1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/icu/source/i18n/choicfmt.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,575 @@
1.4 +/*
1.5 +*******************************************************************************
1.6 +* Copyright (C) 1997-2013, International Business Machines Corporation and *
1.7 +* others. All Rights Reserved. *
1.8 +*******************************************************************************
1.9 +*
1.10 +* File CHOICFMT.CPP
1.11 +*
1.12 +* Modification History:
1.13 +*
1.14 +* Date Name Description
1.15 +* 02/19/97 aliu Converted from java.
1.16 +* 03/20/97 helena Finished first cut of implementation and got rid
1.17 +* of nextDouble/previousDouble and replaced with
1.18 +* boolean array.
1.19 +* 4/10/97 aliu Clean up. Modified to work on AIX.
1.20 +* 06/04/97 helena Fixed applyPattern(), toPattern() and not to include
1.21 +* wchar.h.
1.22 +* 07/09/97 helena Made ParsePosition into a class.
1.23 +* 08/06/97 nos removed overloaded constructor, fixed 'format(array)'
1.24 +* 07/22/98 stephen JDK 1.2 Sync - removed UBool array (doubleFlags)
1.25 +* 02/22/99 stephen Removed character literals for EBCDIC safety
1.26 +********************************************************************************
1.27 +*/
1.28 +
1.29 +#include "unicode/utypes.h"
1.30 +
1.31 +#if !UCONFIG_NO_FORMATTING
1.32 +
1.33 +#include "unicode/choicfmt.h"
1.34 +#include "unicode/numfmt.h"
1.35 +#include "unicode/locid.h"
1.36 +#include "cpputils.h"
1.37 +#include "cstring.h"
1.38 +#include "messageimpl.h"
1.39 +#include "putilimp.h"
1.40 +#include "uassert.h"
1.41 +#include <stdio.h>
1.42 +#include <float.h>
1.43 +
1.44 +// *****************************************************************************
1.45 +// class ChoiceFormat
1.46 +// *****************************************************************************
1.47 +
1.48 +U_NAMESPACE_BEGIN
1.49 +
1.50 +UOBJECT_DEFINE_RTTI_IMPLEMENTATION(ChoiceFormat)
1.51 +
1.52 +// Special characters used by ChoiceFormat. There are two characters
1.53 +// used interchangeably to indicate <=. Either is parsed, but only
1.54 +// LESS_EQUAL is generated by toPattern().
1.55 +#define SINGLE_QUOTE ((UChar)0x0027) /*'*/
1.56 +#define LESS_THAN ((UChar)0x003C) /*<*/
1.57 +#define LESS_EQUAL ((UChar)0x0023) /*#*/
1.58 +#define LESS_EQUAL2 ((UChar)0x2264)
1.59 +#define VERTICAL_BAR ((UChar)0x007C) /*|*/
1.60 +#define MINUS ((UChar)0x002D) /*-*/
1.61 +
1.62 +static const UChar LEFT_CURLY_BRACE = 0x7B; /*{*/
1.63 +static const UChar RIGHT_CURLY_BRACE = 0x7D; /*}*/
1.64 +
1.65 +#ifdef INFINITY
1.66 +#undef INFINITY
1.67 +#endif
1.68 +#define INFINITY ((UChar)0x221E)
1.69 +
1.70 +//static const UChar gPositiveInfinity[] = {INFINITY, 0};
1.71 +//static const UChar gNegativeInfinity[] = {MINUS, INFINITY, 0};
1.72 +#define POSITIVE_INF_STRLEN 1
1.73 +#define NEGATIVE_INF_STRLEN 2
1.74 +
1.75 +// -------------------------------------
1.76 +// Creates a ChoiceFormat instance based on the pattern.
1.77 +
1.78 +ChoiceFormat::ChoiceFormat(const UnicodeString& newPattern,
1.79 + UErrorCode& status)
1.80 +: constructorErrorCode(status),
1.81 + msgPattern(status)
1.82 +{
1.83 + applyPattern(newPattern, status);
1.84 +}
1.85 +
1.86 +// -------------------------------------
1.87 +// Creates a ChoiceFormat instance with the limit array and
1.88 +// format strings for each limit.
