void
ChoiceFormat::parse(const UnicodeString& text,
Formattable& result,
ParsePosition& status) const
{
// find the best number (defined as the one with the longest parse)
int32_t start = status.getIndex();
int32_t furthest = start;
double bestNumber = uprv_getNaN();
double tempNumber = 0.0;
for (int i = 0; i < fCount; ++i) {
int32_t len = fChoiceFormats[i].length();
if (text.compare(start, len, fChoiceFormats[i]) == 0) {
status.setIndex(start + len);
tempNumber = fChoiceLimits[i];
if (status.getIndex() > furthest) {
furthest = status.getIndex();
bestNumber = tempNumber;
if (furthest == text.length())
break;
}
}
}
status.setIndex(furthest);
if (status.getIndex() == start) {
status.setErrorIndex(furthest);
}
result.setDouble(bestNumber);
}
double
ChoiceFormat::parseArgument(
const MessagePattern &pattern, int32_t partIndex,
const UnicodeString &source, ParsePosition &pos) {
// find the best number (defined as the one with the longest parse)
int32_t start = pos.getIndex();
int32_t furthest = start;
double bestNumber = uprv_getNaN();
double tempNumber = 0.0;
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 == source.length()) {
break;
}
}
}
partIndex = msgLimit + 1;
}
if (furthest == start) {
pos.setErrorIndex(start);
} else {
pos.setIndex(furthest);
}
return bestNumber;
}
U_CAPI double U_EXPORT2
unum_parseDoubleCurrency(const UNumberFormat* fmt,
const UChar* text,
int32_t textLength,
int32_t* parsePos, /* 0 = start */
UChar* currency,
UErrorCode* status) {
double doubleVal = 0.0;
currency[0] = 0;
if (U_FAILURE(*status)) {
return doubleVal;
}
const UnicodeString src((UBool)(textLength == -1), text, textLength);
ParsePosition pp;
if (parsePos != NULL) {
pp.setIndex(*parsePos);
}
*status = U_PARSE_ERROR; // assume failure, reset if succeed
LocalPointer<CurrencyAmount> currAmt(((const NumberFormat*)fmt)->parseCurrency(src, pp));
if (pp.getErrorIndex() != -1) {
if (parsePos != NULL) {
*parsePos = pp.getErrorIndex();
}
} else {
if (parsePos != NULL) {
*parsePos = pp.getIndex();
}
if (pp.getIndex() > 0) {
*status = U_ZERO_ERROR;
u_strcpy(currency, currAmt->getISOCurrency());
doubleVal = currAmt->getNumber().getDouble(*status);
}
}
return doubleVal;
}
U_CAPI void U_EXPORT2
udat_parseCalendar(const UDateFormat* format,
UCalendar* calendar,
const UChar* text,
int32_t textLength,
int32_t *parsePos,
UErrorCode *status)
{
if(U_FAILURE(*status)) return;
const UnicodeString src((UBool)(textLength == -1), text, textLength);
ParsePosition pp;
if(parsePos != 0)
pp.setIndex(*parsePos);
((DateFormat*)format)->parse(src, *(Calendar*)calendar, pp);
if(parsePos != 0) {
if(pp.getErrorIndex() == -1)
*parsePos = pp.getIndex();
else {
*parsePos = pp.getErrorIndex();
*status = U_PARSE_ERROR;
}
}
}
UDate
DateFormat::parse(const UnicodeString& text,
ParsePosition& pos) const
{
UDate d = 0; // Error return UDate is 0 (the epoch)
if (fCalendar != NULL) {
int32_t start = pos.getIndex();
// Parse may update TimeZone used by the calendar.
TimeZone *tzsav = (TimeZone*)fCalendar->getTimeZone().clone();
fCalendar->clear();
parse(text, *fCalendar, pos);
if (pos.getIndex() != start) {
UErrorCode ec = U_ZERO_ERROR;
d = fCalendar->getTime(ec);
if (U_FAILURE(ec)) {
// We arrive here if fCalendar is non-lenient and there
// is an out-of-range field. We don't know which field
// was illegal so we set the error index to the start.
