void FTSSpec::scoreDocument(const BSONObj& obj, TermFrequencyMap* term_freqs) const {
if (_textIndexVersion == TEXT_INDEX_VERSION_1) {
return _scoreDocumentV1(obj, term_freqs);
}
FTSElementIterator it(*this, obj);
while (it.more()) {
FTSIteratorValue val = it.next();
std::unique_ptr<FTSTokenizer> tokenizer(val._language->createTokenizer());
_scoreStringV2(tokenizer.get(), val._text, term_freqs, val._weight);
}
}
void FTSSpec::scoreDocument( const BSONObj& obj, TermFrequencyMap* term_freqs ) const {
if ( _textIndexVersion == TEXT_INDEX_VERSION_1 ) {
return _scoreDocumentV1( obj, term_freqs );
}
FTSElementIterator it( *this, obj );
while ( it.more() ) {
FTSIteratorValue val = it.next();
Stemmer stemmer( *val._language );
Tools tools( *val._language, &stemmer, StopWords::getStopWords( *val._language ) );
_scoreStringV2( tools, val._text, term_freqs, val._weight );
}
}
void FTSSpec::scoreDocument( const BSONObj& obj,
const FTSLanguage& parentLanguage,
const string& parentPath,
bool isArray,
TermFrequencyMap* term_freqs ) const {
if ( _textIndexVersion == TEXT_INDEX_VERSION_1 ) {
dassert( parentPath == "" );
dassert( !isArray );
return _scoreDocumentV1( obj, term_freqs );
}
const FTSLanguage& language = _getLanguageToUseV2( obj, parentLanguage );
Stemmer stemmer( language );
Tools tools( language, &stemmer, StopWords::getStopWords( language ) );
// Perform a depth-first traversal of obj, skipping fields not touched by this spec.
BSONObjIterator j( obj );
while ( j.more() ) {
BSONElement elem = j.next();
string fieldName = elem.fieldName();
// Skip "language" specifier fields if wildcard.
if ( wildcard() && languageOverrideField() == fieldName ) {
continue;
}
// Compose the dotted name of the current field:
// 1. parent path empty (top level): use the current field name
// 2. parent path non-empty and obj is an array: use the parent path
// 3. parent path non-empty and obj is a sub-doc: append field name to parent path
string dottedName = ( parentPath.empty() ? fieldName
: isArray ? parentPath
: parentPath + '.' + fieldName );
// Find lower bound of dottedName in _weights. lower_bound leaves us at the first
// weight that could possibly match or be a prefix of dottedName. And if this
// element fails to match, then no subsequent weight can match, since the weights
// are lexicographically ordered.
Weights::const_iterator i = _weights.lower_bound( elem.type() == Object
? dottedName + '.'
: dottedName );
// possibleWeightMatch is set if the weight map contains either a match or some item
// lexicographically larger than fieldName. This boolean acts as a guard on
// dereferences of iterator 'i'.
bool possibleWeightMatch = ( i != _weights.end() );
// Optimize away two cases, when not wildcard:
// 1. lower_bound seeks to end(): no prefix match possible
// 2. lower_bound seeks to a name which is not a prefix
if ( !wildcard() ) {
if ( !possibleWeightMatch ) {
continue;
}
else if ( !_matchPrefix( dottedName, i->first ) ) {
continue;
}
}
// Is the current field an exact match on a weight?
bool exactMatch = ( possibleWeightMatch && i->first == dottedName );
double weight = ( possibleWeightMatch ? i->second : DEFAULT_WEIGHT );
switch ( elem.type() ) {
case String:
// Only index strings on exact match or wildcard.
if ( exactMatch || wildcard() ) {
_scoreStringV2( tools, elem.valuestr(), term_freqs, weight );
}
break;
case Object:
// Only descend into a sub-document on proper prefix or wildcard. Note that
// !exactMatch is a sufficient test for proper prefix match, because of
// matchPrefix() continue block above.
if ( !exactMatch || wildcard() ) {
scoreDocument( elem.Obj(), language, dottedName, false, term_freqs );
}
break;
case Array:
// Only descend into arrays from non-array parents or on wildcard.
if ( !isArray || wildcard() ) {
scoreDocument( elem.Obj(), language, dottedName, true, term_freqs );
}
break;
default:
// Skip over all other BSON types.
break;
}
}
}
