/*
* Initializes the region data from the ICU resource bundles. The region data
* contains the basic relationships such as which regions are known, what the numeric
* codes are, any known aliases, and the territory containment data.
*
* If the region data has already loaded, then this method simply returns without doing
* anything meaningful.
*/
void Region::loadRegionData(UErrorCode &status) {
// Construct service objs first
LocalUHashtablePointer newRegionIDMap(uhash_open(uhash_hashUnicodeString, uhash_compareUnicodeString, NULL, &status));
LocalUHashtablePointer newNumericCodeMap(uhash_open(uhash_hashLong,uhash_compareLong,NULL,&status));
LocalUHashtablePointer newRegionAliases(uhash_open(uhash_hashUnicodeString,uhash_compareUnicodeString,NULL,&status));
LocalPointer<DecimalFormat> df(new DecimalFormat(status), status);
LocalPointer<UVector> continents(new UVector(uprv_deleteUObject, uhash_compareUnicodeString, status), status);
LocalPointer<UVector> groupings(new UVector(uprv_deleteUObject, uhash_compareUnicodeString, status), status);
allRegions = new UVector(uprv_deleteUObject, uhash_compareUnicodeString, status);
LocalUResourceBundlePointer metadata(ures_openDirect(NULL,"metadata",&status));
LocalUResourceBundlePointer metadataAlias(ures_getByKey(metadata.getAlias(),"alias",NULL,&status));
LocalUResourceBundlePointer territoryAlias(ures_getByKey(metadataAlias.getAlias(),"territory",NULL,&status));
LocalUResourceBundlePointer supplementalData(ures_openDirect(NULL,"supplementalData",&status));
LocalUResourceBundlePointer codeMappings(ures_getByKey(supplementalData.getAlias(),"codeMappings",NULL,&status));
LocalUResourceBundlePointer idValidity(ures_getByKey(supplementalData.getAlias(),"idValidity",NULL,&status));
LocalUResourceBundlePointer regionList(ures_getByKey(idValidity.getAlias(),"region",NULL,&status));
LocalUResourceBundlePointer regionRegular(ures_getByKey(regionList.getAlias(),"regular",NULL,&status));
LocalUResourceBundlePointer regionMacro(ures_getByKey(regionList.getAlias(),"macroregion",NULL,&status));
LocalUResourceBundlePointer regionUnknown(ures_getByKey(regionList.getAlias(),"unknown",NULL,&status));
LocalUResourceBundlePointer territoryContainment(ures_getByKey(supplementalData.getAlias(),"territoryContainment",NULL,&status));
LocalUResourceBundlePointer worldContainment(ures_getByKey(territoryContainment.getAlias(),"001",NULL,&status));
LocalUResourceBundlePointer groupingContainment(ures_getByKey(territoryContainment.getAlias(),"grouping",NULL,&status));
if (U_FAILURE(status)) {
return;
}
// now, initialize
df->setParseIntegerOnly(TRUE);
uhash_setValueDeleter(newRegionIDMap.getAlias(), deleteRegion); // regionIDMap owns objs
uhash_setKeyDeleter(newRegionAliases.getAlias(), uprv_deleteUObject); // regionAliases owns the string keys
while ( ures_hasNext(regionRegular.getAlias()) ) {
UnicodeString regionName = ures_getNextUnicodeString(regionRegular.getAlias(),NULL,&status);
int32_t rangeMarkerLocation = regionName.indexOf(RANGE_MARKER);
UChar buf[6];
regionName.extract(buf,6,status);
if ( rangeMarkerLocation > 0 ) {
UChar endRange = regionName.charAt(rangeMarkerLocation+1);
buf[rangeMarkerLocation] = 0;
while ( buf[rangeMarkerLocation-1] <= endRange ) {
LocalPointer<UnicodeString> newRegion(new UnicodeString(buf), status);
allRegions->addElement(newRegion.orphan(),status);
buf[rangeMarkerLocation-1]++;
}
} else {