pos.setIndex(start);
pos.setErrorIndex(start);
d = 0;
}
}
// Restore TimeZone
fCalendar->adoptTimeZone(tzsav);
}
return d;
}
static void
parseRes(Formattable& res,
const UNumberFormat* fmt,
const UChar* text,
int32_t textLength,
int32_t *parsePos /* 0 = start */,
UBool parseCurrency,
UErrorCode *status)
{
if(U_FAILURE(*status))
return;
int32_t len = (textLength == -1 ? u_strlen(text) : textLength);
const UnicodeString src((UChar*)text, len, len);
ParsePosition pp;
if(parsePos != 0)
pp.setIndex(*parsePos);
if (parseCurrency) {
((const NumberFormat*)fmt)->parseCurrency(src, res, pp);
} else {
((const NumberFormat*)fmt)->parse(src, res, pp);
}
if(pp.getErrorIndex() != -1) {
*status = U_PARSE_ERROR;
if(parsePos != 0) {
*parsePos = pp.getErrorIndex();
}
} else if(parsePos != 0) {
*parsePos = pp.getIndex();
}
}
U_CAPI UDate U_EXPORT2
udat_parse( const UDateFormat* format,
const UChar* text,
int32_t textLength,
int32_t *parsePos,
UErrorCode *status)
{
if(U_FAILURE(*status)) return (UDate)0;
const UnicodeString src((UBool)(textLength == -1), text, textLength);
ParsePosition pp;
int32_t stackParsePos = 0;
UDate res;
if(parsePos == NULL) {
parsePos = &stackParsePos;
}
pp.setIndex(*parsePos);
res = ((DateFormat*)format)->parse(src, pp);
if(pp.getErrorIndex() == -1)
*parsePos = pp.getIndex();
else {
*parsePos = pp.getErrorIndex();
*status = U_PARSE_ERROR;
}
return res;
}
static void
parseRes(Formattable& res,
const UNumberFormat* fmt,
const UChar* text,
int32_t textLength,
int32_t *parsePos /* 0 = start */,
UErrorCode *status)
{
if(U_FAILURE(*status))
return;
const UnicodeString src((UBool)(textLength == -1), text, textLength);
ParsePosition pp;
if(parsePos != 0)
pp.setIndex(*parsePos);
((const NumberFormat*)fmt)->parse(src, res, pp);
if(pp.getErrorIndex() != -1) {
*status = U_PARSE_ERROR;
if(parsePos != 0) {
*parsePos = pp.getErrorIndex();
}
} else if(parsePos != 0) {
*parsePos = pp.getIndex();
}
}
UDate
DateFormat::parse(const UnicodeString& text,
ParsePosition& pos) const
{
UDate d = 0; // Error return UDate is 0 (the epoch)
if (fCalendar != NULL) {
Calendar* calClone = fCalendar->clone();
if (calClone != NULL) {
int32_t start = pos.getIndex();
calClone->clear();
parse(text, *calClone, pos);
if (pos.getIndex() != start) {
UErrorCode ec = U_ZERO_ERROR;
d = calClone->getTime(ec);
if (U_FAILURE(ec)) {
// We arrive here if fCalendar => calClone is non-lenient and
// there is an out-of-range field. We don't know which field
// was illegal so we set the error index to the start.
pos.setIndex(start);
pos.setErrorIndex(start);
d = 0;
}
}
delete calClone;
}
}
return d;
}
void
RuleBasedNumberFormat::parse(const UnicodeString& text,
Formattable& result,
ParsePosition& parsePosition) const
{
//TODO: We need a real fix. See #6895 / #6896
if (noParse) {
// skip parsing
parsePosition.setErrorIndex(0);
return;
}
if (!ruleSets) {
parsePosition.setErrorIndex(0);
return;
}
UnicodeString workingText(text, parsePosition.getIndex());
ParsePosition workingPos(0);
ParsePosition high_pp(0);
Formattable high_result;
for (NFRuleSet** p = ruleSets; *p; ++p) {
NFRuleSet *rp = *p;
if (rp->isPublic() && rp->isParseable()) {
ParsePosition working_pp(0);
Formattable working_result;
rp->parse(workingText, working_pp, kMaxDouble, working_result);
if (working_pp.getIndex() > high_pp.getIndex()) {
high_pp = working_pp;
high_result = working_result;
if (high_pp.getIndex() == workingText.length()) {
break;
}
}
}
}
int32_t startIndex = parsePosition.getIndex();
parsePosition.setIndex(startIndex + high_pp.getIndex());
if (high_pp.getIndex() > 0) {
parsePosition.setErrorIndex(-1);
} else {
int32_t errorIndex = (high_pp.getErrorIndex()>0)? high_pp.getErrorIndex(): 0;
parsePosition.setErrorIndex(startIndex + errorIndex);
}
result = high_result;
if (result.getType() == Formattable::kDouble) {
int32_t r = (int32_t)result.getDouble();
if ((double)r == result.getDouble()) {
result.setLong(r);
}
}
}
//
// RBBISymbolTable::parseReference This function from the abstract symbol table interface
// looks for a $variable name in the source text.
