void ParsedNumber::populateFormattable(Formattable& output, parse_flags_t parseFlags) const {
bool sawNaN = 0 != (flags & FLAG_NAN);
bool sawInfinity = 0 != (flags & FLAG_INFINITY);
bool integerOnly = 0 != (parseFlags & PARSE_FLAG_INTEGER_ONLY);
// Check for NaN, infinity, and -0.0
if (sawNaN) {
// Can't use NAN or std::nan because the byte pattern is platform-dependent;
// MSVC sets the sign bit, but Clang and GCC do not
output.setDouble(uprv_getNaN());
return;
}
if (sawInfinity) {
if (0 != (flags & FLAG_NEGATIVE)) {
output.setDouble(-INFINITY);
return;
} else {
output.setDouble(INFINITY);
return;
}
}
U_ASSERT(!quantity.bogus);
if (quantity.isZero() && quantity.isNegative() && !integerOnly) {
output.setDouble(-0.0);
return;
}
// All other numbers
output.adoptDecimalQuantity(new DecimalQuantity(quantity));
}
/**
* If this is a >>> substitution, match only against ruleToUse.
* Otherwise, use the superclass function.
* @param text The string to parse
* @param parsePosition Ignored on entry, updated on exit to point to
* the first unmatched character.
* @param baseValue The partial parse result prior to calling this
* routine.
*/
UBool
ModulusSubstitution::doParse(const UnicodeString& text,
ParsePosition& parsePosition,
double baseValue,
double upperBound,
UBool lenientParse,
Formattable& result) const
{
// if this isn't a >>> substitution, we can just use the
// inherited parse() routine to do the parsing
if (ruleToUse == NULL) {
return NFSubstitution::doParse(text, parsePosition, baseValue, upperBound, lenientParse, result);
// but if it IS a >>> substitution, we have to do it here: we
// use the specific rule's doParse() method, and then we have to
// do some of the other work of NFRuleSet.parse()
} else {
ruleToUse->doParse(text, parsePosition, FALSE, upperBound, result);
if (parsePosition.getIndex() != 0) {
UErrorCode status = U_ZERO_ERROR;
double tempResult = result.getDouble(status);
tempResult = composeRuleValue(tempResult, baseValue);
result.setDouble(tempResult);
}
return TRUE;
}
}
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);
}
virtual UBool doParse(const UnicodeString& /*text*/,
ParsePosition& /*parsePosition*/,
double baseValue,
double /*upperBound*/,
UBool /*lenientParse*/,
Formattable& result) const
{ result.setDouble(baseValue); return TRUE; }
void
ChoiceFormat::parse(const UnicodeString& text,
Formattable& result,
ParsePosition& pos) const
{
result.setDouble(parseArgument(msgPattern, 0, text, pos));
}
/**
* 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;
}
}
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;
}
}