1.89 +
1.90 +ChoiceFormat::ChoiceFormat(const double* limits,
1.91 + const UnicodeString* formats,
1.92 + int32_t cnt )
1.93 +: constructorErrorCode(U_ZERO_ERROR),
1.94 + msgPattern(constructorErrorCode)
1.95 +{
1.96 + setChoices(limits, NULL, formats, cnt, constructorErrorCode);
1.97 +}
1.98 +
1.99 +// -------------------------------------
1.100 +
1.101 +ChoiceFormat::ChoiceFormat(const double* limits,
1.102 + const UBool* closures,
1.103 + const UnicodeString* formats,
1.104 + int32_t cnt )
1.105 +: constructorErrorCode(U_ZERO_ERROR),
1.106 + msgPattern(constructorErrorCode)
1.107 +{
1.108 + setChoices(limits, closures, formats, cnt, constructorErrorCode);
1.109 +}
1.110 +
1.111 +// -------------------------------------
1.112 +// copy constructor
1.113 +
1.114 +ChoiceFormat::ChoiceFormat(const ChoiceFormat& that)
1.115 +: NumberFormat(that),
1.116 + constructorErrorCode(that.constructorErrorCode),
1.117 + msgPattern(that.msgPattern)
1.118 +{
1.119 +}
1.120 +
1.121 +// -------------------------------------
1.122 +// Private constructor that creates a
1.123 +// ChoiceFormat instance based on the
1.124 +// pattern and populates UParseError
1.125 +
1.126 +ChoiceFormat::ChoiceFormat(const UnicodeString& newPattern,
1.127 + UParseError& parseError,
1.128 + UErrorCode& status)
1.129 +: constructorErrorCode(status),
1.130 + msgPattern(status)
1.131 +{
1.132 + applyPattern(newPattern,parseError, status);
1.133 +}
1.134 +// -------------------------------------
1.135 +
1.136 +UBool
1.137 +ChoiceFormat::operator==(const Format& that) const
1.138 +{
1.139 + if (this == &that) return TRUE;
1.140 + if (!NumberFormat::operator==(that)) return FALSE;
1.141 + ChoiceFormat& thatAlias = (ChoiceFormat&)that;
1.142 + return msgPattern == thatAlias.msgPattern;
1.143 +}
1.144 +
1.145 +// -------------------------------------
1.146 +// copy constructor
1.147 +
1.148 +const ChoiceFormat&
1.149 +ChoiceFormat::operator=(const ChoiceFormat& that)
1.150 +{
1.151 + if (this != &that) {
1.152 + NumberFormat::operator=(that);
1.153 + constructorErrorCode = that.constructorErrorCode;
1.154 + msgPattern = that.msgPattern;
1.155 + }
1.156 + return *this;
1.157 +}
1.158 +
1.159 +// -------------------------------------
1.160 +
1.161 +ChoiceFormat::~ChoiceFormat()
1.162 +{
1.163 +}
1.164 +
1.165 +// -------------------------------------
1.166 +
1.167 +/**
1.168 + * Convert a double value to a string without the overhead of NumberFormat.
1.169 + */
1.170 +UnicodeString&
1.171 +ChoiceFormat::dtos(double value,
1.172 + UnicodeString& string)
1.173 +{
1.174 + /* Buffer to contain the digits and any extra formatting stuff. */
1.175 + char temp[DBL_DIG + 16];
1.176 + char *itrPtr = temp;
1.177 + char *expPtr;
1.178 +
1.179 + sprintf(temp, "%.*g", DBL_DIG, value);
1.180 +
1.181 + /* Find and convert the decimal point.
1.182 + Using setlocale on some machines will cause sprintf to use a comma for certain locales.
1.183 + */
1.184 + while (*itrPtr && (*itrPtr == '-' || isdigit(*itrPtr))) {
1.185 + itrPtr++;
1.186 + }
1.187 + if (*itrPtr != 0 && *itrPtr != 'e') {
1.188 + /* We reached something that looks like a decimal point.