// It does not look it up, only scans for it.
// It is used by the UnicodeSet parser.
//
// This implementation is lifted pretty much verbatim
// from the rules based transliterator implementation.
// I didn't see an obvious way of sharing it.
//
UnicodeString RBBISymbolTable::parseReference(const UnicodeString& text,
ParsePosition& pos, int32_t limit) const
{
int32_t start = pos.getIndex();
int32_t i = start;
UnicodeString result;
while (i < limit) {
UChar c = text.charAt(i);
if ((i==start && !u_isIDStart(c)) || !u_isIDPart(c)) {
break;
}
++i;
}
if (i == start) { // No valid name chars
return result; // Indicate failure with empty string
}
pos.setIndex(i);
text.extractBetween(start, i, result);
return result;
}
void RelativeDateFormat::parse( const UnicodeString& text,
Calendar& cal,
ParsePosition& pos) const {
// Can the fDateFormat parse it?
if(fDateFormat != NULL) {
ParsePosition aPos(pos);
fDateFormat->parse(text,cal,aPos);
if((aPos.getIndex() != pos.getIndex()) &&
(aPos.getErrorIndex()==-1)) {
pos=aPos; // copy the sub parse
return; // parsed subfmt OK
}
}
// Linear search the relative strings
for(int n=0; n<fDatesLen; n++) {
if(fDates[n].string != NULL &&
(0==text.compare(pos.getIndex(),
fDates[n].len,
fDates[n].string))) {
UErrorCode status = U_ZERO_ERROR;
// Set the calendar to now+offset
cal.setTime(Calendar::getNow(),status);
cal.add(UCAL_DATE,fDates[n].offset, status);
if(U_FAILURE(status)) {
// failure in setting calendar fields
pos.setErrorIndex(pos.getIndex()+fDates[n].len);
} else {
pos.setIndex(pos.getIndex()+fDates[n].len);
}
return;
}
}
// parse failed
}
UDate
DateFormat::parse(const UnicodeString& text,
ParsePosition& pos) const
{
if (fCalendar != NULL) {
int32_t start = pos.getIndex();
fCalendar->clear();
parse(text, *fCalendar, pos);
if (pos.getIndex() != start) {
UErrorCode ec = U_ZERO_ERROR;
UDate d = fCalendar->getTime(ec);
if (U_SUCCESS(ec)) {
return d; // Successful function exit
}
// We arrive here if fCalendar is non-lenient and there
// is an out-of-range field. We don't know which field
// was illegal so we set the error index to the start.
pos.setIndex(start);
pos.setErrorIndex(start);
}
}
return 0; // Error return UDate is 0 (the epoch)
}
//---------------------------------------------------------------------------------
//
// scanSet Construct a UnicodeSet from the text at the current scan
// position. Advance the scan position to the first character
// after the set.
//
// A new RBBI setref node referring to the set is pushed onto the node
// stack.
//
// The scan position is normally under the control of the state machine
// that controls rule parsing. UnicodeSets, however, are parsed by
// the UnicodeSet constructor, not by the RBBI rule parser.
//
//---------------------------------------------------------------------------------
void RBBIRuleScanner::scanSet() {
UnicodeSet *uset;
ParsePosition pos;
int startPos;
int i;
if (U_FAILURE(*fRB->fStatus)) {
return;
}
pos.setIndex(fScanIndex);
startPos = fScanIndex;
UErrorCode localStatus = U_ZERO_ERROR;
uset = new UnicodeSet(fRB->fRules, pos, USET_IGNORE_SPACE,
fSymbolTable,
localStatus);
if (U_FAILURE(localStatus)) {
// TODO: Get more accurate position of the error from UnicodeSet's return info.
// UnicodeSet appears to not be reporting correctly at this time.
