const CollationKey&
CollationKey::operator=(const CollationKey& other)
{
if (this != &other)
{
if (other.isBogus())
{
return setToBogus();
}
if (other.fBytes != NULL)
{
ensureCapacity(other.fCount);
if (isBogus())
{
return *this;
}
fHashCode = other.fHashCode;
uprv_memcpy(fBytes, other.fBytes, fCount);
}
else
{
fCount = 0;
fBogus = FALSE;
fHashCode = kEmptyHashCode;
}
}
return *this;
}
StatusWith<std::unique_ptr<CollatorInterface>> CollatorFactoryICU::makeFromBSON(
const BSONObj& spec) {
// Parse the locale ID out of the spec.
auto parsedLocaleID = parseLocaleID(spec);
if (!parsedLocaleID.isOK()) {
return parsedLocaleID.getStatus();
}
// If spec = {locale: "simple"}, return a null pointer. A null CollatorInterface indicates
// simple binary compare.
if (parsedLocaleID.getValue() == CollationSpec::kSimpleBinaryComparison) {
if (spec.nFields() > 1) {
return {ErrorCodes::FailedToParse,
str::stream() << "If " << CollationSpec::kLocaleField << "="
<< CollationSpec::kSimpleBinaryComparison
<< ", no other fields should be present in: " << spec};
}
return {nullptr};
}
// Construct an icu::Locale.
auto userLocale = icu::Locale::createFromName(parsedLocaleID.getValue().c_str());
if (userLocale.isBogus()) {
return {ErrorCodes::BadValue,
str::stream() << "Field '" << CollationSpec::kLocaleField
<< "' is not valid in: " << spec};
}
// Construct an icu::Collator.
UErrorCode status = U_ZERO_ERROR;
std::unique_ptr<icu::Collator> icuCollator(icu::Collator::createInstance(userLocale, status));
if (U_FAILURE(status)) {
icu::ErrorCode icuError;
icuError.set(status);
return {ErrorCodes::OperationFailed,
str::stream() << "Failed to create collator: " << icuError.errorName()
<< ". Collation spec: " << spec};
}
Status localeValidationStatus = validateLocaleID(spec, parsedLocaleID.getValue(), *icuCollator);
if (!localeValidationStatus.isOK()) {
return localeValidationStatus;
}
// Construct a CollationSpec using the options provided in spec or the defaults in icuCollator.
// Use userLocale.getName() for the localeID, since it is canonicalized and includes options.
auto parsedSpec = parseToCollationSpec(spec, userLocale.getName(), icuCollator.get());
if (!parsedSpec.isOK()) {
return parsedSpec.getStatus();
}
auto mongoCollator = stdx::make_unique<CollatorInterfaceICU>(std::move(parsedSpec.getValue()),
std::move(icuCollator));
return {std::move(mongoCollator)};
}
/**
* Generic filter-based scanning code for UCD property UnicodeSets.
*/
void UnicodeSet::applyFilter(UnicodeSet::Filter filter,
void* context,
int32_t src,
UErrorCode &status) {
// Walk through all Unicode characters, noting the start
// and end of each range for which filter.contain(c) is
// true. Add each range to a set.
//
// To improve performance, use the INCLUSIONS set, which
// encodes information about character ranges that are known
// to have identical properties. INCLUSIONS contains
// only the first characters of such ranges.
//
// TODO Where possible, instead of scanning over code points,
// use internal property data to initialize UnicodeSets for
// those properties. Scanning code points is slow.
if (U_FAILURE(status)) return;
const UnicodeSet* inclusions = getInclusions(src, status);
if (U_FAILURE(status)) {
return;
}
clear();
UChar32 startHasProperty = -1;
int32_t limitRange = inclusions->getRangeCount();
for (int j=0; j<limitRange; ++j) {
// get current range
UChar32 start = inclusions->getRangeStart(j);
UChar32 end = inclusions->getRangeEnd(j);
// for all the code points in the range, process
for (UChar32 ch = start; ch <= end; ++ch) {
// only add to this UnicodeSet on inflection points --
// where the hasProperty value changes to false
if ((*filter)(ch, context)) {
if (startHasProperty < 0) {
startHasProperty = ch;
}
} else if (startHasProperty >= 0) {
add(startHasProperty, ch-1);
startHasProperty = -1;
}
}
}
if (startHasProperty >= 0) {
add((UChar32)startHasProperty, (UChar32)0x10FFFF);
}
if (isBogus() && U_SUCCESS(status)) {
// We likely ran out of memory. AHHH!