1.189 + In case someone used setlocale(), which changes the decimal point. */
1.190 + *itrPtr = '.';
1.191 + itrPtr++;
1.192 + }
1.193 + /* Search for the exponent */
1.194 + while (*itrPtr && *itrPtr != 'e') {
1.195 + itrPtr++;
1.196 + }
1.197 + if (*itrPtr == 'e') {
1.198 + itrPtr++;
1.199 + /* Verify the exponent sign */
1.200 + if (*itrPtr == '+' || *itrPtr == '-') {
1.201 + itrPtr++;
1.202 + }
1.203 + /* Remove leading zeros. You will see this on Windows machines. */
1.204 + expPtr = itrPtr;
1.205 + while (*itrPtr == '0') {
1.206 + itrPtr++;
1.207 + }
1.208 + if (*itrPtr && expPtr != itrPtr) {
1.209 + /* Shift the exponent without zeros. */
1.210 + while (*itrPtr) {
1.211 + *(expPtr++) = *(itrPtr++);
1.212 + }
1.213 + // NULL terminate
1.214 + *expPtr = 0;
1.215 + }
1.216 + }
1.217 +
1.218 + string = UnicodeString(temp, -1, US_INV); /* invariant codepage */
1.219 + return string;
1.220 +}
1.221 +
1.222 +// -------------------------------------
1.223 +// calls the overloaded applyPattern method.
1.224 +
1.225 +void
1.226 +ChoiceFormat::applyPattern(const UnicodeString& pattern,
1.227 + UErrorCode& status)
1.228 +{
1.229 + msgPattern.parseChoiceStyle(pattern, NULL, status);
1.230 + constructorErrorCode = status;
1.231 +}
1.232 +
1.233 +// -------------------------------------
1.234 +// Applies the pattern to this ChoiceFormat instance.
1.235 +
1.236 +void
1.237 +ChoiceFormat::applyPattern(const UnicodeString& pattern,
1.238 + UParseError& parseError,
1.239 + UErrorCode& status)
1.240 +{
1.241 + msgPattern.parseChoiceStyle(pattern, &parseError, status);
1.242 + constructorErrorCode = status;
1.243 +}
1.244 +// -------------------------------------
1.245 +// Returns the input pattern string.
1.246 +
1.247 +UnicodeString&
1.248 +ChoiceFormat::toPattern(UnicodeString& result) const
1.249 +{
1.250 + return result = msgPattern.getPatternString();
1.251 +}
1.252 +
1.253 +// -------------------------------------
1.254 +// Sets the limit and format arrays.
1.255 +void
1.256 +ChoiceFormat::setChoices( const double* limits,
1.257 + const UnicodeString* formats,
1.258 + int32_t cnt )
1.259 +{
1.260 + UErrorCode errorCode = U_ZERO_ERROR;
1.261 + setChoices(limits, NULL, formats, cnt, errorCode);
1.262 +}
1.263 +
1.264 +// -------------------------------------
1.265 +// Sets the limit and format arrays.
1.266 +void
1.267 +ChoiceFormat::setChoices( const double* limits,
1.268 + const UBool* closures,
1.269 + const UnicodeString* formats,
1.270 + int32_t cnt )
1.271 +{
1.272 + UErrorCode errorCode = U_ZERO_ERROR;
1.273 + setChoices(limits, closures, formats, cnt, errorCode);
1.274 +}
1.275 +
1.276 +void
1.277 +ChoiceFormat::setChoices(const double* limits,
1.278 + const UBool* closures,
1.279 + const UnicodeString* formats,
1.280 + int32_t count,
1.281 + UErrorCode &errorCode) {
1.282 + if (U_FAILURE(errorCode)) {
1.283 + return;
1.284 + }
1.285 + if (limits == NULL || formats == NULL) {
1.286 + errorCode = U_ILLEGAL_ARGUMENT_ERROR;
1.287 + return;
1.288 + }
1.289 + // Reconstruct the original input pattern.
1.290 + // Modified version of the pre-ICU 4.8 toPattern() implementation.