LocalPointer<UnicodeString> newRegion(new UnicodeString(regionName), status);
allRegions->addElement(newRegion.orphan(),status);
}
}
while ( ures_hasNext(regionMacro.getAlias()) ) {
UnicodeString regionName = ures_getNextUnicodeString(regionMacro.getAlias(),NULL,&status);
int32_t rangeMarkerLocation = regionName.indexOf(RANGE_MARKER);
UChar buf[6];
regionName.extract(buf,6,status);
if ( rangeMarkerLocation > 0 ) {
UChar endRange = regionName.charAt(rangeMarkerLocation+1);
buf[rangeMarkerLocation] = 0;
while ( buf[rangeMarkerLocation-1] <= endRange ) {
LocalPointer<UnicodeString> newRegion(new UnicodeString(buf), status);
allRegions->addElement(newRegion.orphan(),status);
buf[rangeMarkerLocation-1]++;
}
} else {
LocalPointer<UnicodeString> newRegion(new UnicodeString(regionName), status);
allRegions->addElement(newRegion.orphan(),status);
}
}
while ( ures_hasNext(regionUnknown.getAlias()) ) {
LocalPointer<UnicodeString> regionName (new UnicodeString(ures_getNextUnicodeString(regionUnknown.getAlias(),NULL,&status),status));
allRegions->addElement(regionName.orphan(),status);
}
while ( ures_hasNext(worldContainment.getAlias()) ) {
UnicodeString *continentName = new UnicodeString(ures_getNextUnicodeString(worldContainment.getAlias(),NULL,&status));
continents->addElement(continentName,status);
}
while ( ures_hasNext(groupingContainment.getAlias()) ) {
UnicodeString *groupingName = new UnicodeString(ures_getNextUnicodeString(groupingContainment.getAlias(),NULL,&status));
groupings->addElement(groupingName,status);
}
for ( int32_t i = 0 ; i < allRegions->size() ; i++ ) {
LocalPointer<Region> r(new Region(), status);
if ( U_FAILURE(status) ) {
return;
}
UnicodeString *regionName = (UnicodeString *)allRegions->elementAt(i);
r->idStr = *regionName;
r->idStr.extract(0,r->idStr.length(),r->id,sizeof(r->id),US_INV);
r->type = URGN_TERRITORY; // Only temporary - figure out the real type later once the aliases are known.
Formattable result;
UErrorCode ps = U_ZERO_ERROR;
df->parse(r->idStr,result,ps);
if ( U_SUCCESS(ps) ) {
r->code = result.getLong(); // Convert string to number
uhash_iput(newNumericCodeMap.getAlias(),r->code,(void *)(r.getAlias()),&status);
r->type = URGN_SUBCONTINENT;
} else {
r->code = -1;
}
void* idStrAlias = (void*)&(r->idStr); // about to orphan 'r'. Save this off.
uhash_put(newRegionIDMap.getAlias(),idStrAlias,(void *)(r.orphan()),&status); // regionIDMap takes ownership
}
// Process the territory aliases
while ( ures_hasNext(territoryAlias.getAlias()) ) {
LocalUResourceBundlePointer res(ures_getNextResource(territoryAlias.getAlias(),NULL,&status));
const char *aliasFrom = ures_getKey(res.getAlias());
LocalPointer<UnicodeString> aliasFromStr(new UnicodeString(aliasFrom, -1, US_INV), status);
UnicodeString aliasTo = ures_getUnicodeStringByKey(res.getAlias(),"replacement",&status);
res.adoptInstead(NULL);
const Region *aliasToRegion = (Region *) uhash_get(newRegionIDMap.getAlias(),&aliasTo);
Region *aliasFromRegion = (Region *)uhash_get(newRegionIDMap.getAlias(),aliasFromStr.getAlias());
if ( aliasToRegion != NULL && aliasFromRegion == NULL ) { // This is just an alias from some string to a region
uhash_put(newRegionAliases.getAlias(),(void *)aliasFromStr.orphan(), (void *)aliasToRegion,&status);
} else {
if ( aliasFromRegion == NULL ) { // Deprecated region code not in the master codes list - so need to create a deprecated region for it.