status = U_MEMORY_ALLOCATION_ERROR;
}
}
UnicodeSet& UnicodeSet::closeOver(int32_t attribute) {
if (isFrozen() || isBogus()) {
return *this;
}
if (attribute & (USET_CASE_INSENSITIVE | USET_ADD_CASE_MAPPINGS)) {
const UCaseProps *csp = ucase_getSingleton();
{
UnicodeSet foldSet(*this);
UnicodeString str;
USetAdder sa = {
foldSet.toUSet(),
_set_add,
_set_addRange,
_set_addString,
NULL, // don't need remove()
NULL // don't need removeRange()
};
// start with input set to guarantee inclusion
// USET_CASE: remove strings because the strings will actually be reduced (folded);
// therefore, start with no strings and add only those needed
if (attribute & USET_CASE_INSENSITIVE) {
foldSet.strings->removeAllElements();
}
int32_t n = getRangeCount();
UChar32 result;
const UChar *full;
int32_t locCache = 0;
for (int32_t i=0; i<n; ++i) {
UChar32 start = getRangeStart(i);
UChar32 end = getRangeEnd(i);
if (attribute & USET_CASE_INSENSITIVE) {
// full case closure
for (UChar32 cp=start; cp<=end; ++cp) {
ucase_addCaseClosure(csp, cp, &sa);
}
} else {
// add case mappings
// (does not add long s for regular s, or Kelvin for k, for example)
for (UChar32 cp=start; cp<=end; ++cp) {
result = ucase_toFullLower(csp, cp, NULL, NULL, &full, "", &locCache);
addCaseMapping(foldSet, result, full, str);
result = ucase_toFullTitle(csp, cp, NULL, NULL, &full, "", &locCache);
addCaseMapping(foldSet, result, full, str);
result = ucase_toFullUpper(csp, cp, NULL, NULL, &full, "", &locCache);
addCaseMapping(foldSet, result, full, str);
result = ucase_toFullFolding(csp, cp, &full, 0);
addCaseMapping(foldSet, result, full, str);
}
}
}
if (strings != NULL && strings->size() > 0) {
if (attribute & USET_CASE_INSENSITIVE) {
for (int32_t j=0; j<strings->size(); ++j) {
str = *(const UnicodeString *) strings->elementAt(j);
str.foldCase();
if(!ucase_addStringCaseClosure(csp, str.getBuffer(), str.length(), &sa)) {
foldSet.add(str); // does not map to code points: add the folded string itself
}
}
} else {
Locale root("");
#if !UCONFIG_NO_BREAK_ITERATION
UErrorCode status = U_ZERO_ERROR;
BreakIterator *bi = BreakIterator::createWordInstance(root, status);
if (U_SUCCESS(status)) {
#endif
const UnicodeString *pStr;
for (int32_t j=0; j<strings->size(); ++j) {
pStr = (const UnicodeString *) strings->elementAt(j);
(str = *pStr).toLower(root);
foldSet.add(str);
#if !UCONFIG_NO_BREAK_ITERATION
(str = *pStr).toTitle(bi, root);
foldSet.add(str);
#endif
(str = *pStr).toUpper(root);
foldSet.add(str);
(str = *pStr).foldCase();
foldSet.add(str);
}
#if !UCONFIG_NO_BREAK_ITERATION
}
delete bi;
#endif
}
}
*this = foldSet;
}
}
return *this;
}
UnicodeSet&
UnicodeSet::applyPropertyAlias(const UnicodeString& prop,
const UnicodeString& value,
UErrorCode& ec) {
if (U_FAILURE(ec) || isFrozen()) return *this;
// prop and value used to be converted to char * using the default
// converter instead of the invariant conversion.
// This should not be necessary because all Unicode property and value
// names use only invariant characters.
// If there are any variant characters, then we won't find them anyway.
