void indri::index::MemoryIndex::_writeFieldExtents( lemur::api::DOCID_T documentID, indri::utility::greedy_vector<indri::parse::TagExtent *>& indexedTags ) {
indri::utility::HashTable< indri::parse::TagExtent *, int> tagIdMap;
// sort fields
std::sort( indexedTags.begin(), indexedTags.end(), indri::parse::LessTagExtent() );
// we'll add to the end of the fields greedy_vector
indri::utility::greedy_vector<indri::index::FieldExtent> & fields = _termList.fields();
// this is used to set the parentOrdinals
int offset = fields.size();
// convert to field extents, set ids, and create the node map
for( size_t i=0; i<indexedTags.size(); i++ ) {
indri::parse::TagExtent * extent = indexedTags[i];
int ordinal = int(i) + 1;
// this is the id for the field type
int tagId = _fieldID( extent->name );
// convert the field
indri::index::FieldExtent converted( tagId, extent->begin, extent->end, extent->number, ordinal);
// add this node to the map;
tagIdMap.insert( extent, ordinal );
// add his field to the field list for the document
fields.push_back( converted );
// add this location to the inverted list for fields - deferred to below -
// _fieldLists[tagId - 1]->addLocation( documentID, extent->begin, extent->end, extent->number, ordinal );
}
// set the parent ordinals
for( size_t j=0; j<indexedTags.size(); j++ ) {
indri::parse::TagExtent * extent = indexedTags[j];
// look up the parent
int parentOrdinal = 0;
int * parentIter;
if ( extent->parent != 0 ) {
parentIter = tagIdMap.find( extent->parent );
if( parentIter != 0 ) {
parentOrdinal = *parentIter;
} else {
parentOrdinal = 0;
}
}
// set the parent
int ordinal = fields[ offset + j ].ordinal;
int tagId = fields[ offset + j ].id;
fields[ offset + j ].parentOrdinal = parentOrdinal;
// add this location to the inverted list for fields
_fieldLists[tagId - 1]->addLocation( documentID, extent->begin, extent->end, extent->number, ordinal, parentOrdinal );
}
}
void indri::infnet::FixedPassageNode::_buildSubextents( const indri::utility::greedy_vector<indri::index::Extent>& extents ) {
_subextents.clear();
for( size_t i=0; i<extents.size(); i++ ) {
_addSubextents( extents[i] );
}
}
// returns a new instance of FieldInfoList which represents field entities in a document index for a specific field, you must delete the instance later. @see FieldInfoList
// Note that not all index types support fields - those that do should override this method.
lemur::api::FieldInfoList *lemur::index::LemurIndriIndex::fieldInfoList(lemur::api::DOCID_T docID, int fieldID) const {
// get the index for this document
indri::collection::Repository::index_state indexes = _repository->indexes();
indri::index::Index* index = _indexWithDocument( indexes, docID );
// ensure we do have an index (i.e. if the docID was invalid...)
if (!index) return NULL;
// and the indri term-list
const indri::index::TermList *tList=index->termList((int)docID);
// ensure we have a term list!
if (!tList) return NULL;
// create a blank field info list object
lemur::api::IndriFieldInfoList *retVal=new lemur::api::IndriFieldInfoList();
// loop through our fields and insert those that match the field ID
const indri::utility::greedy_vector< indri::index::FieldExtent > fieldVec=tList->fields();
int numFields=fieldVec.size();
for (int i=0; i < numFields; i++) {
indri::index::FieldExtent thisField=fieldVec[i];
if (thisField.id==fieldID) {
retVal->add(thisField);
}
}
delete tList;
// and return
return retVal;
}
void IndexWriter::_pushInvertedLists( indri::utility::greedy_vector<WriterIndexContext*>& lists, invertedlist_pqueue& queue ) {
for( int i=0; i<lists.size(); i++ ) {
lists[i]->iterator->nextEntry();
if( !lists[i]->iterator->finished() )
queue.push( lists[i] );
}
}
void indri::index::MemoryIndex::_removeClosedTags( indri::utility::greedy_vector<indri::parse::TagExtent *>& tags, unsigned int position ) {
for( size_t i=0; i<tags.size(); ) {
if( tags[i]->end <= int(position) ) {
tags.erase( tags.begin() + i );
} else {
i++;
}
}
}
void IndexWriter::_storeMatchInformation( indri::utility::greedy_vector<WriterIndexContext*>& lists, int sequence, indri::index::TermData* termData, UINT64 startOffset, UINT64 endOffset ) {