LocalPointer<Region> newRgn(new Region, status);
if ( U_SUCCESS(status) ) {
aliasFromRegion = newRgn.orphan();
} else {
return; // error out
}
aliasFromRegion->idStr.setTo(*aliasFromStr);
aliasFromRegion->idStr.extract(0,aliasFromRegion->idStr.length(),aliasFromRegion->id,sizeof(aliasFromRegion->id),US_INV);
uhash_put(newRegionIDMap.getAlias(),(void *)&(aliasFromRegion->idStr),(void *)aliasFromRegion,&status);
Formattable result;
UErrorCode ps = U_ZERO_ERROR;
df->parse(aliasFromRegion->idStr,result,ps);
if ( U_SUCCESS(ps) ) {
aliasFromRegion->code = result.getLong(); // Convert string to number
uhash_iput(newNumericCodeMap.getAlias(),aliasFromRegion->code,(void *)aliasFromRegion,&status);
} else {
aliasFromRegion->code = -1;
}
aliasFromRegion->type = URGN_DEPRECATED;
} else {
aliasFromRegion->type = URGN_DEPRECATED;
}
{
LocalPointer<UVector> newPreferredValues(new UVector(uprv_deleteUObject, uhash_compareUnicodeString, status), status);
aliasFromRegion->preferredValues = newPreferredValues.orphan();
}
if( U_FAILURE(status)) {
return;
}
UnicodeString currentRegion;
//currentRegion.remove(); TODO: was already 0 length?
for (int32_t i = 0 ; i < aliasTo.length() ; i++ ) {
if ( aliasTo.charAt(i) != 0x0020 ) {
currentRegion.append(aliasTo.charAt(i));
}
if ( aliasTo.charAt(i) == 0x0020 || i+1 == aliasTo.length() ) {
Region *target = (Region *)uhash_get(newRegionIDMap.getAlias(),(void *)¤tRegion);
if (target) {
LocalPointer<UnicodeString> preferredValue(new UnicodeString(target->idStr), status);
aliasFromRegion->preferredValues->addElement((void *)preferredValue.orphan(),status); // may add null if err
}
currentRegion.remove();
}
}
}
}
// Process the code mappings - This will allow us to assign numeric codes to most of the territories.
while ( ures_hasNext(codeMappings.getAlias()) ) {
UResourceBundle *mapping = ures_getNextResource(codeMappings.getAlias(),NULL,&status);
if ( ures_getType(mapping) == URES_ARRAY && ures_getSize(mapping) == 3) {
UnicodeString codeMappingID = ures_getUnicodeStringByIndex(mapping,0,&status);
UnicodeString codeMappingNumber = ures_getUnicodeStringByIndex(mapping,1,&status);
UnicodeString codeMapping3Letter = ures_getUnicodeStringByIndex(mapping,2,&status);
Region *r = (Region *)uhash_get(newRegionIDMap.getAlias(),(void *)&codeMappingID);
if ( r ) {
Formattable result;
UErrorCode ps = U_ZERO_ERROR;
df->parse(codeMappingNumber,result,ps);
if ( U_SUCCESS(ps) ) {
r->code = result.getLong(); // Convert string to number
uhash_iput(newNumericCodeMap.getAlias(),r->code,(void *)r,&status);
}
LocalPointer<UnicodeString> code3(new UnicodeString(codeMapping3Letter), status);
uhash_put(newRegionAliases.getAlias(),(void *)code3.orphan(), (void *)r,&status);
}
}
ures_close(mapping);
}
// Now fill in the special cases for WORLD, UNKNOWN, CONTINENTS, and GROUPINGS
Region *r;
UnicodeString WORLD_ID_STRING(WORLD_ID);
r = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)&WORLD_ID_STRING);
if ( r ) {
r->type = URGN_WORLD;
}
UnicodeString UNKNOWN_REGION_ID_STRING(UNKNOWN_REGION_ID);
r = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)&UNKNOWN_REGION_ID_STRING);
if ( r ) {
r->type = URGN_UNKNOWN;
}
for ( int32_t i = 0 ; i < continents->size() ; i++ ) {
r = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)continents->elementAt(i));
if ( r ) {
r->type = URGN_CONTINENT;
}
}
for ( int32_t i = 0 ; i < groupings->size() ; i++ ) {
r = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)groupings->elementAt(i));
if ( r ) {
r->type = URGN_GROUPING;
}
}
// Special case: The region code "QO" (Outlying Oceania) is a subcontinent code added by CLDR
// even though it looks like a territory code. Need to handle it here.