// Checking first avoids assertion failures in the conversion.
if( !uprv_isInvariantUString(prop.getBuffer(), prop.length()) ||
!uprv_isInvariantUString(value.getBuffer(), value.length())
) {
FAIL(ec);
}
CharString pname, vname;
pname.appendInvariantChars(prop, ec);
vname.appendInvariantChars(value, ec);
if (U_FAILURE(ec)) return *this;
UProperty p;
int32_t v;
UBool mustNotBeEmpty = FALSE, invert = FALSE;
if (value.length() > 0) {
p = u_getPropertyEnum(pname.data());
if (p == UCHAR_INVALID_CODE) FAIL(ec);
// Treat gc as gcm
if (p == UCHAR_GENERAL_CATEGORY) {
p = UCHAR_GENERAL_CATEGORY_MASK;
}
if ((p >= UCHAR_BINARY_START && p < UCHAR_BINARY_LIMIT) ||
(p >= UCHAR_INT_START && p < UCHAR_INT_LIMIT) ||
(p >= UCHAR_MASK_START && p < UCHAR_MASK_LIMIT)) {
v = u_getPropertyValueEnum(p, vname.data());
if (v == UCHAR_INVALID_CODE) {
// Handle numeric CCC
if (p == UCHAR_CANONICAL_COMBINING_CLASS ||
p == UCHAR_TRAIL_CANONICAL_COMBINING_CLASS ||
p == UCHAR_LEAD_CANONICAL_COMBINING_CLASS) {
char* end;
double value = uprv_strtod(vname.data(), &end);
v = (int32_t) value;
if (v != value || v < 0 || *end != 0) {
// non-integral or negative value, or trailing junk
FAIL(ec);
}
// If the resultant set is empty then the numeric value
// was invalid.
mustNotBeEmpty = TRUE;
} else {
FAIL(ec);
}
}
}
else {
switch (p) {
case UCHAR_NUMERIC_VALUE:
{
char* end;
double value = uprv_strtod(vname.data(), &end);
if (*end != 0) {
FAIL(ec);
}
applyFilter(numericValueFilter, &value, UPROPS_SRC_CHAR, ec);
return *this;
}
case UCHAR_NAME:
{
// Must munge name, since u_charFromName() does not do
// 'loose' matching.
char buf[128]; // it suffices that this be > uprv_getMaxCharNameLength
if (!mungeCharName(buf, vname.data(), sizeof(buf))) FAIL(ec);
UChar32 ch = u_charFromName(U_EXTENDED_CHAR_NAME, buf, &ec);
if (U_SUCCESS(ec)) {
clear();
add(ch);
return *this;
} else {
FAIL(ec);
}
}
case UCHAR_UNICODE_1_NAME:
// ICU 49 deprecates the Unicode_1_Name property APIs.
FAIL(ec);
case UCHAR_AGE:
{
// Must munge name, since u_versionFromString() does not do
// 'loose' matching.
char buf[128];
if (!mungeCharName(buf, vname.data(), sizeof(buf))) FAIL(ec);
UVersionInfo version;
u_versionFromString(version, buf);
applyFilter(versionFilter, &version, UPROPS_SRC_PROPSVEC, ec);
return *this;
}
case UCHAR_SCRIPT_EXTENSIONS:
v = u_getPropertyValueEnum(UCHAR_SCRIPT, vname.data());
if (v == UCHAR_INVALID_CODE) {
FAIL(ec);
}
// fall through to calling applyIntPropertyValue()
break;
default:
// p is a non-binary, non-enumerated property that we
// don't support (yet).
FAIL(ec);
}
}
}
else {
// value is empty. Interpret as General Category, Script, or
// Binary property.
p = UCHAR_GENERAL_CATEGORY_MASK;
v = u_getPropertyValueEnum(p, pname.data());
if (v == UCHAR_INVALID_CODE) {
p = UCHAR_SCRIPT;
v = u_getPropertyValueEnum(p, pname.data());
if (v == UCHAR_INVALID_CODE) {
p = u_getPropertyEnum(pname.data());
if (p >= UCHAR_BINARY_START && p < UCHAR_BINARY_LIMIT) {
v = 1;
} else if (0 == uprv_comparePropertyNames(ANY, pname.data())) {
set(MIN_VALUE, MAX_VALUE);
return *this;
} else if (0 == uprv_comparePropertyNames(ASCII, pname.data())) {
set(0, 0x7F);
return *this;
} else if (0 == uprv_comparePropertyNames(ASSIGNED, pname.data())) {
// [:Assigned:]=[:^Cn:]
p = UCHAR_GENERAL_CATEGORY_MASK;
v = U_GC_CN_MASK;
invert = TRUE;
} else {
FAIL(ec);
}
}
}
}
applyIntPropertyValue(p, v, ec);
if(invert) {
complement();
}
if (U_SUCCESS(ec) && (mustNotBeEmpty && isEmpty())) {
// mustNotBeEmpty is set to true if an empty set indicates
// invalid input.
