive comparisons for complex-type names
michael@0:             if(isChoice(typeIndex)) {
michael@0:                 argType=UMSGPAT_ARG_TYPE_CHOICE;
michael@0:             } else if(isPlural(typeIndex)) {
michael@0:                 argType=UMSGPAT_ARG_TYPE_PLURAL;
michael@0:             } else if(isSelect(typeIndex)) {
michael@0:                 argType=UMSGPAT_ARG_TYPE_SELECT;
michael@0:             }
michael@0:         } else if(length==13) {
michael@0:             if(isSelect(typeIndex) && isOrdinal(typeIndex+6)) {
michael@0:                 argType=UMSGPAT_ARG_TYPE_SELECTORDINAL;
michael@0:             }
michael@0:         }
michael@0:         // change the ARG_START type from NONE to argType
michael@0:         partsList->a[argStart].value=(int16_t)argType;
michael@0:         if(argType==UMSGPAT_ARG_TYPE_SIMPLE) {
michael@0:             addPart(UMSGPAT_PART_TYPE_ARG_TYPE, typeIndex, length, 0, errorCode);
michael@0:         }
michael@0:         // look for an argument style (pattern)
michael@0:         if(c==u_rightCurlyBrace) {
michael@0:             if(argType!=UMSGPAT_ARG_TYPE_SIMPLE) {
michael@0:                 setParseError(parseError, nameIndex);  // No style field for complex argument.
michael@0:                 errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:         } else /* ',' */ {
michael@0:             ++index;
michael@0:             if(argType==UMSGPAT_ARG_TYPE_SIMPLE) {
michael@0:                 index=parseSimpleStyle(index, parseError, errorCode);
michael@0:             } else if(argType==UMSGPAT_ARG_TYPE_CHOICE) {
michael@0:                 index=parseChoiceStyle(index, nestingLevel, parseError, errorCode);
michael@0:             } else {
michael@0:                 index=parsePluralOrSelectStyle(argType, index, nestingLevel, parseError, errorCode);
michael@0:             }
michael@0:         }
michael@0:     }
michael@0:     // Argument parsing stopped on the '}'.
michael@0:     addLimitPart(argStart, UMSGPAT_PART_TYPE_ARG_LIMIT, index, 1, argType, errorCode);
michael@0:     return index+1;
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::parseSimpleStyle(int32_t index, UParseError *parseError, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t start=index;
michael@0:     int32_t nestedBraces=0;
michael@0:     while(index<msg.length()) {
michael@0:         UChar c=msg.charAt(index++);
michael@0:         if(c==u_apos) {
michael@0:             // Treat apostrophe as quoting but include it in the style part.
michael@0:             // Find the end of the quoted literal text.
michael@0:             index=msg.indexOf(u_apos, index);
michael@0:             if(index<0) {
michael@0:                 // Quoted literal argument style text reaches to the end of the message.
michael@0:                 setParseError(parseError, start);
michael@0:                 errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:             // skip the quote-ending apostrophe
michael@0:             ++index;
michael@0:         } else if(c==u_leftCurlyBrace) {
michael@0:             ++nestedBraces;
michael@0:         } else if(c==u_rightCurlyBrace) {
michael@0:             if(nestedBraces>0) {
michael@0:                 --nestedBraces;
michael@0:             } else {
michael@0:                 int32_t length=--index-start;
michael@0:                 if(length>Part::MAX_LENGTH) {
michael@0:                     setParseError(parseError, start);  // Argument style text too long.
michael@0:                     errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:                     return 0;
michael@0:                 }
michael@0:                 addPart(UMSGPAT_PART_TYPE_ARG_STYLE, start, length, 0, errorCode);
michael@0:                 return index;
michael@0:             }
michael@0:         }  // c is part of literal text
michael@0:     }
michael@0:     setParseError(parseError, 0);  // Unmatched '{' braces in message.
