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);
}
}
}
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);
}
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;
}
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;
}
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;
}
void
PluralFormat::parseObject(const UnicodeString& /*source*/,
Formattable& /*result*/,
ParsePosition& pos) const
{
// Parsing not supported.
pos.setErrorIndex(pos.getIndex());
}
void
SelectFormat::parseObject(const UnicodeString& /*source*/,
Formattable& /*result*/,
ParsePosition& pos) const
{
// TODO: not yet supported in icu4j and icu4c
pos.setErrorIndex(pos.getIndex());
}
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)
}
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);
}
}
}