ec = U_ILLEGAL_ARGUMENT_ERROR;
}
if (isBogus() && U_SUCCESS(ec)) {
// We likely ran out of memory. AHHH!
ec = U_MEMORY_ALLOCATION_ERROR;
}
return *this;
}
/**
* Parse the pattern from the given RuleCharacterIterator. The
* iterator is advanced over the parsed pattern.
* @param chars iterator over the pattern characters. Upon return
* it will be advanced to the first character after the parsed
* pattern, or the end of the iteration if all characters are
* parsed.
* @param symbols symbol table to use to parse and dereference
* variables, or null if none.
* @param rebuiltPat the pattern that was parsed, rebuilt or
* copied from the input pattern, as appropriate.
* @param options a bit mask of zero or more of the following:
* IGNORE_SPACE, CASE.
*/
void UnicodeSet::applyPattern(RuleCharacterIterator& chars,
const SymbolTable* symbols,
UnicodeString& rebuiltPat,
uint32_t options,
UnicodeSet& (UnicodeSet::*caseClosure)(int32_t attribute),
UErrorCode& ec) {
if (U_FAILURE(ec)) return;
// Syntax characters: [ ] ^ - & { }
// Recognized special forms for chars, sets: c-c s-s s&s
int32_t opts = RuleCharacterIterator::PARSE_VARIABLES |
RuleCharacterIterator::PARSE_ESCAPES;
if ((options & USET_IGNORE_SPACE) != 0) {
opts |= RuleCharacterIterator::SKIP_WHITESPACE;
}
UnicodeString patLocal, buf;
UBool usePat = FALSE;
UnicodeSetPointer scratch;
RuleCharacterIterator::Pos backup;
// mode: 0=before [, 1=between [...], 2=after ]
// lastItem: 0=none, 1=char, 2=set
int8_t lastItem = 0, mode = 0;
UChar32 lastChar = 0;
UChar op = 0;
UBool invert = FALSE;
clear();
while (mode != 2 && !chars.atEnd()) {
U_ASSERT((lastItem == 0 && op == 0) ||
(lastItem == 1 && (op == 0 || op == HYPHEN /*'-'*/)) ||
(lastItem == 2 && (op == 0 || op == HYPHEN /*'-'*/ ||
op == INTERSECTION /*'&'*/)));
UChar32 c = 0;
UBool literal = FALSE;
UnicodeSet* nested = 0; // alias - do not delete
// -------- Check for property pattern
// setMode: 0=none, 1=unicodeset, 2=propertypat, 3=preparsed
int8_t setMode = 0;
if (resemblesPropertyPattern(chars, opts)) {
setMode = 2;
}
// -------- Parse '[' of opening delimiter OR nested set.
// If there is a nested set, use `setMode' to define how
// the set should be parsed. If the '[' is part of the
// opening delimiter for this pattern, parse special
// strings "[", "[^", "[-", and "[^-". Check for stand-in
// characters representing a nested set in the symbol
// table.
else {
// Prepare to backup if necessary
chars.getPos(backup);
c = chars.next(opts, literal, ec);
if (U_FAILURE(ec)) return;
if (c == 0x5B /*'['*/ && !literal) {
if (mode == 1) {
chars.setPos(backup); // backup
setMode = 1;
} else {
// Handle opening '[' delimiter
mode = 1;
patLocal.append((UChar) 0x5B /*'['*/);
chars.getPos(backup); // prepare to backup
c = chars.next(opts, literal, ec);
if (U_FAILURE(ec)) return;
if (c == 0x5E /*'^'*/ && !literal) {
invert = TRUE;
patLocal.append((UChar) 0x5E /*'^'*/);
chars.getPos(backup); // prepare to backup
c = chars.next(opts, literal, ec);
if (U_FAILURE(ec)) return;
}
// Fall through to handle special leading '-';
// otherwise restart loop for nested [], \p{}, etc.