michael@0:     errorCode=U_UNMATCHED_BRACES;
michael@0:     return 0;
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::parseChoiceStyle(int32_t index, int32_t nestingLevel,
michael@0:                                  UParseError *parseError, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t start=index;
michael@0:     index=skipWhiteSpace(index);
michael@0:     if(index==msg.length() || msg.charAt(index)==u_rightCurlyBrace) {
michael@0:         setParseError(parseError, 0);  // Missing choice argument pattern.
michael@0:         errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     for(;;) {
michael@0:         // The choice argument style contains |-separated (number, separator, message) triples.
michael@0:         // Parse the number.
michael@0:         int32_t numberIndex=index;
michael@0:         index=skipDouble(index);
michael@0:         int32_t length=index-numberIndex;
michael@0:         if(length==0) {
michael@0:             setParseError(parseError, start);  // Bad choice pattern syntax.
michael@0:             errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         if(length>Part::MAX_LENGTH) {
michael@0:             setParseError(parseError, numberIndex);  // Choice number too long.
michael@0:             errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         parseDouble(numberIndex, index, TRUE, parseError, errorCode);  // adds ARG_INT or ARG_DOUBLE
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return 0;
michael@0:         }
michael@0:         // Parse the separator.
michael@0:         index=skipWhiteSpace(index);
michael@0:         if(index==msg.length()) {
michael@0:             setParseError(parseError, start);  // Bad choice pattern syntax.
michael@0:             errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         UChar c=msg.charAt(index);
michael@0:         if(!(c==u_pound || c==u_lessThan || c==u_lessOrEqual)) {  // U+2264 is <=
michael@0:             setParseError(parseError, start);  // Expected choice separator (#<\u2264) instead of c.
michael@0:             errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         addPart(UMSGPAT_PART_TYPE_ARG_SELECTOR, index, 1, 0, errorCode);
michael@0:         // Parse the message fragment.
michael@0:         index=parseMessage(++index, 0, nestingLevel+1, UMSGPAT_ARG_TYPE_CHOICE, parseError, errorCode);
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return 0;
michael@0:         }
michael@0:         // parseMessage(..., CHOICE) returns the index of the terminator, or msg.length().
michael@0:         if(index==msg.length()) {
michael@0:             return index;
michael@0:         }
michael@0:         if(msg.charAt(index)==u_rightCurlyBrace) {
michael@0:             if(!inMessageFormatPattern(nestingLevel)) {
michael@0:                 setParseError(parseError, start);  // Bad choice pattern syntax.
michael@0:                 errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:             return index;
michael@0:         }  // else the terminator is '|'
michael@0:         index=skipWhiteSpace(index+1);
michael@0:     }
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::parsePluralOrSelectStyle(UMessagePatternArgType argType,
michael@0:                                          int32_t index, int32_t nestingLevel,
michael@0:                                          UParseError *parseError, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t start=index;
michael@0:     UBool isEmpty=TRUE;
michael@0:     UBool hasOther=FALSE;
michael@0:     for(;;) {
michael@0:         // First, collect the selector looking for a small set of terminators.
michael@0:         // It would be a little faster to consider the syntax of each possible
michael@0:         // token right here, but that makes the code too complicated.
michael@0:         index=skipWhiteSpace(index);
michael@0:         UBool eos=index==msg.length();
michael@0:         if(eos || msg.charAt(index)==u_rightCurlyBrace) {
michael@0:             if(eos==inMessageFormatPattern(nestingLevel)) {
michael@0:                 setParseError(parseError, start);  // Bad plural/select pattern syntax.
michael@0:                 errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:             if(!hasOther) {
michael@0:                 setParseError(parseError, 0);  // Missing 'other' keyword in plural/select pattern.
michael@0:                 errorCode=U_DEFAULT_KEYWORD_MISSING;
michael@0:                 return 0;
michael@0:             }
michael@0:             return index;
michael@0:         }
michael@0:         int32_t selectorIndex=index;
michael@0:         if(UMSGPAT_ARG_TYPE_HAS_PLURAL_STYLE(argType) && msg.charAt(selectorIndex)==u_equal) {
michael@0:             // explicit-value plural selector: =double
michael@0:             index=skipDouble(index+1);
michael@0:             int32_t length=index-selectorIndex;
michael@0:             if(length==1) {
michael@0:                 setParseError(parseError, start);  // Bad plural/select pattern syntax.