#ifdef RBBI_DEBUG
RBBIDebugPrintf("UnicodeSet parse postion.ErrorIndex = %d\n", pos.getIndex());
#endif
error(localStatus);
delete uset;
return;
}
// Verify that the set contains at least one code point.
//
if (uset->isEmpty()) {
// This set is empty.
// Make it an error, because it almost certainly is not what the user wanted.
// Also, avoids having to think about corner cases in the tree manipulation code
// that occurs later on.
error(U_BRK_RULE_EMPTY_SET);
delete uset;
return;
}
// Advance the RBBI parse postion over the UnicodeSet pattern.
// Don't just set fScanIndex because the line/char positions maintained
// for error reporting would be thrown off.
i = pos.getIndex();
for (;;) {
if (fNextIndex >= i) {
break;
}
nextCharLL();
}
if (U_SUCCESS(*fRB->fStatus)) {
RBBINode *n;
n = pushNewNode(RBBINode::setRef);
n->fFirstPos = startPos;
n->fLastPos = fNextIndex;
fRB->fRules.extractBetween(n->fFirstPos, n->fLastPos, n->fText);
// findSetFor() serves several purposes here:
// - Adopts storage for the UnicodeSet, will be responsible for deleting.
// - Mantains collection of all sets in use, needed later for establishing
// character categories for run time engine.
// - Eliminates mulitiple instances of the same set.
// - Creates a new uset node if necessary (if this isn't a duplicate.)
findSetFor(n->fText, n, uset);
}
}
void RelativeDateFormat::parse( const UnicodeString& text,
Calendar& cal,
ParsePosition& pos) const {
int32_t startIndex = pos.getIndex();
if (fDatePattern.isEmpty()) {
// no date pattern, try parsing as time
fDateTimeFormatter->applyPattern(fTimePattern);
fDateTimeFormatter->parse(text,cal,pos);
} else if (fTimePattern.isEmpty() || fCombinedFormat == NULL) {
// no time pattern or way to combine, try parsing as date
// first check whether text matches a relativeDayString
UBool matchedRelative = FALSE;
for (int n=0; n < fDatesLen && !matchedRelative; n++) {
if (fDates[n].string != NULL &&
text.compare(startIndex, fDates[n].len, fDates[n].string) == 0) {
// it matched, handle the relative day string
UErrorCode status = U_ZERO_ERROR;
matchedRelative = TRUE;
// Set the calendar to now+offset
cal.setTime(Calendar::getNow(),status);
cal.add(UCAL_DATE,fDates[n].offset, status);
if(U_FAILURE(status)) {
// failure in setting calendar field, set offset to beginning of rel day string
pos.setErrorIndex(startIndex);
} else {
pos.setIndex(startIndex + fDates[n].len);
}
}
}
if (!matchedRelative) {
// just parse as normal date
fDateTimeFormatter->applyPattern(fDatePattern);
fDateTimeFormatter->parse(text,cal,pos);
}
} else {
// Here we replace any relativeDayString in text with the equivalent date
// formatted per fDatePattern, then parse text normally using the combined pattern.
UnicodeString modifiedText(text);
FieldPosition fPos;
int32_t dateStart = 0, origDateLen = 0, modDateLen = 0;
UErrorCode status = U_ZERO_ERROR;
for (int n=0; n < fDatesLen; n++) {
int32_t relativeStringOffset;
if (fDates[n].string != NULL &&
(relativeStringOffset = modifiedText.indexOf(fDates[n].string, fDates[n].len, startIndex)) >= startIndex) {
// it matched, replace the relative date with a real one for parsing
UnicodeString dateString;
Calendar * tempCal = cal.clone();
// Set the calendar to now+offset
tempCal->setTime(Calendar::getNow(),status);
tempCal->add(UCAL_DATE,fDates[n].offset, status);
if(U_FAILURE(status)) {
pos.setErrorIndex(startIndex);
delete tempCal;
return;
}
fDateTimeFormatter->applyPattern(fDatePattern);
fDateTimeFormatter->format(*tempCal, dateString, fPos);
dateStart = relativeStringOffset;
origDateLen = fDates[n].len;
modDateLen = dateString.length();