if (c == HYPHEN /*'-'*/) {
literal = TRUE;
// Fall through to handle literal '-' below
} else {
chars.setPos(backup); // backup
continue;
}
}
} else if (symbols != 0) {
const UnicodeFunctor *m = symbols->lookupMatcher(c);
if (m != 0) {
const UnicodeSet *ms = dynamic_cast<const UnicodeSet *>(m);
if (ms == NULL) {
ec = U_MALFORMED_SET;
return;
}
// casting away const, but `nested' won't be modified
// (important not to modify stored set)
nested = const_cast<UnicodeSet*>(ms);
setMode = 3;
}
}
}
// -------- Handle a nested set. This either is inline in
// the pattern or represented by a stand-in that has
// previously been parsed and was looked up in the symbol
// table.
if (setMode != 0) {
if (lastItem == 1) {
if (op != 0) {
// syntaxError(chars, "Char expected after operator");
ec = U_MALFORMED_SET;
return;
}
add(lastChar, lastChar);
_appendToPat(patLocal, lastChar, FALSE);
lastItem = 0;
op = 0;
}
if (op == HYPHEN /*'-'*/ || op == INTERSECTION /*'&'*/) {
patLocal.append(op);
}
if (nested == 0) {
// lazy allocation
if (!scratch.allocate()) {
ec = U_MEMORY_ALLOCATION_ERROR;
return;
}
nested = scratch.pointer();
}
switch (setMode) {
case 1:
nested->applyPattern(chars, symbols, patLocal, options, caseClosure, ec);
break;
case 2:
chars.skipIgnored(opts);
nested->applyPropertyPattern(chars, patLocal, ec);
if (U_FAILURE(ec)) return;
break;
case 3: // `nested' already parsed
nested->_toPattern(patLocal, FALSE);
break;
}
usePat = TRUE;
if (mode == 0) {
// Entire pattern is a category; leave parse loop
*this = *nested;
mode = 2;
break;
}
switch (op) {
case HYPHEN: /*'-'*/
removeAll(*nested);
break;
case INTERSECTION: /*'&'*/
retainAll(*nested);
break;
case 0:
addAll(*nested);
break;
}
op = 0;
lastItem = 2;
continue;
}
if (mode == 0) {
// syntaxError(chars, "Missing '['");
ec = U_MALFORMED_SET;
return;
}
// -------- Parse special (syntax) characters. If the
// current character is not special, or if it is escaped,
// then fall through and handle it below.
if (!literal) {
switch (c) {
case 0x5D /*']'*/:
if (lastItem == 1) {
add(lastChar, lastChar);
_appendToPat(patLocal, lastChar, FALSE);
}
// Treat final trailing '-' as a literal
if (op == HYPHEN /*'-'*/) {
add(op, op);
patLocal.append(op);
} else if (op == INTERSECTION /*'&'*/) {
// syntaxError(chars, "Trailing '&'");
ec = U_MALFORMED_SET;
return;
}
patLocal.append((UChar) 0x5D /*']'*/);
mode = 2;
continue;
case HYPHEN /*'-'*/:
if (op == 0) {
if (lastItem != 0) {
op = (UChar) c;
continue;
} else {
// Treat final trailing '-' as a literal
add(c, c);
c = chars.next(opts, literal, ec);
if (U_FAILURE(ec)) return;
if (c == 0x5D /*']'*/ && !literal) {
patLocal.append(HYPHEN_RIGHT_BRACE, 2);
mode = 2;
continue;
}
}
}
// syntaxError(chars, "'-' not after char or set");
ec = U_MALFORMED_SET;
return;
case INTERSECTION /*'&'*/:
if (lastItem == 2 && op == 0) {
op = (UChar) c;
continue;
}
// syntaxError(chars, "'&' not after set");
ec = U_MALFORMED_SET;
return;
case 0x5E /*'^'*/:
// syntaxError(chars, "'^' not after '['");
ec = U_MALFORMED_SET;
return;
case 0x7B /*'{'*/:
if (op != 0) {
// syntaxError(chars, "Missing operand after operator");