1.291 + UnicodeString result;
1.292 + for (int32_t i = 0; i < count; ++i) {
1.293 + if (i != 0) {
1.294 + result += VERTICAL_BAR;
1.295 + }
1.296 + UnicodeString buf;
1.297 + if (uprv_isPositiveInfinity(limits[i])) {
1.298 + result += INFINITY;
1.299 + } else if (uprv_isNegativeInfinity(limits[i])) {
1.300 + result += MINUS;
1.301 + result += INFINITY;
1.302 + } else {
1.303 + result += dtos(limits[i], buf);
1.304 + }
1.305 + if (closures != NULL && closures[i]) {
1.306 + result += LESS_THAN;
1.307 + } else {
1.308 + result += LESS_EQUAL;
1.309 + }
1.310 + // Append formats[i], using quotes if there are special
1.311 + // characters. Single quotes themselves must be escaped in
1.312 + // either case.
1.313 + const UnicodeString& text = formats[i];
1.314 + int32_t textLength = text.length();
1.315 + int32_t nestingLevel = 0;
1.316 + for (int32_t j = 0; j < textLength; ++j) {
1.317 + UChar c = text[j];
1.318 + if (c == SINGLE_QUOTE && nestingLevel == 0) {
1.319 + // Double each top-level apostrophe.
1.320 + result.append(c);
1.321 + } else if (c == VERTICAL_BAR && nestingLevel == 0) {
1.322 + // Surround each pipe symbol with apostrophes for quoting.
1.323 + // If the next character is an apostrophe, then that will be doubled,
1.324 + // and although the parser will see the apostrophe pairs beginning
1.325 + // and ending one character earlier than our doubling, the result
1.326 + // is as desired.
1.327 + // | -> '|'
1.328 + // |' -> '|'''
1.329 + // |'' -> '|''''' etc.
1.330 + result.append(SINGLE_QUOTE).append(c).append(SINGLE_QUOTE);
1.331 + continue; // Skip the append(c) at the end of the loop body.
1.332 + } else if (c == LEFT_CURLY_BRACE) {
1.333 + ++nestingLevel;
1.334 + } else if (c == RIGHT_CURLY_BRACE && nestingLevel > 0) {
1.335 + --nestingLevel;
1.336 + }
1.337 + result.append(c);
1.338 + }
1.339 + }
1.340 + // Apply the reconstructed pattern.
1.341 + applyPattern(result, errorCode);
1.342 +}
1.343 +
1.344 +// -------------------------------------
1.345 +// Gets the limit array.
1.346 +
1.347 +const double*
1.348 +ChoiceFormat::getLimits(int32_t& cnt) const
1.349 +{
1.350 + cnt = 0;
1.351 + return NULL;
1.352 +}
1.353 +
1.354 +// -------------------------------------
1.355 +// Gets the closures array.
1.356 +
1.357 +const UBool*
1.358 +ChoiceFormat::getClosures(int32_t& cnt) const
1.359 +{
1.360 + cnt = 0;
1.361 + return NULL;
1.362 +}
1.363 +
1.364 +// -------------------------------------
1.365 +// Gets the format array.
1.366 +
1.367 +const UnicodeString*
1.368 +ChoiceFormat::getFormats(int32_t& cnt) const
1.369 +{
1.370 + cnt = 0;
1.371 + return NULL;
1.372 +}
1.373 +
1.374 +// -------------------------------------
1.375 +// Formats an int64 number, it's actually formatted as
1.376 +// a double. The returned format string may differ
1.377 +// from the input number because of this.
1.378 +
1.379 +UnicodeString&
1.380 +ChoiceFormat::format(int64_t number,
1.381 + UnicodeString& appendTo,
1.382 + FieldPosition& status) const
1.383 +{
1.384 + return format((double) number, appendTo, status);
1.385 +}
1.386 +
1.387 +// -------------------------------------
1.388 +// Formats an int32_t number, it's actually formatted as
1.389 +// a double.
1.390 +
1.391 +UnicodeString&
1.392 +ChoiceFormat::format(int32_t number,
1.393 + UnicodeString& appendTo,
1.394 + FieldPosition& status) const
1.395 +{
1.396 + return format((double) number, appendTo, status);
1.397 +}
1.398 +
1.399 +// -------------------------------------
1.400 +// Formats a double number.