modifiedText.replace(dateStart, origDateLen, dateString);
delete tempCal;
break;
}
}
UnicodeString combinedPattern;
fCombinedFormat->format(fTimePattern, fDatePattern, combinedPattern, status);
fDateTimeFormatter->applyPattern(combinedPattern);
fDateTimeFormatter->parse(modifiedText,cal,pos);
// Adjust offsets
UBool noError = (pos.getErrorIndex() < 0);
int32_t offset = (noError)? pos.getIndex(): pos.getErrorIndex();
if (offset >= dateStart + modDateLen) {
// offset at or after the end of the replaced text,
// correct by the difference between original and replacement
offset -= (modDateLen - origDateLen);
} else if (offset >= dateStart) {
// offset in the replaced text, set it to the beginning of that text
// (i.e. the beginning of the relative day string)
offset = dateStart;
}
if (noError) {
pos.setIndex(offset);
} else {
pos.setErrorIndex(offset);
}
}
}
void
TimeUnitFormat::parseObject(const UnicodeString& source,
Formattable& result,
ParsePosition& pos) const {
Formattable resultNumber(0.0);
UBool withNumberFormat = false;
TimeUnit::UTimeUnitFields resultTimeUnit = TimeUnit::UTIMEUNIT_FIELD_COUNT;
int32_t oldPos = pos.getIndex();
int32_t newPos = -1;
int32_t longestParseDistance = 0;
UnicodeString* countOfLongestMatch = NULL;
#ifdef TMUTFMT_DEBUG
char res[1000];
source.extract(0, source.length(), res, "UTF-8");
std::cout << "parse source: " << res << "\n";
#endif
// parse by iterating through all available patterns
// and looking for the longest match.
for (TimeUnit::UTimeUnitFields i = TimeUnit::UTIMEUNIT_YEAR;
i < TimeUnit::UTIMEUNIT_FIELD_COUNT;
i = (TimeUnit::UTimeUnitFields)(i+1)) {
Hashtable* countToPatterns = fTimeUnitToCountToPatterns[i];
int32_t elemPos = UHASH_FIRST;
const UHashElement* elem = NULL;
while ((elem = countToPatterns->nextElement(elemPos)) != NULL){
const UHashTok keyTok = elem->key;
UnicodeString* count = (UnicodeString*)keyTok.pointer;
#ifdef TMUTFMT_DEBUG
count->extract(0, count->length(), res, "UTF-8");
std::cout << "parse plural count: " << res << "\n";
#endif
const UHashTok valueTok = elem->value;
// the value is a pair of MessageFormat*
MessageFormat** patterns = (MessageFormat**)valueTok.pointer;
for (UTimeUnitFormatStyle style = UTMUTFMT_FULL_STYLE; style < UTMUTFMT_FORMAT_STYLE_COUNT;
style = (UTimeUnitFormatStyle)(style + 1)) {
MessageFormat* pattern = patterns[style];
pos.setErrorIndex(-1);
pos.setIndex(oldPos);
// see if we can parse
Formattable parsed;
pattern->parseObject(source, parsed, pos);
if (pos.getErrorIndex() != -1 || pos.getIndex() == oldPos) {
continue;
}
#ifdef TMUTFMT_DEBUG
std::cout << "parsed.getType: " << parsed.getType() << "\n";
#endif
Formattable tmpNumber(0.0);
if (pattern->getArgTypeCount() != 0) {
Formattable& temp = parsed[0];
if (temp.getType() == Formattable::kString) {
UnicodeString tmpString;
UErrorCode pStatus = U_ZERO_ERROR;
getNumberFormat().parse(temp.getString(tmpString), tmpNumber, pStatus);
if (U_FAILURE(pStatus)) {
continue;
}
} else if (temp.isNumeric()) {
tmpNumber = temp;
} else {
continue;
}
}
int32_t parseDistance = pos.getIndex() - oldPos;
if (parseDistance > longestParseDistance) {
if (pattern->getArgTypeCount() != 0) {
resultNumber = tmpNumber;
withNumberFormat = true;
} else {
withNumberFormat = false;
}
resultTimeUnit = i;
newPos = pos.getIndex();
longestParseDistance = parseDistance;
countOfLongestMatch = count;
}
}
}
}
/* After find the longest match, parse the number.
* Result number could be null for the pattern without number pattern.
* such as unit pattern in Arabic.
* When result number is null, use plural rule to set the number.
