X-Git-Url: https://git.saurik.com/apple/icu.git/blobdiff_plain/374ca955a76ecab1204ca8bfa63ff9238d998416..51004dcb01e06fef634b61be77ed73dd61cb6db9:/icuSources/i18n/choicfmt.cpp diff --git a/icuSources/i18n/choicfmt.cpp b/icuSources/i18n/choicfmt.cpp index eeac7f38..79a722bd 100644 --- a/icuSources/i18n/choicfmt.cpp +++ b/icuSources/i18n/choicfmt.cpp @@ -1,6 +1,6 @@ /* ******************************************************************************* -* Copyright (C) 1997-2004, International Business Machines Corporation and * +* Copyright (C) 1997-2012, International Business Machines Corporation and * * others. All Rights Reserved. * ******************************************************************************* * @@ -31,9 +31,12 @@ #include "unicode/numfmt.h" #include "unicode/locid.h" #include "cpputils.h" -#include "ustrfmt.h" #include "cstring.h" +#include "messageimpl.h" #include "putilimp.h" +#include "uassert.h" +#include +#include // ***************************************************************************** // class ChoiceFormat @@ -52,10 +55,17 @@ UOBJECT_DEFINE_RTTI_IMPLEMENTATION(ChoiceFormat) #define LESS_EQUAL2 ((UChar)0x2264) #define VERTICAL_BAR ((UChar)0x007C) /*|*/ #define MINUS ((UChar)0x002D) /*-*/ + +static const UChar LEFT_CURLY_BRACE = 0x7B; /*{*/ +static const UChar RIGHT_CURLY_BRACE = 0x7D; /*}*/ + +#ifdef INFINITY +#undef INFINITY +#endif #define INFINITY ((UChar)0x221E) -static const UChar gPositiveInfinity[] = {INFINITY, 0}; -static const UChar gNegativeInfinity[] = {MINUS, INFINITY, 0}; +//static const UChar gPositiveInfinity[] = {INFINITY, 0}; +//static const UChar gNegativeInfinity[] = {MINUS, INFINITY, 0}; #define POSITIVE_INF_STRLEN 1 #define NEGATIVE_INF_STRLEN 2 @@ -64,10 +74,8 @@ static const UChar gNegativeInfinity[] = {MINUS, INFINITY, 0}; ChoiceFormat::ChoiceFormat(const UnicodeString& newPattern, UErrorCode& status) -: fChoiceLimits(0), - fClosures(0), - fChoiceFormats(0), - fCount(0) +: constructorErrorCode(status), + msgPattern(status) { applyPattern(newPattern, status); } @@ -79,12 +87,10 @@ ChoiceFormat::ChoiceFormat(const UnicodeString& newPattern, ChoiceFormat::ChoiceFormat(const double* limits, const UnicodeString* formats, int32_t cnt ) -: fChoiceLimits(0), - fClosures(0), - fChoiceFormats(0), - fCount(0) +: constructorErrorCode(U_ZERO_ERROR), + msgPattern(constructorErrorCode) { - setChoices(limits, formats, cnt ); + setChoices(limits, NULL, formats, cnt, constructorErrorCode); } // ------------------------------------- @@ -93,12 +99,10 @@ ChoiceFormat::ChoiceFormat(const double* limits, const UBool* closures, const UnicodeString* formats, int32_t cnt ) -: fChoiceLimits(0), - fClosures(0), - fChoiceFormats(0), - fCount(0) +: constructorErrorCode(U_ZERO_ERROR), + msgPattern(constructorErrorCode) { - setChoices(limits, closures, formats, cnt ); + setChoices(limits, closures, formats, cnt, constructorErrorCode); } // ------------------------------------- @@ -106,11 +110,9 @@ ChoiceFormat::ChoiceFormat(const double* limits, ChoiceFormat::ChoiceFormat(const ChoiceFormat& that) : NumberFormat(that), - fChoiceLimits(0), - fClosures(0), - fChoiceFormats(0) + constructorErrorCode(that.constructorErrorCode), + msgPattern(that.msgPattern) { - *this = that; } // ------------------------------------- @@ -121,10 +123,8 @@ ChoiceFormat::ChoiceFormat(const ChoiceFormat& that) ChoiceFormat::ChoiceFormat(const UnicodeString& newPattern, UParseError& parseError, UErrorCode& status) -: fChoiceLimits(0), - fClosures(0), - fChoiceFormats(0), - fCount(0) +: constructorErrorCode(status), + msgPattern(status) { applyPattern(newPattern,parseError, status); } @@ -136,16 +136,7 @@ ChoiceFormat::operator==(const Format& that) const if (this == &that) return TRUE; if (!NumberFormat::operator==(that)) return FALSE; ChoiceFormat& thatAlias = (ChoiceFormat&)that; - if (fCount != thatAlias.fCount) return FALSE; - // Checks the limits, the corresponding format string and LE or LT flags. - // LE means less than and equal to, LT means less than. - for (int32_t i = 0; i < fCount; i++) { - if ((fChoiceLimits[i] != thatAlias.fChoiceLimits[i]) || - (fClosures[i] != thatAlias.fClosures[i]) || - (fChoiceFormats[i] != thatAlias.fChoiceFormats[i])) - return FALSE; - } - return TRUE; + return msgPattern == thatAlias.msgPattern; } // ------------------------------------- @@ -156,21 +147,8 @@ ChoiceFormat::operator=(const ChoiceFormat& that) { if (this != &that) { NumberFormat::operator=(that); - fCount = that.fCount; - uprv_free(fChoiceLimits); - fChoiceLimits = NULL; - uprv_free(fClosures); - fClosures = NULL; - delete [] fChoiceFormats; - fChoiceFormats = NULL; - - fChoiceLimits = (double*) uprv_malloc( sizeof(double) * fCount); - fClosures = (UBool*) uprv_malloc( sizeof(UBool) * fCount); - fChoiceFormats = new UnicodeString[fCount]; - - uprv_arrayCopy(that.fChoiceLimits, fChoiceLimits, fCount); - uprv_arrayCopy(that.fClosures, fClosures, fCount); - uprv_arrayCopy(that.fChoiceFormats, fChoiceFormats, fCount); + constructorErrorCode = that.constructorErrorCode; + msgPattern = that.msgPattern; } return *this; } @@ -179,40 +157,61 @@ ChoiceFormat::operator=(const ChoiceFormat& that) ChoiceFormat::~ChoiceFormat() { - uprv_free(fChoiceLimits); - fChoiceLimits = NULL; - uprv_free(fClosures); - fClosures = NULL; - delete [] fChoiceFormats; - fChoiceFormats = NULL; - fCount = 0; -} - -/** - * Convert a string to a double value - */ -double -ChoiceFormat::stod(const UnicodeString& string) -{ - char source[256]; - char* end; - - string.extract(0, string.length(), source, (int32_t)sizeof(source), US_INV); /* invariant codepage */ - return uprv_strtod(source,&end); } // ------------------------------------- /** - * Convert a double value to a string + * Convert a double value to a string without the overhead of NumberFormat. */ UnicodeString& ChoiceFormat::dtos(double value, UnicodeString& string) { - char temp[256]; + /* Buffer to contain the digits and any extra formatting stuff. */ + char temp[DBL_DIG + 16]; + char *itrPtr = temp; + char *expPtr; + + sprintf(temp, "%.*g", DBL_DIG, value); + + /* Find and convert the decimal point. + Using setlocale on some machines will cause sprintf to use a comma for certain locales. + */ + while (*itrPtr && (*itrPtr == '-' || isdigit(*itrPtr))) { + itrPtr++; + } + if (*itrPtr != 0 && *itrPtr != 'e') { + /* We reached something that looks like a decimal point. + In case someone used setlocale(), which changes the decimal point. */ + *itrPtr = '.'; + itrPtr++; + } + /* Search for the exponent */ + while (*itrPtr && *itrPtr != 'e') { + itrPtr++; + } + if (*itrPtr == 'e') { + itrPtr++; + /* Verify the exponent sign */ + if (*itrPtr == '+' || *itrPtr == '-') { + itrPtr++; + } + /* Remove leading zeros. You will see this on Windows machines. */ + expPtr = itrPtr; + while (*itrPtr == '0') { + itrPtr++; + } + if (*itrPtr && expPtr != itrPtr) { + /* Shift the exponent without