michael@0:                 errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:             if(length>Part::MAX_LENGTH) {
michael@0:                 setParseError(parseError, selectorIndex);  // Argument selector too long.
michael@0:                 errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:             addPart(UMSGPAT_PART_TYPE_ARG_SELECTOR, selectorIndex, length, 0, errorCode);
michael@0:             parseDouble(selectorIndex+1, index, FALSE,
michael@0:                         parseError, errorCode);  // adds ARG_INT or ARG_DOUBLE
michael@0:         } else {
michael@0:             index=skipIdentifier(index);
michael@0:             int32_t length=index-selectorIndex;
michael@0:             if(length==0) {
michael@0:                 setParseError(parseError, start);  // Bad plural/select pattern syntax.
michael@0:                 errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                 return 0;
michael@0:             }
michael@0:             // Note: The ':' in "offset:" is just beyond the skipIdentifier() range.
michael@0:             if( UMSGPAT_ARG_TYPE_HAS_PLURAL_STYLE(argType) && length==6 && index<msg.length() &&
michael@0:                 0==msg.compare(selectorIndex, 7, kOffsetColon, 0, 7)
michael@0:             ) {
michael@0:                 // plural offset, not a selector
michael@0:                 if(!isEmpty) {
michael@0:                     // Plural argument 'offset:' (if present) must precede key-message pairs.
michael@0:                     setParseError(parseError, start);
michael@0:                     errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                     return 0;
michael@0:                 }
michael@0:                 // allow whitespace between offset: and its value
michael@0:                 int32_t valueIndex=skipWhiteSpace(index+1);  // The ':' is at index.
michael@0:                 index=skipDouble(valueIndex);
michael@0:                 if(index==valueIndex) {
michael@0:                     setParseError(parseError, start);  // Missing value for plural 'offset:'.
michael@0:                     errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:                     return 0;
michael@0:                 }
michael@0:                 if((index-valueIndex)>Part::MAX_LENGTH) {
michael@0:                     setParseError(parseError, valueIndex);  // Plural offset value too long.
michael@0:                     errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:                     return 0;
michael@0:                 }
michael@0:                 parseDouble(valueIndex, index, FALSE,
michael@0:                             parseError, errorCode);  // adds ARG_INT or ARG_DOUBLE
michael@0:                 if(U_FAILURE(errorCode)) {
michael@0:                     return 0;
michael@0:                 }
michael@0:                 isEmpty=FALSE;
michael@0:                 continue;  // no message fragment after the offset
michael@0:             } else {
michael@0:                 // normal selector word
michael@0:                 if(length>Part::MAX_LENGTH) {
michael@0:                     setParseError(parseError, selectorIndex);  // Argument selector too long.
michael@0:                     errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:                     return 0;
michael@0:                 }
michael@0:                 addPart(UMSGPAT_PART_TYPE_ARG_SELECTOR, selectorIndex, length, 0, errorCode);
michael@0:                 if(0==msg.compare(selectorIndex, length, kOther, 0, 5)) {
michael@0:                     hasOther=TRUE;
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return 0;
michael@0:         }
michael@0: 
michael@0:         // parse the message fragment following the selector
michael@0:         index=skipWhiteSpace(index);
michael@0:         if(index==msg.length() || msg.charAt(index)!=u_leftCurlyBrace) {
michael@0:             setParseError(parseError, selectorIndex);  // No message fragment after plural/select selector.