UnicodeString OUTLYING_OCEANIA_REGION_ID_STRING(OUTLYING_OCEANIA_REGION_ID);
r = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)&OUTLYING_OCEANIA_REGION_ID_STRING);
if ( r ) {
r->type = URGN_SUBCONTINENT;
}
// Load territory containment info from the supplemental data.
while ( ures_hasNext(territoryContainment.getAlias()) ) {
LocalUResourceBundlePointer mapping(ures_getNextResource(territoryContainment.getAlias(),NULL,&status));
if( U_FAILURE(status) ) {
return; // error out
}
const char *parent = ures_getKey(mapping.getAlias());
if (uprv_strcmp(parent, "containedGroupings") == 0 || uprv_strcmp(parent, "deprecated") == 0) {
continue; // handle new pseudo-parent types added in ICU data per cldrbug 7808; for now just skip.
// #11232 is to do something useful with these.
}
UnicodeString parentStr = UnicodeString(parent, -1 , US_INV);
Region *parentRegion = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)&parentStr);
for ( int j = 0 ; j < ures_getSize(mapping.getAlias()); j++ ) {
UnicodeString child = ures_getUnicodeStringByIndex(mapping.getAlias(),j,&status);
Region *childRegion = (Region *) uhash_get(newRegionIDMap.getAlias(),(void *)&child);
if ( parentRegion != NULL && childRegion != NULL ) {
// Add the child region to the set of regions contained by the parent
if (parentRegion->containedRegions == NULL) {
parentRegion->containedRegions = new UVector(uprv_deleteUObject, uhash_compareUnicodeString, status);
}
LocalPointer<UnicodeString> childStr(new UnicodeString(), status);
if( U_FAILURE(status) ) {
return; // error out
}
childStr->fastCopyFrom(childRegion->idStr);
parentRegion->containedRegions->addElement((void *)childStr.orphan(),status);
// Set the parent region to be the containing region of the child.
// Regions of type GROUPING can't be set as the parent, since another region
// such as a SUBCONTINENT, CONTINENT, or WORLD must always be the parent.
if ( parentRegion->type != URGN_GROUPING) {
childRegion->containingRegion = parentRegion;
}
}
}
}
// Create the availableRegions lists
int32_t pos = UHASH_FIRST;
while ( const UHashElement* element = uhash_nextElement(newRegionIDMap.getAlias(),&pos)) {
Region *ar = (Region *)element->value.pointer;
if ( availableRegions[ar->type] == NULL ) {
LocalPointer<UVector> newAr(new UVector(uprv_deleteUObject, uhash_compareUnicodeString, status), status);
availableRegions[ar->type] = newAr.orphan();
}
LocalPointer<UnicodeString> arString(new UnicodeString(ar->idStr), status);
if( U_FAILURE(status) ) {
return; // error out
}
availableRegions[ar->type]->addElement((void *)arString.orphan(),status);
}
ucln_i18n_registerCleanup(UCLN_I18N_REGION, region_cleanup);
// copy hashtables
numericCodeMap = newNumericCodeMap.orphan();
regionIDMap = newRegionIDMap.orphan();
regionAliases = newRegionAliases.orphan();
}
NumberFormat*
NumberFormat::makeInstance(const Locale& desiredLocale,
UNumberFormatStyle style,
UBool mustBeDecimalFormat,
UErrorCode& status) {
if (U_FAILURE(status)) return NULL;
if (style < 0 || style >= UNUM_FORMAT_STYLE_COUNT) {
status = U_ILLEGAL_ARGUMENT_ERROR;
return NULL;
}
// Some styles are not supported. This is a result of merging
// the @draft ICU 4.2 NumberFormat::EStyles into the long-existing UNumberFormatStyle.
// Ticket #8503 is for reviewing/fixing/merging the two relevant implementations:
// this one and unum_open().