*/
if (withNumberFormat == false && longestParseDistance != 0) {
// set the number using plurrual count
if (0 == countOfLongestMatch->compare(PLURAL_COUNT_ZERO, 4)) {
resultNumber = Formattable(0.0);
} else if (0 == countOfLongestMatch->compare(PLURAL_COUNT_ONE, 3)) {
resultNumber = Formattable(1.0);
} else if (0 == countOfLongestMatch->compare(PLURAL_COUNT_TWO, 3)) {
resultNumber = Formattable(2.0);
} else {
// should not happen.
// TODO: how to handle?
resultNumber = Formattable(3.0);
}
}
if (longestParseDistance == 0) {
pos.setIndex(oldPos);
pos.setErrorIndex(0);
} else {
UErrorCode status = U_ZERO_ERROR;
TimeUnitAmount* tmutamt = new TimeUnitAmount(resultNumber, resultTimeUnit, status);
if (U_SUCCESS(status)) {
result.adoptObject(tmutamt);
pos.setIndex(newPos);
pos.setErrorIndex(-1);
} else {
pos.setIndex(oldPos);
pos.setErrorIndex(0);
}
}
}
void
TimeUnitFormat::parseObject(const UnicodeString& source,
Formattable& result,
ParsePosition& pos) const {
double resultNumber = -1;
UBool withNumberFormat = false;
TimeUnit::UTimeUnitFields resultTimeUnit = TimeUnit::UTIMEUNIT_FIELD_COUNT;
int32_t oldPos = pos.getIndex();
int32_t newPos = -1;
int32_t longestParseDistance = 0;
UnicodeString* countOfLongestMatch = NULL;
#ifdef TMUTFMT_DEBUG
char res[1000];
source.extract(0, source.length(), res, "UTF-8");
std::cout << "parse source: " << res << "\n";
#endif
// parse by iterating through all available patterns
// and looking for the longest match.
for (TimeUnit::UTimeUnitFields i = TimeUnit::UTIMEUNIT_YEAR;
i < TimeUnit::UTIMEUNIT_FIELD_COUNT;
i = (TimeUnit::UTimeUnitFields)(i+1)) {
Hashtable* countToPatterns = fTimeUnitToCountToPatterns[i];
int32_t elemPos = -1;
const UHashElement* elem = NULL;
while ((elem = countToPatterns->nextElement(elemPos)) != NULL){
const UHashTok keyTok = elem->key;
UnicodeString* count = (UnicodeString*)keyTok.pointer;
#ifdef TMUTFMT_DEBUG
count->extract(0, count->length(), res, "UTF-8");
std::cout << "parse plural count: " << res << "\n";
#endif
const UHashTok valueTok = elem->value;
// the value is a pair of MessageFormat*
MessageFormat** patterns = (MessageFormat**)valueTok.pointer;
for (EStyle style = kFull; style < kTotal; style = (EStyle)(style + 1)) {
MessageFormat* pattern = patterns[style];
pos.setErrorIndex(-1);
pos.setIndex(oldPos);
// see if we can parse
Formattable parsed;
pattern->parseObject(source, parsed, pos);
if (pos.getErrorIndex() != -1 || pos.getIndex() == oldPos) {
continue;
}
#ifdef TMUTFMT_DEBUG
std::cout << "parsed.getType: " << parsed.getType() << "\n";
#endif
double tmpNumber = 0;
if (pattern->getArgTypeCount() != 0) {
// pattern with Number as beginning, such as "{0} d".
// check to make sure that the timeUnit is consistent
Formattable& temp = parsed[0];
if (temp.getType() == Formattable::kDouble) {
tmpNumber = temp.getDouble();
} else if (temp.getType() == Formattable::kLong) {
tmpNumber = temp.getLong();
} else {
continue;
}
UnicodeString select = fPluralRules->select(tmpNumber);
#ifdef TMUTFMT_DEBUG
select.extract(0, select.length(), res, "UTF-8");
std::cout << "parse plural select count: " << res << "\n";
#endif
if (*count != select) {
continue;
}
}
int32_t parseDistance = pos.getIndex() - oldPos;
if (parseDistance > longestParseDistance) {
if (pattern->getArgTypeCount() != 0) {
resultNumber = tmpNumber;
withNumberFormat = true;
} else {
withNumberFormat = false;
}
resultTimeUnit = i;
newPos = pos.getIndex();
longestParseDistance = parseDistance;
countOfLongestMatch = count;
}
}
}
}
/* After find the longest match, parse the number.