michael@0:             errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0:         index=parseMessage(index, 1, nestingLevel+1, argType, parseError, errorCode);
michael@0:         if(U_FAILURE(errorCode)) {
michael@0:             return 0;
michael@0:         }
michael@0:         isEmpty=FALSE;
michael@0:     }
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::parseArgNumber(const UnicodeString &s, int32_t start, int32_t limit) {
michael@0:     // If the identifier contains only ASCII digits, then it is an argument _number_
michael@0:     // and must not have leading zeros (except "0" itself).
michael@0:     // Otherwise it is an argument _name_.
michael@0:     if(start>=limit) {
michael@0:         return UMSGPAT_ARG_NAME_NOT_VALID;
michael@0:     }
michael@0:     int32_t number;
michael@0:     // Defer numeric errors until we know there are only digits.
michael@0:     UBool badNumber;
michael@0:     UChar c=s.charAt(start++);
michael@0:     if(c==0x30) {
michael@0:         if(start==limit) {
michael@0:             return 0;
michael@0:         } else {
michael@0:             number=0;
michael@0:             badNumber=TRUE;  // leading zero
michael@0:         }
michael@0:     } else if(0x31<=c && c<=0x39) {
michael@0:         number=c-0x30;
michael@0:         badNumber=FALSE;
michael@0:     } else {
michael@0:         return UMSGPAT_ARG_NAME_NOT_NUMBER;
michael@0:     }
michael@0:     while(start<limit) {
michael@0:         c=s.charAt(start++);
michael@0:         if(0x30<=c && c<=0x39) {
michael@0:             if(number>=INT32_MAX/10) {
michael@0:                 badNumber=TRUE;  // overflow
michael@0:             }
michael@0:             number=number*10+(c-0x30);
michael@0:         } else {
michael@0:             return UMSGPAT_ARG_NAME_NOT_NUMBER;
michael@0:         }
michael@0:     }
michael@0:     // There are only ASCII digits.
michael@0:     if(badNumber) {
michael@0:         return UMSGPAT_ARG_NAME_NOT_VALID;
michael@0:     } else {
michael@0:         return number;
michael@0:     }
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::parseDouble(int32_t start, int32_t limit, UBool allowInfinity,
michael@0:                             UParseError *parseError, UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return;
michael@0:     }
michael@0:     U_ASSERT(start<limit);
michael@0:     // fake loop for easy exit and single throw statement
michael@0:     for(;;) { /*loop doesn't iterate*/
michael@0:         // fast path for small integers and infinity
michael@0:         int32_t value=0;
michael@0:         int32_t isNegative=0;  // not boolean so that we can easily add it to value
michael@0:         int32_t index=start;
michael@0:         UChar c=msg.charAt(index++);
michael@0:         if(c==u_minus) {
michael@0:             isNegative=1;
michael@0:             if(index==limit) {
michael@0:                 break;  // no number
michael@0:             }
michael@0:             c=msg.charAt(index++);
michael@0:         } else if(c==u_plus) {
michael@0:             if(index==limit) {
michael@0:                 break;  // no number
michael@0:             }
michael@0:             c=msg.charAt(index++);
michael@0:         }
michael@0:         if(c==0x221e) {  // infinity
michael@0:             if(allowInfinity && index==limit) {
michael@0:                 double infinity=uprv_getInfinity();
michael@0:                 addArgDoublePart(
michael@0:                     isNegative!=0 ? -infinity : infinity,
michael@0:                     start, limit-start, errorCode);
michael@0:                 return;
michael@0:             } else {
michael@0:                 break;
michael@0:             }
michael@0:         }
michael@0:         // try to parse the number as a small integer but fall back to a double
michael@0:         while('0'<=c && c<='9') {
michael@0:             value=value*10+(c-'0');
michael@0:             if(value>(Part::MAX_VALUE+isNegative)) {
michael@0:                 break;  // not a small-enough integer
michael@0:             }
michael@0:             if(index==limit) {
michael@0:                 addPart(UMSGPAT_PART_TYPE_ARG_INT, start, limit-start,
michael@0:                         isNegative!=0 ? -value : value, errorCode);
michael@0:                 return;
michael@0:             }
michael@0:             c=msg.charAt(index++);
michael@0:         }
michael@0:         // Let Double.parseDouble() throw a NumberFormatException.