// The UNUM_PATTERN_ styles are not supported here
// because this method does not take a pattern string.
if (!isStyleSupported(style)) {
status = U_UNSUPPORTED_ERROR;
return NULL;
}
#if U_PLATFORM_USES_ONLY_WIN32_API
if (!mustBeDecimalFormat) {
char buffer[8];
int32_t count = desiredLocale.getKeywordValue("compat", buffer, sizeof(buffer), status);
// if the locale has "@compat=host", create a host-specific NumberFormat
if (U_SUCCESS(status) && count > 0 && uprv_strcmp(buffer, "host") == 0) {
Win32NumberFormat *f = NULL;
UBool curr = TRUE;
switch (style) {
case UNUM_DECIMAL:
curr = FALSE;
// fall-through
case UNUM_CURRENCY:
case UNUM_CURRENCY_ISO: // do not support plural formatting here
case UNUM_CURRENCY_PLURAL:
f = new Win32NumberFormat(desiredLocale, curr, status);
if (U_SUCCESS(status)) {
return f;
}
delete f;
break;
default:
break;
}
}
}
#endif
// Use numbering system cache hashtable
umtx_initOnce(gNSCacheInitOnce, &nscacheInit);
// Get cached numbering system
LocalPointer<NumberingSystem> ownedNs;
NumberingSystem *ns = NULL;
if (NumberingSystem_cache != NULL) {
// TODO: Bad hash key usage, see ticket #8504.
int32_t hashKey = desiredLocale.hashCode();
Mutex lock(&nscacheMutex);
ns = (NumberingSystem *)uhash_iget(NumberingSystem_cache, hashKey);
if (ns == NULL) {
ns = NumberingSystem::createInstance(desiredLocale,status);
uhash_iput(NumberingSystem_cache, hashKey, (void*)ns, &status);
}
} else {
ownedNs.adoptInstead(NumberingSystem::createInstance(desiredLocale,status));
ns = ownedNs.getAlias();
}
// check results of getting a numbering system
if (U_FAILURE(status)) {
return NULL;
}
if (mustBeDecimalFormat && ns->isAlgorithmic()) {
status = U_UNSUPPORTED_ERROR;
return NULL;
}
LocalPointer<DecimalFormatSymbols> symbolsToAdopt;
UnicodeString pattern;
LocalUResourceBundlePointer ownedResource(ures_open(NULL, desiredLocale.getName(), &status));
if (U_FAILURE(status)) {
// We don't appear to have resource data available -- use the last-resort data
status = U_USING_FALLBACK_WARNING;
// When the data is unavailable, and locale isn't passed in, last resort data is used.
symbolsToAdopt.adoptInstead(new DecimalFormatSymbols(status));
if (symbolsToAdopt.isNull()) {
status = U_MEMORY_ALLOCATION_ERROR;
return NULL;
}
// Creates a DecimalFormat instance with the last resort number patterns.
pattern.setTo(TRUE, gLastResortNumberPatterns[style], -1);
}
else {
// Loads the decimal symbols of the desired locale.
symbolsToAdopt.adoptInstead(new DecimalFormatSymbols(desiredLocale, status));
if (symbolsToAdopt.isNull()) {
status = U_MEMORY_ALLOCATION_ERROR;
return NULL;
}
UResourceBundle *resource = ownedResource.orphan();
UResourceBundle *numElements = ures_getByKeyWithFallback(resource, gNumberElements, NULL, &status);
resource = ures_getByKeyWithFallback(numElements, ns->getName(), resource, &status);
resource = ures_getByKeyWithFallback(resource, gPatterns, resource, &status);
ownedResource.adoptInstead(resource);
int32_t patLen = 0;
const UChar *patResStr = ures_getStringByKeyWithFallback(resource, gFormatKeys[style], &patLen, &status);
// Didn't find a pattern specific to the numbering system, so fall back to "latn"
if ( status == U_MISSING_RESOURCE_ERROR && uprv_strcmp(gLatn,ns->getName())) {
status = U_ZERO_ERROR;
resource = ures_getByKeyWithFallback(numElements, gLatn, resource, &status);
resource = ures_getByKeyWithFallback(resource, gPatterns, resource, &status);
patResStr = ures_getStringByKeyWithFallback(resource, gFormatKeys[style], &patLen, &status);
}
ures_close(numElements);
// Creates the specified decimal format style of the desired locale.
