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);
}
}
}
UBool
NFRuleSet::parse(const UnicodeString& text, ParsePosition& pos, double upperBound, Formattable& result) const
{
// try matching each rule in the rule set against the text being
// parsed. Whichever one matches the most characters is the one
// that determines the value we return.
result.setLong(0);
// dump out if there's no text to parse
if (text.length() == 0) {
return 0;
}
ParsePosition highWaterMark;
ParsePosition workingPos = pos;
#ifdef RBNF_DEBUG
fprintf(stderr, "<nfrs> %x '", this);
dumpUS(stderr, name);
fprintf(stderr, "' text '");
dumpUS(stderr, text);
fprintf(stderr, "'\n");
fprintf(stderr, " parse negative: %d\n", this, negativeNumberRule != 0);
#endif
// start by trying the negative number rule (if there is one)
if (negativeNumberRule) {
Formattable tempResult;
#ifdef RBNF_DEBUG
fprintf(stderr, " <nfrs before negative> %x ub: %g\n", negativeNumberRule, upperBound);
#endif
UBool success = negativeNumberRule->doParse(text, workingPos, 0, upperBound, tempResult);
#ifdef RBNF_DEBUG
fprintf(stderr, " <nfrs after negative> success: %d wpi: %d\n", success, workingPos.getIndex());
#endif
if (success && workingPos.getIndex() > highWaterMark.getIndex()) {
result = tempResult;
highWaterMark = workingPos;
}
workingPos = pos;
}
#ifdef RBNF_DEBUG
fprintf(stderr, "<nfrs> continue fractional with text '");
dumpUS(stderr, text);
fprintf(stderr, "' hwm: %d\n", highWaterMark.getIndex());
#endif
// then try each of the fraction rules
{
for (int i = 0; i < 3; i++) {
if (fractionRules[i]) {
Formattable tempResult;
UBool success = fractionRules[i]->doParse(text, workingPos, 0, upperBound, tempResult);
if (success && (workingPos.getIndex() > highWaterMark.getIndex())) {
result = tempResult;
highWaterMark = workingPos;
}
workingPos = pos;
}
}
}
#ifdef RBNF_DEBUG
fprintf(stderr, "<nfrs> continue other with text '");
dumpUS(stderr, text);
fprintf(stderr, "' hwm: %d\n", highWaterMark.getIndex());
#endif
// finally, go through the regular rules one at a time. We start
// at the end of the list because we want to try matching the most
// sigificant rule first (this helps ensure that we parse
// "five thousand three hundred six" as
// "(five thousand) (three hundred) (six)" rather than
// "((five thousand three) hundred) (six)"). Skip rules whose
// base values are higher than the upper bound (again, this helps
// limit ambiguity by making sure the rules that match a rule's
// are less significant than the rule containing the substitutions)/
{
int64_t ub = util64_fromDouble(upperBound);
#ifdef RBNF_DEBUG
{
char ubstr[64];
util64_toa(ub, ubstr, 64);
char ubstrhex[64];
util64_toa(ub, ubstrhex, 64, 16);
fprintf(stderr, "ub: %g, i64: %s (%s)\n", upperBound, ubstr, ubstrhex);
}
#endif
for (int32_t i = rules.size(); --i >= 0 && highWaterMark.getIndex() < text.length();) {
if ((!fIsFractionRuleSet) && (rules[i]->getBaseValue() >= ub)) {
continue;
}
Formattable tempResult;
UBool success = rules[i]->doParse(text, workingPos, fIsFractionRuleSet, upperBound, tempResult);
if (success && workingPos.getIndex() > highWaterMark.getIndex()) {
result = tempResult;
highWaterMark = workingPos;
}
workingPos = pos;
}
}
#ifdef RBNF_DEBUG
fprintf(stderr, "<nfrs> exit\n");
#endif
// finally, update the parse postion we were passed to point to the
// first character we didn't use, and return the result that
// corresponds to that string of characters
pos = highWaterMark;
return 1;
}
void
PluralFormatTest::pluralFormatExtendedTest(void) {
const char *targets[] = {
"There are no widgets.",
"There is one widget.",
"There is a bling widget and one other widget.",
"There is a bling widget and 2 other widgets.",
"There is a bling widget and 3 other widgets.",
"Widgets, five (5-1=4) there be.",
"There is a bling widget and 5 other widgets.",
"There is a bling widget and 6 other widgets.",
};
const char* fmt =
"offset:1.0 "
"=0 {There are no widgets.} "
"=1.0 {There is one widget.} "
"=5 {Widgets, five (5-1=#) there be.} "
"one {There is a bling widget and one other widget.} "
"other {There is a bling widget and # other widgets.}";
UErrorCode status = U_ZERO_ERROR;
UnicodeString fmtString(fmt, -1, US_INV);
PluralFormat pf(Locale::getEnglish(), fmtString, status);
MessageFormat mf(UNICODE_STRING_SIMPLE("{0,plural,").append(fmtString).append((UChar)0x7d /* '}' */),
Locale::getEnglish(), status);
Formattable args;
FieldPosition ignore;
if (U_FAILURE(status)) {
dataerrln("Failed to apply pattern - %s", u_errorName(status));
return;
}
for (int32_t i = 0; i < 7; ++i) {
UnicodeString result = pf.format(i, status);
if (U_FAILURE(status)) {
errln("PluralFormat.format(value %d) failed - %s", i, u_errorName(status));
return;
}
UnicodeString expected(targets[i], -1, US_INV);
if (expected != result) {
UnicodeString message("PluralFormat.format(): Expected '", -1, US_INV);
message.append(expected);
message.append(UnicodeString("' but got '", -1, US_INV));
message.append(result);
message.append("'", -1, US_INV);
errln(message);
}
args.setLong(i);
mf.format(&args, 1, result.remove(), ignore, status);
if (U_FAILURE(status)) {
errln("MessageFormat.format(value %d) failed - %s", i, u_errorName(status));
return;
}
if (expected != result) {
UnicodeString message("MessageFormat.format(): Expected '", -1, US_INV);
message.append(expected);
message.append(UnicodeString("' but got '", -1, US_INV));
message.append(result);
message.append("'", -1, US_INV);
errln(message);
}
}
}
/**
* Parses a string using the rule set or DecimalFormat belonging
* to this substitution. If there's a match, a mathematical
* operation (the inverse of the one used in formatting) is
* performed on the result of the parse and the value passed in
* and returned as the result. The parse position is updated to
* point to the first unmatched character in the string.