* Result number could be null for the pattern without number pattern.
* such as unit pattern in Arabic.
* When result number is null, use plural rule to set the number.
*/
if (withNumberFormat == false && longestParseDistance != 0) {
// set the number using plurrual count
if ( *countOfLongestMatch == PLURAL_COUNT_ZERO ) {
resultNumber = 0;
} else if ( *countOfLongestMatch == PLURAL_COUNT_ONE ) {
resultNumber = 1;
} else if ( *countOfLongestMatch == PLURAL_COUNT_TWO ) {
resultNumber = 2;
} else {
// should not happen.
// TODO: how to handle?
resultNumber = 3;
}
}
if (longestParseDistance == 0) {
pos.setIndex(oldPos);
pos.setErrorIndex(0);
} else {
UErrorCode status = U_ZERO_ERROR;
TimeUnitAmount* tmutamt = new TimeUnitAmount(resultNumber, resultTimeUnit, status);
if (U_SUCCESS(status)) {
result.adoptObject(tmutamt);
pos.setIndex(newPos);
pos.setErrorIndex(-1);
} else {
pos.setIndex(oldPos);
pos.setErrorIndex(0);
}
}
}
static jobject parseRBNFImpl(JNIEnv *env, jclass clazz, jint addr, jstring text,
jobject position, jboolean lenient) {
// LOGI("ENTER parseRBNFImpl");
const char * parsePositionClassName = "java/text/ParsePosition";
const char * longClassName = "java/lang/Long";
const char * doubleClassName = "java/lang/Double";
UErrorCode status = U_ZERO_ERROR;
UNumberFormat *fmt = (UNumberFormat *)(int)addr;
jchar *str = (UChar *)env->GetStringChars(text, NULL);
int strlength = env->GetStringLength(text);
jclass parsePositionClass = env->FindClass(parsePositionClassName);
jclass longClass = env->FindClass(longClassName);
jclass doubleClass = env->FindClass(doubleClassName);
jmethodID getIndexMethodID = env->GetMethodID(parsePositionClass,
"getIndex", "()I");
jmethodID setIndexMethodID = env->GetMethodID(parsePositionClass,
"setIndex", "(I)V");
jmethodID setErrorIndexMethodID = env->GetMethodID(parsePositionClass,
"setErrorIndex", "(I)V");
jmethodID longInitMethodID = env->GetMethodID(longClass, "<init>", "(J)V");
jmethodID dblInitMethodID = env->GetMethodID(doubleClass, "<init>", "(D)V");
int parsePos = env->CallIntMethod(position, getIndexMethodID, NULL);
// make sure the ParsePosition is valid. Actually icu4c would parse a number
// correctly even if the parsePosition is set to -1, but since the RI fails
// for that case we have to fail too
if(parsePos < 0 || parsePos > strlength) {
return NULL;
}
Formattable res;
const UnicodeString src((UChar*)str, strlength, strlength);
ParsePosition pp;
pp.setIndex(parsePos);
if(lenient) {
unum_setAttribute(fmt, UNUM_LENIENT_PARSE, JNI_TRUE);
}
((const NumberFormat*)fmt)->parse(src, res, pp);
if(lenient) {
unum_setAttribute(fmt, UNUM_LENIENT_PARSE, JNI_FALSE);
}
env->ReleaseStringChars(text, str);
if(pp.getErrorIndex() == -1) {
parsePos = pp.getIndex();
} else {
env->CallVoidMethod(position, setErrorIndexMethodID,
(jint) pp.getErrorIndex());
return NULL;
}
Formattable::Type numType;
numType = res.getType();
UErrorCode fmtStatus;
double resultDouble;
long resultLong;
int64_t resultInt64;
switch(numType) {
case Formattable::kDouble:
resultDouble = res.getDouble();
env->CallVoidMethod(position, setIndexMethodID, (jint) parsePos);
return env->NewObject(doubleClass, dblInitMethodID,
(jdouble) resultDouble);
case Formattable::kLong:
resultLong = res.getLong();
env->CallVoidMethod(position, setIndexMethodID, (jint) parsePos);
return env->NewObject(longClass, longInitMethodID,
(jlong) resultLong);
case Formattable::kInt64:
resultInt64 = res.getInt64();
env->CallVoidMethod(position, setIndexMethodID, (jint) parsePos);
return env->NewObject(longClass, longInitMethodID,
(jlong) resultInt64);
default:
break;
}
return NULL;
}
UBool
NumeratorSubstitution::doParse(const UnicodeString& text,
