void indri::collection::Repository::_buildChain(indri::api::Parameters& parameters, indri::api::Parameters* options) {
// Extract url from metadata before case normalizing.
// this could be parameterized.
if (parameters.get("injectURL", true))
_transformations.push_back(new indri::parse::URLTextAnnotator());
bool dontNormalize = parameters.exists("normalize") && (false == (bool) parameters["normalize"]);
if (dontNormalize == false) {
_transformations.push_back(new indri::parse::NormalizationTransformation());
_transformations.push_back(new indri::parse::UTF8CaseNormalizationTransformation());
}
for(size_t i=0; i<_fields.size(); i++) {
if (_fields[i].parserName == "NumericFieldAnnotator") {
_transformations.push_back(new indri::parse::NumericFieldAnnotator(_fields[i].name));
}
else if (_fields[i].parserName == "DateFieldAnnotator") {
_transformations.push_back(new indri::parse::DateFieldAnnotator(_fields[i].name));
}
}
if (_parameters.exists("stopper.word")) {
indri::api::Parameters stop = _parameters["stopper.word"];
_transformations.push_back(new indri::parse::StopperTransformation(stop));
}
// the transient chain stopwords need to precede the stemmer.
if (options) {
if (options->exists("stopper.word")) {
indri::api::Parameters stop = (*options)["stopper.word"];
_transformations.push_back(new indri::parse::StopperTransformation(stop));
}
}
if (_parameters.exists("stemmer.name")) {
std::string stemmerName = std::string(_parameters["stemmer.name"]);
indri::api::Parameters stemmerParams = _parameters["stemmer"];
_transformations.push_back(indri::parse::StemmerFactory::get(stemmerName, stemmerParams));
}
}
void indri::collection::Repository::_openIndexes(indri::api::Parameters& params, const std::string& parentPath) {
try {
indri::api::Parameters container = params["indexes"];
_active = new index_vector;
_states.push_back(_active);
_indexCount = params.get("indexCount", 0);
if (container.exists("index")) {
indri::api::Parameters indexes = container["index"];
for(size_t i=0; i<indexes.size(); i++) {
indri::api::Parameters indexSpec = indexes[i];
indri::index::DiskIndex* diskIndex = new indri::index::DiskIndex();
std::string indexName = (std::string) indexSpec;
diskIndex->open(parentPath, indexName);
_active->push_back(diskIndex);
}
}
} catch(lemur::api::Exception& e) {
LEMUR_RETHROW(e, "_openIndexes: Couldn't open DiskIndexes because:");
}
}
std::string indri::collection::Repository::_stemmerName(indri::api::Parameters& parameters) {
return parameters.get("stemmer.name", "");
}
void processFields( indri::api::Parameters ¶m ) {
g_timer.start();
std::string index = param.get("index");
std::cout << "Opening: " << index << std::endl;
// make sure this path doesn't exist.
std::string idx2 = index + ".new"; // temp target index.
// presumes a single input oa file for the entire collection.
std::string offsetAnnotationsPath = param.get("annotations");
/// these need to be combined with existing.
// fields to add
// these need to supply numeric/parental/ordinal/etc...
if (param.exists("addField"))
addFields = param["addField"];
// fields to remove
// these only need to be a list of names.
if (param.exists("removeField")) {
indri::api::Parameters slice = param["removeField"];
for (size_t i = 0; i < slice.size(); i++) {
if( slice[i].exists("name") ) {
removeNames.push_back( slice[i]["name"] );
}
}
}
// need to know the file class environment to get the
// conflations right.
std::string className = param.get("fileclass", "");
indri::collection::Repository sourceRepo;
indri::collection::Repository targetRepo;
indri::parse::OffsetAnnotationAnnotator oa_annotator;
indri::parse::FileClassEnvironmentFactory _fileClassFactory;
// Open source repo
sourceRepo.openRead(index);
// Copy its parameters, create target repo, adding or removing
// fields.
repo.loadFile( indri::file::Path::combine( index, "manifest" ) );
int mem = param.get("memory", INT64(100*1024*1024));
repo.set("memory", mem);
adding = addFields.exists("field");
_mergeFields();
// Create the offset annotator.
fce = _fileClassFactory.get( className );
indri::parse::Conflater* conflater = 0;
if( fce ) {
conflater = fce->conflater;
}
if (adding)
{
oa_annotator.setConflater( conflater );
oa_annotator.open( offsetAnnotationsPath );
}
targetRepo.create(idx2, &repo);
// for each document in the source repo, fetch ParsedDocument
// construct full rep, apply annotator, insert into
// target repo.
_index = sourceRepo.indexes()->front(); // presume 1
_docIter = _index->termListFileIterator();
_docIter->startIteration();
// ought to deal with deleted documents here...
// if there are deleted documents, regular add to collection
// if not, only rewrite the indexes, then rename the collection.
indri::index::DeletedDocumentList& deleted = sourceRepo.deletedList();
UINT64 delCount = deleted.deletedCount();
if (delCount > 0)
{
// either warn, compact and then process, or
// do it the old way... FIXME!
std::cerr << "Deleted documents detected... compact with dumpindex first." << std::endl;
return;
}
for (UINT64 docid = 1; docid <= _index->documentCount(); docid++)
{
if ((docid % 500) == 0) {
g_timer.printElapsedSeconds(std::cout);
std::cout << ": " << docid << "\r";
std::cout.flush();
}
parsed = sourceRepo.collection()->retrieve(docid);
// combine field and term data with parsed document
_mergeData();
// apply annotator
if (adding)
parsed = oa_annotator.transform(parsed);
targetRepo.addDocument(parsed, false);
// TagList allocs memory for the tags...
for (size_t i = 0; i < parsed->tags.size(); i++)
delete(parsed->tags[i]);
delete(parsed);
_docIter->nextEntry();
}
std::cout << std::endl;
g_timer.printElapsedSeconds(std::cout);
std::cout << ": " << _index->documentCount() << std::endl;
g_timer.printElapsedSeconds(std::cout);
std::cout << ": closing" << std::endl;
targetRepo.close();
sourceRepo.close();
std::string oldcollectionPath = indri::file::Path::combine( index, "collection" );
std::string newcollectionPath = indri::file::Path::combine( idx2, "collection" );
// clone the collection
indri::file::Path::remove(newcollectionPath);
indri::file::Path::rename(oldcollectionPath, newcollectionPath);
// rename target repo to source repo.
indri::file::Path::remove(index);
indri::file::Path::rename(idx2, index);
g_timer.printElapsedSeconds(std::cout);
std::cout << ": done" << std::endl;
}
