int32_t SelectFormat::findSubMessage(const MessagePattern& pattern, int32_t partIndex,
const UnicodeString& keyword, UErrorCode& ec) {
if (U_FAILURE(ec)) {
return 0;
}
UnicodeString other(FALSE, SELECT_KEYWORD_OTHER, 5);
int32_t count = pattern.countParts();
int32_t msgStart=0;
// Iterate over (ARG_SELECTOR, message) pairs until ARG_LIMIT or end of select-only pattern.
do {
const MessagePattern::Part& part=pattern.getPart(partIndex++);
const UMessagePatternPartType type=part.getType();
if(type==UMSGPAT_PART_TYPE_ARG_LIMIT) {
break;
}
// part is an ARG_SELECTOR followed by a message
if(pattern.partSubstringMatches(part, keyword)) {
// keyword matches
return partIndex;
} else if(msgStart==0 && pattern.partSubstringMatches(part, other)) {
msgStart=partIndex;
}
partIndex=pattern.getLimitPartIndex(partIndex);
} while(++partIndex<count);
return msgStart;
}
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;
}
int32_t
ChoiceFormat::matchStringUntilLimitPart(
const MessagePattern &pattern, int32_t partIndex, int32_t limitPartIndex,
const UnicodeString &source, int32_t sourceOffset) {
int32_t matchingSourceLength = 0;
const UnicodeString &msgString = pattern.getPatternString();
int32_t prevIndex = pattern.getPart(partIndex).getLimit();
for (;;) {
const MessagePattern::Part &part = pattern.getPart(++partIndex);
if (partIndex == limitPartIndex || part.getType() == UMSGPAT_PART_TYPE_SKIP_SYNTAX) {
int32_t index = part.getIndex();
int32_t length = index - prevIndex;
if (length != 0 && 0 != source.compare(sourceOffset, length, msgString, prevIndex, length)) {
return -1; // mismatch
}
matchingSourceLength += length;
if (partIndex == limitPartIndex) {
return matchingSourceLength;
}
prevIndex = part.getLimit(); // SKIP_SYNTAX
}
}
}
int32_t
ChoiceFormat::findSubMessage(const MessagePattern &pattern, int32_t partIndex, double number) {
int32_t count = pattern.countParts();
int32_t msgStart;
// Iterate over (ARG_INT|DOUBLE, ARG_SELECTOR, message) tuples
// until ARG_LIMIT or end of choice-only pattern.
// Ignore the first number and selector and start the loop on the first message.
partIndex += 2;
for (;;) {
// Skip but remember the current sub-message.
msgStart = partIndex;
partIndex = pattern.getLimitPartIndex(partIndex);
if (++partIndex >= count) {
// Reached the end of the choice-only pattern.
// Return with the last sub-message.
break;
}
const MessagePattern::Part &part = pattern.getPart(partIndex++);
UMessagePatternPartType type = part.getType();
if (type == UMSGPAT_PART_TYPE_ARG_LIMIT) {
// Reached the end of the ChoiceFormat style.
// Return with the last sub-message.
break;
}
// part is an ARG_INT or ARG_DOUBLE
U_ASSERT(MessagePattern::Part::hasNumericValue(type));
double boundary = pattern.getNumericValue(part);
// Fetch the ARG_SELECTOR character.
int32_t selectorIndex = pattern.getPatternIndex(partIndex++);
UChar boundaryChar = pattern.getPatternString().charAt(selectorIndex);
if (boundaryChar == LESS_THAN ? !(number > boundary) : !(number >= boundary)) {
// The number is in the interval between the previous boundary and the current one.
// Return with the sub-message between them.
// The !(a>b) and !(a>=b) comparisons are equivalent to
// (a<=b) and (a<b) except they "catch" NaN.
break;
}
}
return msgStart;
}
int32_t PluralFormat::findSubMessage(const MessagePattern& pattern, int32_t partIndex,
const PluralSelector& selector, void *context,
double number, UErrorCode& ec) {
if (U_FAILURE(ec)) {
return 0;
}
int32_t count=pattern.countParts();
double offset;
const MessagePattern::Part* part=&pattern.getPart(partIndex);
if (MessagePattern::Part::hasNumericValue(part->getType())) {
offset=pattern.getNumericValue(*part);
++partIndex;
} else {
offset=0;
}
// The keyword is empty until we need to match against a non-explicit, not-"other" value.
// Then we get the keyword from the selector.
// (In other words, we never call the selector if we match against an explicit value,
// or if the only non-explicit keyword is "other".)
UnicodeString keyword;
UnicodeString other(FALSE, OTHER_STRING, 5);
// When we find a match, we set msgStart>0 and also set this boolean to true
// to avoid matching the keyword again (duplicates are allowed)
// while we continue to look for an explicit-value match.
UBool haveKeywordMatch=FALSE;
// msgStart is 0 until we find any appropriate sub-message.
// We remember the first "other" sub-message if we have not seen any
// appropriate sub-message before.
// We remember the first matching-keyword sub-message if we have not seen
// one of those before.
// (The parser allows [does not check for] duplicate keywords.
// We just have to make sure to take the first one.)
// We avoid matching the keyword twice by also setting haveKeywordMatch=true
// at the first keyword match.
// We keep going until we find an explicit-value match or reach the end of the plural style.
int32_t msgStart=0;
// Iterate over (ARG_SELECTOR [ARG_INT|ARG_DOUBLE] message) tuples
// until ARG_LIMIT or end of plural-only pattern.
do {
part=&pattern.getPart(partIndex++);
const UMessagePatternPartType type = part->getType();
if(type==UMSGPAT_PART_TYPE_ARG_LIMIT) {
break;
}
U_ASSERT (type==UMSGPAT_PART_TYPE_ARG_SELECTOR);
// part is an ARG_SELECTOR followed by an optional explicit value, and then a message
if(MessagePattern::Part::hasNumericValue(pattern.getPartType(partIndex))) {
// explicit value like "=2"
part=&pattern.getPart(partIndex++);
if(number==pattern.getNumericValue(*part)) {
// matches explicit value
return partIndex;
}
} else if(!haveKeywordMatch) {
// plural keyword like "few" or "other"
// Compare "other" first and call the selector if this is not "other".
if(pattern.partSubstringMatches(*part, other)) {
if(msgStart==0) {
msgStart=partIndex;
if(0 == keyword.compare(other)) {
// This is the first "other" sub-message,
// and the selected keyword is also "other".
// Do not match "other" again.
haveKeywordMatch=TRUE;
}
}
} else {
if(keyword.isEmpty()) {
keyword=selector.select(context, number-offset, ec);
if(msgStart!=0 && (0 == keyword.compare(other))) {
// We have already seen an "other" sub-message.
// Do not match "other" again.
haveKeywordMatch=TRUE;
// Skip keyword matching but do getLimitPartIndex().
}
}
if(!haveKeywordMatch && pattern.partSubstringMatches(*part, keyword)) {
// keyword matches
msgStart=partIndex;
// Do not match this keyword again.
haveKeywordMatch=TRUE;
}
}
}
partIndex=pattern.getLimitPartIndex(partIndex);
} while(++partIndex<count);
return msgStart;
}