zeros. */ + while (*itrPtr) { + *(expPtr++) = *(itrPtr++); + } + // NULL terminate + *expPtr = 0; + } + } - uprv_dtostr(value, temp, 3, TRUE); string = UnicodeString(temp, -1, US_INV); /* invariant codepage */ return string; } @@ -224,8 +223,8 @@ void ChoiceFormat::applyPattern(const UnicodeString& pattern, UErrorCode& status) { - UParseError parseError; - applyPattern(pattern, parseError, status); + msgPattern.parseChoiceStyle(pattern, NULL, status); + constructorErrorCode = status; } // ------------------------------------- @@ -236,261 +235,18 @@ ChoiceFormat::applyPattern(const UnicodeString& pattern, UParseError& parseError, UErrorCode& status) { - if (U_FAILURE(status)) - { - return; - } - - // Clear error struct - parseError.offset = -1; - parseError.preContext[0] = parseError.postContext[0] = (UChar)0; - - // Perform 2 passes. The first computes the number of limits in - // this pattern (fCount), which is 1 more than the number of - // literal VERTICAL_BAR characters. - int32_t count = 1; - int32_t i; - for (i=0; i 0 && limit <= newLimits[k-1]) { - // Each limit must be strictly > than the previous - // limit. One exception: Two subsequent limits may be - // == if the first closure is FALSE and the second - // closure is TRUE. This places the limit value in - // the second interval. - if (!(limit == newLimits[k-1] && - !newClosures[k-1] && - newClosures[k])) { - goto error; - } - } - - buf.truncate(0); - } else if (c == VERTICAL_BAR) { - if (inNumber) { - goto error; - } - inNumber = TRUE; - - newFormats[k] = buf; - ++k; - buf.truncate(0); - } else { - buf += c; - } - } - - if (k != (count-1) || inNumber || inQuote) { - goto error; - } - newFormats[k] = buf; - - // Don't modify this object until the parse succeeds - uprv_free(fChoiceLimits); - uprv_free(fClosures); - delete[] fChoiceFormats; - fCount = count; - fChoiceLimits = newLimits; - fClosures = newClosures; - fChoiceFormats = newFormats; - return; - -error: - status = U_ILLEGAL_ARGUMENT_ERROR; - syntaxError(pattern,i,parseError); - uprv_free(newLimits); - uprv_free(newClosures); - delete[] newFormats; - return; - + msgPattern.parseChoiceStyle(pattern, &parseError, status); + constructorErrorCode = status; } // ------------------------------------- -// Reconstruct the original input pattern. +// Returns the input pattern string. UnicodeString& ChoiceFormat::toPattern(UnicodeString& result) const { - result.remove(); - for (int32_t i = 0; i < fCount; ++i) { - if (i != 0) { - result += VERTICAL_BAR; - } - UnicodeString buf; - if (uprv_isPositiveInfinity(fChoiceLimits[i])) { - result += INFINITY; - } else if (uprv_isNegativeInfinity(fChoiceLimits[i])) { - result += MINUS; - result += INFINITY; - } else { - result += dtos(fChoiceLimits[i], buf); - } - if (fClosures[i]) { - result += LESS_THAN; - } else { - result += LESS_EQUAL; - } - // Append fChoiceFormats[i], using quotes if there are special - // characters. Single quotes themselves must be escaped in - // either case. - const UnicodeString& text = fChoiceFormats[i]; - UBool needQuote = text.indexOf(LESS_THAN) >= 0 - || text.indexOf(LESS_EQUAL) >= 0 - || text.indexOf(LESS_EQUAL2) >= 0 - || text.indexOf(VERTICAL_BAR) >= 0; - if (needQuote) { - result += SINGLE_QUOTE; - } - if (text.indexOf(SINGLE_QUOTE) < 0) { - result += text; - } - else { - for (int32_t j = 0; j < text.length(); ++j) { - UChar c = text[j]; - result += c; - if (c == SINGLE_QUOTE) { - result += c; - } - } - } - if (needQuote) { - result += SINGLE_QUOTE; - } - } - - return result; + return result = msgPattern.getPatternString(); } -#ifdef U_USE_CHOICE_FORMAT_DEPRECATES -// ------------------------------------- -// Adopts the limit and format arrays. - -void -ChoiceFormat::adoptChoices(double *limits, - UnicodeString *formats, - int32_t cnt ) -{ - adoptChoices(limits, (UBool *)0, formats, cnt); -} - -// ------------------------------------- -// Adopts the limit and format arrays. - -void -ChoiceFormat::adoptChoices(double *limits, - UBool *closures, - UnicodeString *formats, - int32_t cnt ) -{ - if(limits == 0 || formats == 0) - return; - - uprv_free(fChoiceLimits); - uprv_free(fClosures); - delete [] fChoiceFormats; - fChoiceLimits = limits; - fClosures = closures; - fChoiceFormats = formats; - fCount = cnt; - - if (fClosures == 0) { - fClosures = (UBool*) uprv_malloc( sizeof(UBool) * fCount); - int32_t i; - for (i=0; i '|' + // |' -> '|''' + // |'' -> '|''''' etc. + result.append(SINGLE_QUOTE).append(c).append(SINGLE_QUOTE); + continue; // Skip the append(c) at the end of the loop body. + } else if (c == LEFT_CURLY_BRACE) { + ++nestingLevel; + } else if (c == RIGHT_CURLY_BRACE && nestingLevel > 0) { + --nestingLevel; + } + result.append(c); } } + // Apply the reconstructed pattern. + applyPattern(result, errorCode); } // ------------------------------------- @@ -542,8 +344,8 @@ ChoiceFormat::setChoices( const double* limits, const double* ChoiceFormat::getLimits(int32_t& cnt) const { - cnt = fCount; - return fChoiceLimits; + cnt = 0; + return NULL; } // ------------------------------------- @@ -552,8 +354,8 @@ ChoiceFormat::getLimits(int32_t& cnt) const const UBool* ChoiceFormat::getClosures(int32_t& cnt) const { - cnt = fCount; - return fClosures; + cnt = 0; + return NULL; } // ------------------------------------- @@ -562,8 +364,8 @@ ChoiceFormat::getClosures(int32_t& cnt) const const UnicodeString* ChoiceFormat::getFormats(int32_t& cnt) const { - cnt = fCount; - return fChoiceFormats; + cnt = 0; + return NULL; } // ------------------------------------- @@ -580,9 +382,8 @@ ChoiceFormat::format(int64_t number, } // ------------------------------------- -// Formats a long number, it's actually formatted as -// a double. The returned format string may differ -// from the input number because of this. +// Formats an int32_t number, it's actually formatted as +// a double. UnicodeString& ChoiceFormat::format(int32_t number, @@ -600,26 +401,63 @@ ChoiceFormat::format(double number, UnicodeString& appendTo, FieldPosition& /*pos*/) const { - // find the number - int32_t i; - for (i = 0; i < fCount; ++i) { - if (fClosures[i]) { - if (!(number > fChoiceLimits[i])) { - // same as number <= fChoiceLimits, except catches NaN - break; - } - } else if (!(number >= fChoiceLimits[i])) { - // same as number < fChoiceLimits, except catches NaN + if (msgPattern.countParts() == 0) { + // No pattern was applied, or it failed. + return appendTo; + } + // Get the appropriate sub-message. + int32_t msgStart = findSubMessage(msgPattern, 0, number); + if (!MessageImpl::jdkAposMode(msgPattern)) { + int32_t patternStart = msgPattern.getPart(msgStart).getLimit(); + int32_t msgLimit = msgPattern.getLimitPartIndex(msgStart); + appendTo.append(msgPattern.getPatternString(), + patternStart, + msgPattern.getPatternIndex(msgLimit) - patternStart); + return appendTo; + } + // JDK compatibility mode: Remove SKIP_SYNTAX. + return