pattern.setTo(TRUE, patResStr, patLen);
}
if (U_FAILURE(status)) {
return NULL;
}
if(style==UNUM_CURRENCY || style == UNUM_CURRENCY_ISO){
const UChar* currPattern = symbolsToAdopt->getCurrencyPattern();
if(currPattern!=NULL){
pattern.setTo(currPattern, u_strlen(currPattern));
}
}
NumberFormat *f;
if (ns->isAlgorithmic()) {
UnicodeString nsDesc;
UnicodeString nsRuleSetGroup;
UnicodeString nsRuleSetName;
Locale nsLoc;
URBNFRuleSetTag desiredRulesType = URBNF_NUMBERING_SYSTEM;
nsDesc.setTo(ns->getDescription());
int32_t firstSlash = nsDesc.indexOf(gSlash);
int32_t lastSlash = nsDesc.lastIndexOf(gSlash);
if ( lastSlash > firstSlash ) {
CharString nsLocID;
nsLocID.appendInvariantChars(nsDesc.tempSubString(0, firstSlash), status);
nsRuleSetGroup.setTo(nsDesc,firstSlash+1,lastSlash-firstSlash-1);
nsRuleSetName.setTo(nsDesc,lastSlash+1);
nsLoc = Locale::createFromName(nsLocID.data());
UnicodeString SpelloutRules = UNICODE_STRING_SIMPLE("SpelloutRules");
if ( nsRuleSetGroup.compare(SpelloutRules) == 0 ) {
desiredRulesType = URBNF_SPELLOUT;
}
} else {
nsLoc = desiredLocale;
nsRuleSetName.setTo(nsDesc);
}
RuleBasedNumberFormat *r = new RuleBasedNumberFormat(desiredRulesType,nsLoc,status);
if (r == NULL) {
status = U_MEMORY_ALLOCATION_ERROR;
return NULL;
}
r->setDefaultRuleSet(nsRuleSetName,status);
f = r;
} else {
// replace single currency sign in the pattern with double currency sign
// if the style is UNUM_CURRENCY_ISO
if (style == UNUM_CURRENCY_ISO) {
pattern.findAndReplace(UnicodeString(TRUE, gSingleCurrencySign, 1),
UnicodeString(TRUE, gDoubleCurrencySign, 2));
}
// "new DecimalFormat()" does not adopt the symbols if its memory allocation fails.
DecimalFormatSymbols *syms = symbolsToAdopt.orphan();
f = new DecimalFormat(pattern, syms, style, status);
if (f == NULL) {
delete syms;
status = U_MEMORY_ALLOCATION_ERROR;
return NULL;
}
}
f->setLocaleIDs(ures_getLocaleByType(ownedResource.getAlias(), ULOC_VALID_LOCALE, &status),
ures_getLocaleByType(ownedResource.getAlias(), ULOC_ACTUAL_LOCALE, &status));
if (U_FAILURE(status)) {
delete f;
return NULL;
}
return f;
}
BreakIterator *
SimpleFilteredBreakIteratorBuilder::build(BreakIterator* adoptBreakIterator, UErrorCode& status) {
LocalPointer<BreakIterator> adopt(adoptBreakIterator);
if(U_FAILURE(status)) {
return NULL;
}
LocalPointer<UCharsTrieBuilder> builder(new UCharsTrieBuilder(status));
LocalPointer<UCharsTrieBuilder> builder2(new UCharsTrieBuilder(status));
int32_t revCount = 0;
int32_t fwdCount = 0;
int32_t subCount = fSet.size();
LocalArray<UnicodeString> ustrs(new UnicodeString[subCount]);
LocalArray<int> partials(new int[subCount]);
LocalPointer<UCharsTrie> backwardsTrie; // i.e. ".srM" for Mrs.