michael@0:         char numberChars[128];
michael@0:         int32_t capacity=(int32_t)sizeof(numberChars);
michael@0:         int32_t length=limit-start;
michael@0:         if(length>=capacity) {
michael@0:             break;  // number too long
michael@0:         }
michael@0:         msg.extract(start, length, numberChars, capacity, US_INV);
michael@0:         if((int32_t)uprv_strlen(numberChars)<length) {
michael@0:             break;  // contains non-invariant character that was turned into NUL
michael@0:         }
michael@0:         char *end;
michael@0:         double numericValue=uprv_strtod(numberChars, &end);
michael@0:         if(end!=(numberChars+length)) {
michael@0:             break;  // parsing error
michael@0:         }
michael@0:         addArgDoublePart(numericValue, start, length, errorCode);
michael@0:         return;
michael@0:     }
michael@0:     setParseError(parseError, start /*, limit*/);  // Bad syntax for numeric value.
michael@0:     errorCode=U_PATTERN_SYNTAX_ERROR;
michael@0:     return;
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::skipWhiteSpace(int32_t index) {
michael@0:     const UChar *s=msg.getBuffer();
michael@0:     int32_t msgLength=msg.length();
michael@0:     const UChar *t=PatternProps::skipWhiteSpace(s+index, msgLength-index);
michael@0:     return (int32_t)(t-s);
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::skipIdentifier(int32_t index) {
michael@0:     const UChar *s=msg.getBuffer();
michael@0:     int32_t msgLength=msg.length();
michael@0:     const UChar *t=PatternProps::skipIdentifier(s+index, msgLength-index);
michael@0:     return (int32_t)(t-s);
michael@0: }
michael@0: 
michael@0: int32_t
michael@0: MessagePattern::skipDouble(int32_t index) {
michael@0:     int32_t msgLength=msg.length();
michael@0:     while(index<msgLength) {
michael@0:         UChar c=msg.charAt(index);
michael@0:         // U+221E: Allow the infinity symbol, for ChoiceFormat patterns.
michael@0:         if((c<0x30 && c!=u_plus && c!=u_minus && c!=u_dot) || (c>0x39 && c!=u_e && c!=u_E && c!=0x221e)) {
michael@0:             break;
michael@0:         }
michael@0:         ++index;
michael@0:     }
michael@0:     return index;
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::isArgTypeChar(UChar32 c) {
michael@0:     return (u_a<=c && c<=u_z) || (u_A<=c && c<=u_Z);
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::isChoice(int32_t index) {
michael@0:     UChar c;
michael@0:     return
michael@0:         ((c=msg.charAt(index++))==u_c || c==u_C) &&
michael@0:         ((c=msg.charAt(index++))==u_h || c==u_H) &&
michael@0:         ((c=msg.charAt(index++))==u_o || c==u_O) &&
michael@0:         ((c=msg.charAt(index++))==u_i || c==u_I) &&
michael@0:         ((c=msg.charAt(index++))==u_c || c==u_C) &&
michael@0:         ((c=msg.charAt(index))==u_e || c==u_E);
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::isPlural(int32_t index) {
michael@0:     UChar c;
michael@0:     return
michael@0:         ((c=msg.charAt(index++))==u_p || c==u_P) &&
michael@0:         ((c=msg.charAt(index++))==u_l || c==u_L) &&
michael@0:         ((c=msg.charAt(index++))==u_u || c==u_U) &&
michael@0:         ((c=msg.charAt(index++))==u_r || c==u_R) &&
michael@0:         ((c=msg.charAt(index++))==u_a || c==u_A) &&
michael@0:         ((c=msg.charAt(index))==u_l || c==u_L);
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::isSelect(int32_t index) {
michael@0:     UChar c;
michael@0:     return
michael@0:         ((c=msg.charAt(index++))==u_s || c==u_S) &&
michael@0:         ((c=msg.charAt(index++))==u_e || c==u_E) &&
michael@0:         ((c=msg.charAt(index++))==u_l || c==u_L) &&