MessageImpl::appendSubMessageWithoutSkipSyntax(msgPattern, msgStart, appendTo); +} + +int32_t +ChoiceFormat::findSubMessage(const MessagePattern &pattern, int32_t partIndex, double number) { + int32_t count = pattern.countParts(); + int32_t msgStart; + // Iterate over (ARG_INT|DOUBLE, ARG_SELECTOR, message) tuples + // until ARG_LIMIT or end of choice-only pattern. + // Ignore the first number and selector and start the loop on the first message. + partIndex += 2; + for (;;) { + // Skip but remember the current sub-message. + msgStart = partIndex; + partIndex = pattern.getLimitPartIndex(partIndex); + if (++partIndex >= count) { + // Reached the end of the choice-only pattern. + // Return with the last sub-message. + break; + } + const MessagePattern::Part &part = pattern.getPart(partIndex++); + UMessagePatternPartType type = part.getType(); + if (type == UMSGPAT_PART_TYPE_ARG_LIMIT) { + // Reached the end of the ChoiceFormat style. + // Return with the last sub-message. + break; + } + // part is an ARG_INT or ARG_DOUBLE + U_ASSERT(MessagePattern::Part::hasNumericValue(type)); + double boundary = pattern.getNumericValue(part); + // Fetch the ARG_SELECTOR character. + int32_t selectorIndex = pattern.getPatternIndex(partIndex++); + UChar boundaryChar = pattern.getPatternString().charAt(selectorIndex); + if (boundaryChar == LESS_THAN ? !(number > boundary) : !(number >= boundary)) { + // The number is in the interval between the previous boundary and the current one. + // Return with the sub-message between them. + // The !(a>b) and !(a>=b) comparisons are equivalent to + // (a<=b) and (a furthest) { - furthest = status.getIndex(); + int32_t count = pattern.countParts(); + while (partIndex < count && pattern.getPartType(partIndex) != UMSGPAT_PART_TYPE_ARG_LIMIT) { + tempNumber = pattern.getNumericValue(pattern.getPart(partIndex)); + partIndex += 2; // skip the numeric part and ignore the ARG_SELECTOR + int32_t msgLimit = pattern.getLimitPartIndex(partIndex); + int32_t len = matchStringUntilLimitPart(pattern, partIndex, msgLimit, source, start); + if (len >= 0) { + int32_t newIndex = start + len; + if (newIndex > furthest) { + furthest = newIndex; bestNumber = tempNumber; - if (furthest == text.length()) + if (furthest == source.length()) { break; + } } } + partIndex = msgLimit + 1; } - status.setIndex(furthest); - if (status.getIndex() == start) { - status.setErrorIndex(furthest); + if (furthest == start) { + pos.setErrorIndex(start); + } else { + pos.setIndex(furthest); + } + return bestNumber; +} + +int32_t +ChoiceFormat::matchStringUntilLimitPart( + const MessagePattern &pattern, int32_t partIndex, int32_t limitPartIndex, + const UnicodeString &source, int32_t sourceOffset) { + int32_t matchingSourceLength = 0; + const UnicodeString &msgString = pattern.getPatternString(); + int32_t prevIndex = pattern.getPart(partIndex).getLimit(); + for (;;) { + const MessagePattern::Part &part = pattern.getPart(++partIndex); + if (partIndex == limitPartIndex || part.getType() == UMSGPAT_PART_TYPE_SKIP_SYNTAX) { + int32_t index = part.getIndex(); + int32_t length = index - prevIndex; + if (length != 0 && 0 != source.compare(sourceOffset, length, msgString, prevIndex, length)) { + return -1; // mismatch + } + matchingSourceLength += length; + if (partIndex == limitPartIndex) { + return matchingSourceLength; + } + prevIndex = part.getLimit(); // SKIP_SYNTAX + } } - result.setDouble(bestNumber); } // -------------------------------------