* @param text The string to parse
* @param parsePosition On entry, ignored, but assumed to be 0.
* On exit, this is updated to point to the first unmatched
* character (or 0 if the substitution didn't match)
* @param baseValue A partial parse result that should be
* combined with the result of this parse
* @param upperBound When searching the rule set for a rule
* matching the string passed in, only rules with base values
* lower than this are considered
* @param lenientParse If true and matching against rules fails,
* the substitution will also try matching the text against
* numerals using a default-costructed NumberFormat. If false,
* no extra work is done. (This value is false whenever the
* formatter isn't in lenient-parse mode, but is also false
* under some conditions even when the formatter _is_ in
* lenient-parse mode.)
* @return If there's a match, this is the result of composing
* baseValue with whatever was returned from matching the
* characters. This will be either a Long or a Double. If there's
* no match this is new Long(0) (not null), and parsePosition
* is left unchanged.
*/
UBool
NFSubstitution::doParse(const UnicodeString& text,
ParsePosition& parsePosition,
double baseValue,
double upperBound,
UBool lenientParse,
Formattable& result) const
{
#ifdef RBNF_DEBUG
fprintf(stderr, "<nfsubs> %x bv: %g ub: %g\n", this, baseValue, upperBound);
#endif
// figure out the highest base value a rule can have and match
// the text being parsed (this varies according to the type of
// substitutions: multiplier, modulus, and numerator substitutions
// restrict the search to rules with base values lower than their
// own; same-value substitutions leave the upper bound wherever
// it was, and the others allow any rule to match
upperBound = calcUpperBound(upperBound);
// use our rule set to parse the text. If that fails and
// lenient parsing is enabled (this is always false if the
// formatter's lenient-parsing mode is off, but it may also
// be false even when the formatter's lenient-parse mode is
// on), then also try parsing the text using a default-
// constructed NumberFormat
if (ruleSet != NULL) {
ruleSet->parse(text, parsePosition, upperBound, result);
if (lenientParse && !ruleSet->isFractionRuleSet() && parsePosition.getIndex() == 0) {
UErrorCode status = U_ZERO_ERROR;
NumberFormat* fmt = NumberFormat::createInstance(status);
if (U_SUCCESS(status)) {
fmt->parse(text, result, parsePosition);
}
delete fmt;
}
// ...or use our DecimalFormat to parse the text
} else if (numberFormat != NULL) {
numberFormat->parse(text, result, parsePosition);
}
// if the parse was successful, we've already advanced the caller's
// parse position (this is the one function that doesn't have one
// of its own). Derive a parse result and return it as a Long,
// if possible, or a Double
if (parsePosition.getIndex() != 0) {
UErrorCode status = U_ZERO_ERROR;
double tempResult = result.getDouble(status);
// composeRuleValue() produces a full parse result from
// the partial parse result passed to this function from
// the caller (this is either the owning rule's base value
// or the partial result obtained from composing the
// owning rule's base value with its other substitution's
// parse result) and the partial parse result obtained by
// matching the substitution (which will be the same value
// the caller would get by parsing just this part of the
// text with RuleBasedNumberFormat.parse() ). How the two
// values are used to derive the full parse result depends
// on the types of substitutions: For a regular rule, the
// ultimate result is its multiplier substitution's result
// times the rule's divisor (or the rule's base value) plus
// the modulus substitution's result (which will actually
// supersede part of the rule's base value). For a negative-
// number rule, the result is the negative of its substitution's
// result. For a fraction rule, it's the sum of its two
// substitution results. For a rule in a fraction rule set,
// it's the numerator substitution's result divided by
// the rule's base value. Results from same-value substitutions
// propagate back upard, and null substitutions don't affect
// the result.
tempResult = composeRuleValue(tempResult, baseValue);
result.setDouble(tempResult);
return TRUE;
// if the parse was UNsuccessful, return 0
} else {
result.setLong(0);
return FALSE;
}
}