ParsePosition& parsePosition,
double baseValue,
double upperBound,
UBool /*lenientParse*/,
Formattable& result) const
{
// we don't have to do anything special to do the parsing here,
// but we have to turn lenient parsing off-- if we leave it on,
// it SERIOUSLY messes up the algorithm
// if withZeros is true, we need to count the zeros
// and use that to adjust the parse result
UErrorCode status = U_ZERO_ERROR;
int32_t zeroCount = 0;
UnicodeString workText(text);
if (withZeros) {
ParsePosition workPos(1);
Formattable temp;
while (workText.length() > 0 && workPos.getIndex() != 0) {
workPos.setIndex(0);
getRuleSet()->parse(workText, workPos, 1, temp); // parse zero or nothing at all
if (workPos.getIndex() == 0) {
// we failed, either there were no more zeros, or the number was formatted with digits
// either way, we're done
break;
}
++zeroCount;
parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex());
workText.remove(0, workPos.getIndex());
while (workText.length() > 0 && workText.charAt(0) == gSpace) {
workText.remove(0, 1);
parsePosition.setIndex(parsePosition.getIndex() + 1);
}
}
workText = text;
workText.remove(0, (int32_t)parsePosition.getIndex());
parsePosition.setIndex(0);
}
// we've parsed off the zeros, now let's parse the rest from our current position
NFSubstitution::doParse(workText, parsePosition, withZeros ? 1 : baseValue, upperBound, FALSE, result);
if (withZeros) {
// any base value will do in this case. is there a way to
// force this to not bother trying all the base values?
// compute the 'effective' base and prescale the value down
int64_t n = result.getLong(status); // force conversion!
int64_t d = 1;
int32_t pow = 0;
while (d <= n) {
d *= 10;
++pow;
}
// now add the zeros
while (zeroCount > 0) {
d *= 10;
--zeroCount;
}
// d is now our true denominator
result.setDouble((double)n/(double)d);
}
return TRUE;
}
UBool
FractionalPartSubstitution::doParse(const UnicodeString& text,
ParsePosition& parsePosition,
double baseValue,
double /*upperBound*/,
UBool lenientParse,
Formattable& resVal) const
{
// if we're not in byDigits mode, we can just use the inherited
// doParse()
if (!byDigits) {
return NFSubstitution::doParse(text, parsePosition, baseValue, 0, lenientParse, resVal);
// if we ARE in byDigits mode, parse the text one digit at a time
// using this substitution's owning rule set (we do this by setting
// upperBound to 10 when calling doParse() ) until we reach
// nonmatching text
} else {
UnicodeString workText(text);
ParsePosition workPos(1);
double result = 0;
int32_t digit;
// double p10 = 0.1;
DigitList dl;
NumberFormat* fmt = NULL;
while (workText.length() > 0 && workPos.getIndex() != 0) {
workPos.setIndex(0);
Formattable temp;
getRuleSet()->parse(workText, workPos, 10, temp);
UErrorCode status = U_ZERO_ERROR;
digit = temp.getLong(status);
// digit = temp.getType() == Formattable::kLong ?
// temp.getLong() :
// (int32_t)temp.getDouble();
if (lenientParse && workPos.getIndex() == 0) {
if (!fmt) {
status = U_ZERO_ERROR;
fmt = NumberFormat::createInstance(status);
if (U_FAILURE(status)) {
delete fmt;
fmt = NULL;
}
}
if (fmt) {
fmt->parse(workText, temp, workPos);
digit = temp.getLong(status);
}
}
if (workPos.getIndex() != 0) {
dl.append((char)('0' + digit));
// result += digit * p10;
// p10 /= 10;
parsePosition.setIndex(parsePosition.getIndex() + workPos.getIndex());
workText.removeBetween(0, workPos.getIndex());
while (workText.length() > 0 && workText.charAt(0) == gSpace) {
workText.removeBetween(0, 1);
parsePosition.setIndex(parsePosition.getIndex() + 1);
}
}
}
delete fmt;
result = dl.fCount == 0 ? 0 : dl.getDouble();
result = composeRuleValue(result, baseValue);
resVal.setDouble(result);
return TRUE;
}
}