LocalPointer<UCharsTrie> forwardsPartialTrie; // Has ".a" for "a.M."
int n=0;
for ( set<UnicodeString>::iterator i = fSet.begin();
i != fSet.end();
i++) {
const UnicodeString &abbr = *i;
ustrs[n] = abbr;
partials[n] = 0; // default: not partial
n++;
}
// first pass - find partials.
for(int i=0;i<subCount;i++) {
int nn = ustrs[i].indexOf(kFULLSTOP); // TODO: non-'.' abbreviations
if(nn>-1 && (nn+1)!=ustrs[i].length()) {
//if(true) u_printf("Is a partial: /%S/\n", ustrs[i].getTerminatedBuffer());
// is partial.
// is it unique?
int sameAs = -1;
for(int j=0;j<subCount;j++) {
if(j==i) continue;
if(ustrs[i].compare(0,nn+1,ustrs[j],0,nn+1)==0) {
//if(true) u_printf("Prefix match: /%S/ to %d\n", ustrs[j].getTerminatedBuffer(), nn+1);
//UBool otherIsPartial = ((nn+1)!=ustrs[j].length()); // true if ustrs[j] doesn't end at nn
if(partials[j]==0) { // hasn't been processed yet
partials[j] = kSuppressInReverse | kAddToForward;
//if(true) u_printf("Suppressing: /%S/\n", ustrs[j].getTerminatedBuffer());
} else if(partials[j] & kSuppressInReverse) {
sameAs = j; // the other entry is already in the reverse table.
}
}
}
//if(debug2) u_printf("for partial /%S/ same=%d partials=%d\n", ustrs[i].getTerminatedBuffer(), sameAs, partials[i]);
UnicodeString prefix(ustrs[i], 0, nn+1);
if(sameAs == -1 && partials[i] == 0) {
// first one - add the prefix to the reverse table.
prefix.reverse();
builder->add(prefix, kPARTIAL, status);
revCount++;
//if(debug2) u_printf("Added Partial: /%S/ from /%S/ status=%s\n", prefix.getTerminatedBuffer(), ustrs[i].getTerminatedBuffer(), u_errorName(status));
partials[i] = kSuppressInReverse | kAddToForward;
} else {
//if(debug2) u_printf(" // not adding partial for /%S/ from /%S/\n", prefix.getTerminatedBuffer(), ustrs[i].getTerminatedBuffer());
}
}
}
for(int i=0;i<subCount;i++) {
if(partials[i]==0) {
ustrs[i].reverse();
builder->add(ustrs[i], kMATCH, status);
revCount++;
//if(debug2) u_printf("Added: /%S/ status=%s\n", ustrs[i].getTerminatedBuffer(), u_errorName(status));
} else {
//if(debug2) u_printf(" Adding fwd: /%S/\n", ustrs[i].getTerminatedBuffer());
// an optimization would be to only add the portion after the '.'
// for example, for "Ph.D." we store ".hP" in the reverse table. We could just store "D." in the forward,
// instead of "Ph.D." since we already know the "Ph." part is a match.
// would need the trie to be able to hold 0-length strings, though.
builder2->add(ustrs[i], kMATCH, status); // forward
fwdCount++;
//ustrs[i].reverse();
////if(debug2) u_printf("SUPPRESS- not Added(%d): /%S/ status=%s\n",partials[i], ustrs[i].getTerminatedBuffer(), u_errorName(status));
}
}
//if(debug) u_printf(" %s has %d abbrs.\n", fJSONSource.c_str(), subCount);
if(revCount>0) {
backwardsTrie.adoptInstead(builder->build(USTRINGTRIE_BUILD_FAST, status));
if(U_FAILURE(status)) {
//printf("Error %s building backwards\n", u_errorName(status));
return NULL;
}
}
if(fwdCount>0) {
forwardsPartialTrie.adoptInstead(builder2->build(USTRINGTRIE_BUILD_FAST, status));
if(U_FAILURE(status)) {
//printf("Error %s building forwards\n", u_errorName(status));
return NULL;
}
}
return new ULISentenceBreakIterator(adopt.orphan(), forwardsPartialTrie.orphan(), backwardsTrie.orphan(), status);
}