1.401 +
1.402 +UnicodeString&
1.403 +ChoiceFormat::format(double number,
1.404 + UnicodeString& appendTo,
1.405 + FieldPosition& /*pos*/) const
1.406 +{
1.407 + if (msgPattern.countParts() == 0) {
1.408 + // No pattern was applied, or it failed.
1.409 + return appendTo;
1.410 + }
1.411 + // Get the appropriate sub-message.
1.412 + int32_t msgStart = findSubMessage(msgPattern, 0, number);
1.413 + if (!MessageImpl::jdkAposMode(msgPattern)) {
1.414 + int32_t patternStart = msgPattern.getPart(msgStart).getLimit();
1.415 + int32_t msgLimit = msgPattern.getLimitPartIndex(msgStart);
1.416 + appendTo.append(msgPattern.getPatternString(),
1.417 + patternStart,
1.418 + msgPattern.getPatternIndex(msgLimit) - patternStart);
1.419 + return appendTo;
1.420 + }
1.421 + // JDK compatibility mode: Remove SKIP_SYNTAX.
1.422 + return MessageImpl::appendSubMessageWithoutSkipSyntax(msgPattern, msgStart, appendTo);
1.423 +}
1.424 +
1.425 +int32_t
1.426 +ChoiceFormat::findSubMessage(const MessagePattern &pattern, int32_t partIndex, double number) {
1.427 + int32_t count = pattern.countParts();
1.428 + int32_t msgStart;
1.429 + // Iterate over (ARG_INT|DOUBLE, ARG_SELECTOR, message) tuples
1.430 + // until ARG_LIMIT or end of choice-only pattern.
1.431 + // Ignore the first number and selector and start the loop on the first message.
1.432 + partIndex += 2;
1.433 + for (;;) {
1.434 + // Skip but remember the current sub-message.
1.435 + msgStart = partIndex;
1.436 + partIndex = pattern.getLimitPartIndex(partIndex);
1.437 + if (++partIndex >= count) {
1.438 + // Reached the end of the choice-only pattern.
1.439 + // Return with the last sub-message.
1.440 + break;
1.441 + }
1.442 + const MessagePattern::Part &part = pattern.getPart(partIndex++);
1.443 + UMessagePatternPartType type = part.getType();
1.444 + if (type == UMSGPAT_PART_TYPE_ARG_LIMIT) {
1.445 + // Reached the end of the ChoiceFormat style.
1.446 + // Return with the last sub-message.
1.447 + break;
1.448 + }
1.449 + // part is an ARG_INT or ARG_DOUBLE
1.450 + U_ASSERT(MessagePattern::Part::hasNumericValue(type));
1.451 + double boundary = pattern.getNumericValue(part);
1.452 + // Fetch the ARG_SELECTOR character.
1.453 + int32_t selectorIndex = pattern.getPatternIndex(partIndex++);
1.454 + UChar boundaryChar = pattern.getPatternString().charAt(selectorIndex);
1.455 + if (boundaryChar == LESS_THAN ? !(number > boundary) : !(number >= boundary)) {
1.456 + // The number is in the interval between the previous boundary and the current one.
1.457 + // Return with the sub-message between them.
1.458 + // The !(a>b) and !(a>=b) comparisons are equivalent to
1.459 + // (a<=b) and (a<b) except they "catch" NaN.
1.460 + break;
1.461 + }
1.462 + }
1.463 + return msgStart;
1.464 +}
1.465 +
1.466 +// -------------------------------------
1.467 +// Formats an array of objects. Checks if the data type of the objects
1.468 +// to get the right value for formatting.