michael@0:         ((c=msg.charAt(index++))==u_e || c==u_E) &&
michael@0:         ((c=msg.charAt(index++))==u_c || c==u_C) &&
michael@0:         ((c=msg.charAt(index))==u_t || c==u_T);
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::isOrdinal(int32_t index) {
michael@0:     UChar c;
michael@0:     return
michael@0:         ((c=msg.charAt(index++))==u_o || c==u_O) &&
michael@0:         ((c=msg.charAt(index++))==u_r || c==u_R) &&
michael@0:         ((c=msg.charAt(index++))==u_d || c==u_D) &&
michael@0:         ((c=msg.charAt(index++))==u_i || c==u_I) &&
michael@0:         ((c=msg.charAt(index++))==u_n || c==u_N) &&
michael@0:         ((c=msg.charAt(index++))==u_a || c==u_A) &&
michael@0:         ((c=msg.charAt(index))==u_l || c==u_L);
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::inMessageFormatPattern(int32_t nestingLevel) {
michael@0:     return nestingLevel>0 || partsList->a[0].type==UMSGPAT_PART_TYPE_MSG_START;
michael@0: }
michael@0: 
michael@0: UBool
michael@0: MessagePattern::inTopLevelChoiceMessage(int32_t nestingLevel, UMessagePatternArgType parentType) {
michael@0:     return
michael@0:         nestingLevel==1 &&
michael@0:         parentType==UMSGPAT_ARG_TYPE_CHOICE &&
michael@0:         partsList->a[0].type!=UMSGPAT_PART_TYPE_MSG_START;
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::addPart(UMessagePatternPartType type, int32_t index, int32_t length,
michael@0:                         int32_t value, UErrorCode &errorCode) {
michael@0:     if(partsList->ensureCapacityForOneMore(partsLength, errorCode)) {
michael@0:         Part &part=partsList->a[partsLength++];
michael@0:         part.type=type;
michael@0:         part.index=index;
michael@0:         part.length=(uint16_t)length;
michael@0:         part.value=(int16_t)value;
michael@0:         part.limitPartIndex=0;
michael@0:     }
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::addLimitPart(int32_t start,
michael@0:                              UMessagePatternPartType type, int32_t index, int32_t length,
michael@0:                              int32_t value, UErrorCode &errorCode) {
michael@0:     partsList->a[start].limitPartIndex=partsLength;
michael@0:     addPart(type, index, length, value, errorCode);
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::addArgDoublePart(double numericValue, int32_t start, int32_t length,
michael@0:                                  UErrorCode &errorCode) {
michael@0:     if(U_FAILURE(errorCode)) {
michael@0:         return;
michael@0:     }
michael@0:     int32_t numericIndex=numericValuesLength;
michael@0:     if(numericValuesList==NULL) {
michael@0:         numericValuesList=new MessagePatternDoubleList();
michael@0:         if(numericValuesList==NULL) {
michael@0:             errorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:             return;
michael@0:         }
michael@0:     } else if(!numericValuesList->ensureCapacityForOneMore(numericValuesLength, errorCode)) {
michael@0:         return;
michael@0:     } else {
michael@0:         if(numericIndex>Part::MAX_VALUE) {
michael@0:             errorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             return;
michael@0:         }
michael@0:     }
michael@0:     numericValuesList->a[numericValuesLength++]=numericValue;
michael@0:     addPart(UMSGPAT_PART_TYPE_ARG_DOUBLE, start, length, numericIndex, errorCode);
michael@0: }
michael@0: 
michael@0: void
michael@0: MessagePattern::setParseError(UParseError *parseError, int32_t index) {
michael@0:     if(parseError==NULL) {
michael@0:         return;
michael@0:     }
michael@0:     parseError->offset=index;
michael@0: 
michael@0:     // Set preContext to some of msg before index.