bool isFrequent = termData->corpus.totalCount > FREQUENT_TERM_COUNT;
if( isFrequent )
_isFrequentCount++;
for( int i=0; i<lists.size(); i++ ) {
WriterIndexContext* list = lists[i];
indri::index::DiskDocListIterator* iterator = dynamic_cast<DiskDocListIterator*>(lists[i]->iterator->currentEntry()->iterator);
bool isMemoryIndex = (iterator == 0);
bool wasFrequent = (isMemoryIndex || iterator->isFrequent());
if( !wasFrequent )
list->wasInfrequentCount++;
if( wasFrequent )
list->wasFrequentCount++;
list->sequenceCount++;
if( !wasFrequent ) {
if( !isFrequent ) {
// common case--remaining infrequent
assert( sequence - _isFrequentCount - 1 >= 0 );
assert( ((sequence -_isFrequentCount - 1) + _isFrequentCount + 1) <= _corpus.uniqueTerms );
list->bitmap->add( list->wasInfrequentCount - 1, sequence - _isFrequentCount - 1 );
} else if( isFrequent ) {
// becoming frequent
list->newlyFrequent->add( list->wasInfrequentCount - 1, termData->term );
}
}
}
if( isFrequent ) {
indri::index::DiskTermData* diskTermData = disktermdata_create( _fields.size() );
::termdata_merge( diskTermData->termData, termData, _fields.size() );
diskTermData->startOffset = startOffset;
diskTermData->length = endOffset - startOffset;
strcpy( const_cast<char*>(diskTermData->termData->term), termData->term );
_topTerms.push_back( diskTermData );
} else {
indri::index::DiskTermData diskTermData;
diskTermData.termData = termData;
diskTermData.startOffset = startOffset;
diskTermData.length = endOffset - startOffset;
diskTermData.termID = sequence - _topTerms.size();
_storeTermEntry( _infrequentTerms, &diskTermData );
}
}
/// sorts the extents by the beginning extent position (if available)
void sortbegin(indri::utility::greedy_vector<indri::index::Extent>& extents){
int sorted=0;
int lastbegin = 0; int lastend = extents.size();
while(lastbegin<lastend){
int i=lastbegin;
int end=lastend-1;
lastbegin=lastend;
lastend=i;
for (;i<end;i++){
if (extents[i].begin > extents[i+1].begin){
indri::index::Extent x(extents[i]);
extents[i]=extents[i+1];extents[i+1]=x;
if(lastbegin>i)lastbegin=i;
if(lastend<i+1)lastend=i+1;
}
}
}
}
void indri::index::MemoryIndex::_addOpenTags( indri::utility::greedy_vector<indri::parse::TagExtent *>& indexedTags,
indri::utility::greedy_vector<indri::parse::TagExtent *>& openTags,
indri::utility::greedy_vector<indri::parse::TagExtent *>& extents,
unsigned int& extentIndex,
unsigned int position ) {
for( ; extentIndex < extents.size(); extentIndex++ ) {
indri::parse::TagExtent* extent = extents[extentIndex];
if( extent->begin > (int)position )
break;
int tagId = _fieldID( extent->name );
if( tagId == 0 )
continue;
openTags.push_back( extent );
indexedTags.push_back( extent );
}
}
void _fetchText( indri::utility::greedy_vector<TagExtent *>& tags, indri::utility::greedy_vector<char*>& terms ) {
// now, fetch the additional terms
char line[65536];
_buffer.clear();
for( int i=0; i<_count; i++ ) {
// LINK
_in.getline( line, sizeof line-1 );
// LINKDOCNO
_in.getline( line, sizeof line-1 );
// TEXT=
_in.getline( line, sizeof line-1 );
int textLen = strlen(line+6);
strcpy( _buffer.write(textLen+1), line+6 );
_buffer.unwrite(1);
assert( *(_buffer.front()+_buffer.position()-1) == '\"' && "Last character should be a quote" );
}
*(_buffer.write(1)) = 0;
// now there's a bunch of text in _buffer, space separated, with each
// link separated by a " symbol
char* beginWord = 0;
int beginIndex = 0;
char* buffer = _buffer.front();
for( unsigned int i=0; i<_buffer.position(); i++ ) {
if( isalnum(buffer[i]) && !beginWord ) {
beginWord = buffer+i;
if(!beginIndex)
beginIndex = terms.size();
} else if( isspace(buffer[i]) ) {
buffer[i] = 0;
if( beginWord )
terms.push_back( beginWord );
beginWord = 0;
} else if( buffer[i] == '\"' ) {
buffer[i] = 0;
if( beginWord )
terms.push_back( beginWord );
beginWord = 0;
TagExtent * extent = new TagExtent;
extent->name = "inlink";
extent->begin = beginIndex;
extent->end = terms.size();
extent->number = 0;
extent->parent = 0;
assert( extent->begin <= extent->end );
if( beginIndex ) {
tags.push_back(extent);
if( terms.size() > 125000 )
break;
}
beginIndex = 0;
}
}
}