1.469 +
1.470 +UnicodeString&
1.471 +ChoiceFormat::format(const Formattable* objs,
1.472 + int32_t cnt,
1.473 + UnicodeString& appendTo,
1.474 + FieldPosition& pos,
1.475 + UErrorCode& status) const
1.476 +{
1.477 + if(cnt < 0) {
1.478 + status = U_ILLEGAL_ARGUMENT_ERROR;
1.479 + return appendTo;
1.480 + }
1.481 + if (msgPattern.countParts() == 0) {
1.482 + status = U_INVALID_STATE_ERROR;
1.483 + return appendTo;
1.484 + }
1.485 +
1.486 + for (int32_t i = 0; i < cnt; i++) {
1.487 + double objDouble = objs[i].getDouble(status);
1.488 + if (U_SUCCESS(status)) {
1.489 + format(objDouble, appendTo, pos);
1.490 + }
1.491 + }
1.492 +
1.493 + return appendTo;
1.494 +}
1.495 +
1.496 +// -------------------------------------
1.497 +
1.498 +void
1.499 +ChoiceFormat::parse(const UnicodeString& text,
1.500 + Formattable& result,
1.501 + ParsePosition& pos) const
1.502 +{
1.503 + result.setDouble(parseArgument(msgPattern, 0, text, pos));
1.504 +}
1.505 +
1.506 +double
1.507 +ChoiceFormat::parseArgument(
1.508 + const MessagePattern &pattern, int32_t partIndex,
1.509 + const UnicodeString &source, ParsePosition &pos) {
1.510 + // find the best number (defined as the one with the longest parse)
1.511 + int32_t start = pos.getIndex();
1.512 + int32_t furthest = start;
1.513 + double bestNumber = uprv_getNaN();
1.514 + double tempNumber = 0.0;
1.515 + int32_t count = pattern.countParts();
1.516 + while (partIndex < count && pattern.getPartType(partIndex) != UMSGPAT_PART_TYPE_ARG_LIMIT) {
1.517 + tempNumber = pattern.getNumericValue(pattern.getPart(partIndex));
1.518 + partIndex += 2; // skip the numeric part and ignore the ARG_SELECTOR
1.519 + int32_t msgLimit = pattern.getLimitPartIndex(partIndex);
1.520 + int32_t len = matchStringUntilLimitPart(pattern, partIndex, msgLimit, source, start);
1.521 + if (len >= 0) {
1.522 + int32_t newIndex = start + len;
1.523 + if (newIndex > furthest) {
1.524 + furthest = newIndex;
1.525 + bestNumber = tempNumber;
1.526 + if (furthest == source.length()) {
1.527 + break;
1.528 + }
1.529 + }
1.530 + }
1.531 + partIndex = msgLimit + 1;
1.532 + }
1.533 + if (furthest == start) {
1.534 + pos.setErrorIndex(start);
1.535 + } else {
1.536 + pos.setIndex(furthest);
1.537 + }
1.538 + return bestNumber;
1.539 +}
1.540 +
1.541 +int32_t
1.542 +ChoiceFormat::matchStringUntilLimitPart(
1.543 + const MessagePattern &pattern, int32_t partIndex, int32_t limitPartIndex,
1.544 + const UnicodeString &source, int32_t sourceOffset) {
1.545 + int32_t matchingSourceLength = 0;
1.546 + const UnicodeString &msgString = pattern.getPatternString();
1.547 + int32_t prevIndex = pattern.getPart(partIndex).getLimit();
1.548 + for (;;) {
1.549 + const MessagePattern::Part &part = pattern.getPart(++partIndex);
1.550 + if (partIndex == limitPartIndex || part.getType() == UMSGPAT_PART_TYPE_SKIP_SYNTAX) {
1.551 + int32_t index = part.getIndex();
1.552 + int32_t length = index - prevIndex;
1.553 + if (length != 0 && 0 != source.compare(sourceOffset, length, msgString, prevIndex, length)) {
1.554 + return -1; // mismatch
1.555 + }
1.556 + matchingSourceLength += length;
1.557 + if (partIndex == limitPartIndex) {
1.558 + return matchingSourceLength;
1.559 + }
1.560 + prevIndex = part.getLimit(); // SKIP_SYNTAX
1.561 + }
1.562 + }
1.563 +}
1.564 +
1.565 +// -------------------------------------
1.566 +
1.567 +Format*
1.568 +ChoiceFormat::clone() const
1.569 +{
1.570 + ChoiceFormat *aCopy = new ChoiceFormat(*this);
1.571 + return aCopy;
1.572 +}
1.573 +
1.574 +U_NAMESPACE_END
1.575 +
1.576 +#endif /* #if !UCONFIG_NO_FORMATTING */
1.577 +
1.578 +//eof