michael@0:     // Avoid splitting a surrogate pair.
michael@0:     int32_t length=index;
michael@0:     if(length>=U_PARSE_CONTEXT_LEN) {
michael@0:         length=U_PARSE_CONTEXT_LEN-1;
michael@0:         if(length>0 && U16_IS_TRAIL(msg[index-length])) {
michael@0:             --length;
michael@0:         }
michael@0:     }
michael@0:     msg.extract(index-length, length, parseError->preContext);
michael@0:     parseError->preContext[length]=0;
michael@0: 
michael@0:     // Set postContext to some of msg starting at index.
michael@0:     length=msg.length()-index;
michael@0:     if(length>=U_PARSE_CONTEXT_LEN) {
michael@0:         length=U_PARSE_CONTEXT_LEN-1;
michael@0:         if(length>0 && U16_IS_LEAD(msg[index+length-1])) {
michael@0:             --length;
michael@0:         }
michael@0:     }
michael@0:     msg.extract(index, length, parseError->postContext);
michael@0:     parseError->postContext[length]=0;
michael@0: }
michael@0: 
michael@0: // MessageImpl ------------------------------------------------------------- ***
michael@0: 
michael@0: void
michael@0: MessageImpl::appendReducedApostrophes(const UnicodeString &s, int32_t start, int32_t limit,
michael@0:                                       UnicodeString &sb) {
michael@0:     int32_t doubleApos=-1;
michael@0:     for(;;) {
michael@0:         int32_t i=s.indexOf(u_apos, start);
michael@0:         if(i<0 || i>=limit) {
michael@0:             sb.append(s, start, limit-start);
michael@0:             break;
michael@0:         }
michael@0:         if(i==doubleApos) {
michael@0:             // Double apostrophe at start-1 and start==i, append one.
michael@0:             sb.append(u_apos);
michael@0:             ++start;
michael@0:             doubleApos=-1;
michael@0:         } else {
michael@0:             // Append text between apostrophes and skip this one.
michael@0:             sb.append(s, start, i-start);
michael@0:             doubleApos=start=i+1;
michael@0:         }
michael@0:     }
michael@0: }
michael@0: 
michael@0: // Ported from second half of ICU4J SelectFormat.format(String).
michael@0: UnicodeString &
michael@0: MessageImpl::appendSubMessageWithoutSkipSyntax(const MessagePattern &msgPattern,
michael@0:                                                int32_t msgStart,
michael@0:                                                UnicodeString &result) {
michael@0:     const UnicodeString &msgString=msgPattern.getPatternString();
michael@0:     int32_t prevIndex=msgPattern.getPart(msgStart).getLimit();
michael@0:     for(int32_t i=msgStart;;) {
michael@0:         const MessagePattern::Part &part=msgPattern.getPart(++i);
michael@0:         UMessagePatternPartType type=part.getType();
michael@0:         int32_t index=part.getIndex();
michael@0:         if(type==UMSGPAT_PART_TYPE_MSG_LIMIT) {
michael@0:             return result.append(msgString, prevIndex, index-prevIndex);
michael@0:         } else if(type==UMSGPAT_PART_TYPE_SKIP_SYNTAX) {
michael@0:             result.append(msgString, prevIndex, index-prevIndex);
michael@0:             prevIndex=part.getLimit();
michael@0:         } else if(type==UMSGPAT_PART_TYPE_ARG_START) {
michael@0:             result.append(msgString, prevIndex, index-prevIndex);
michael@0:             prevIndex=index;
michael@0:             i=msgPattern.getLimitPartIndex(i);
michael@0:             index=msgPattern.getPart(i).getLimit();
michael@0:             appendReducedApostrophes(msgString, prevIndex, index, result);
michael@0:             prevIndex=index;
michael@0:         }
michael@0:     }
michael@0: }
michael@0: 
michael@0: U_NAMESPACE_END
michael@0: 
michael@0: #endif  // !UCONFIG_NO_FORMATTING