ec = U_MALFORMED_SET;
return;
}
if (lastItem == 1) {
add(lastChar, lastChar);
_appendToPat(patLocal, lastChar, FALSE);
}
lastItem = 0;
buf.truncate(0);
{
UBool ok = FALSE;
while (!chars.atEnd()) {
c = chars.next(opts, literal, ec);
if (U_FAILURE(ec)) return;
if (c == 0x7D /*'}'*/ && !literal) {
ok = TRUE;
break;
}
buf.append(c);
}
if (buf.length() < 1 || !ok) {
// syntaxError(chars, "Invalid multicharacter string");
ec = U_MALFORMED_SET;
return;
}
}
// We have new string. Add it to set and continue;
// we don't need to drop through to the further
// processing
add(buf);
patLocal.append((UChar) 0x7B /*'{'*/);
_appendToPat(patLocal, buf, FALSE);
patLocal.append((UChar) 0x7D /*'}'*/);
continue;
case SymbolTable::SYMBOL_REF:
// symbols nosymbols
// [a-$] error error (ambiguous)
// [a$] anchor anchor
// [a-$x] var "x"* literal '$'
// [a-$.] error literal '$'
// *We won't get here in the case of var "x"
{
chars.getPos(backup);
c = chars.next(opts, literal, ec);
if (U_FAILURE(ec)) return;
UBool anchor = (c == 0x5D /*']'*/ && !literal);
if (symbols == 0 && !anchor) {
c = SymbolTable::SYMBOL_REF;
chars.setPos(backup);
break; // literal '$'
}
if (anchor && op == 0) {
if (lastItem == 1) {
add(lastChar, lastChar);
_appendToPat(patLocal, lastChar, FALSE);
}
add(U_ETHER);
usePat = TRUE;
patLocal.append((UChar) SymbolTable::SYMBOL_REF);
patLocal.append((UChar) 0x5D /*']'*/);
mode = 2;
continue;
}
// syntaxError(chars, "Unquoted '$'");
ec = U_MALFORMED_SET;
return;
}
default:
break;
}
}
// -------- Parse literal characters. This includes both
// escaped chars ("\u4E01") and non-syntax characters
// ("a").
switch (lastItem) {
case 0:
lastItem = 1;
lastChar = c;
break;
case 1:
if (op == HYPHEN /*'-'*/) {
if (lastChar >= c) {
// Don't allow redundant (a-a) or empty (b-a) ranges;
// these are most likely typos.
// syntaxError(chars, "Invalid range");
ec = U_MALFORMED_SET;
return;
}
add(lastChar, c);
_appendToPat(patLocal, lastChar, FALSE);
patLocal.append(op);
_appendToPat(patLocal, c, FALSE);
lastItem = 0;
op = 0;
} else {
add(lastChar, lastChar);
_appendToPat(patLocal, lastChar, FALSE);
lastChar = c;
}
break;
case 2:
if (op != 0) {
// syntaxError(chars, "Set expected after operator");
ec = U_MALFORMED_SET;
return;
}
lastChar = c;
lastItem = 1;
break;
}
}
if (mode != 2) {
// syntaxError(chars, "Missing ']'");
ec = U_MALFORMED_SET;
return;
}
chars.skipIgnored(opts);
/**
* Handle global flags (invert, case insensitivity). If this
* pattern should be compiled case-insensitive, then we need
* to close over case BEFORE COMPLEMENTING. This makes
* patterns like /[^abc]/i work.
*/
if ((options & USET_CASE_INSENSITIVE) != 0) {
(this->*caseClosure)(USET_CASE_INSENSITIVE);
}
else if ((options & USET_ADD_CASE_MAPPINGS) != 0) {
(this->*caseClosure)(USET_ADD_CASE_MAPPINGS);
}
if (invert) {
complement();
}
// Use the rebuilt pattern (patLocal) only if necessary. Prefer the
// generated pattern.
if (usePat) {
rebuiltPat.append(patLocal);
} else {
_generatePattern(rebuiltPat, FALSE);
}
if (isBogus() && U_SUCCESS(ec)) {
// We likely ran out of memory. AHHH!
ec = U_MEMORY_ALLOCATION_ERROR;
}
}
