int main( int argc, char** argv ) {
try {
indri::api::Parameters& parameters = indri::api::Parameters::instance();
parameters.loadCommandLine( argc, argv );
if( parameters.get( "version", false ) ) {
std::cout << INDRI_DISTRIBUTION << std::endl;
}
require_parameter( "corpus", parameters );
require_parameter( "links", parameters );
require_parameter( "output", parameters );
std::string corpusPath = parameters[ "corpus" ];
std::string linkPath = parameters[ "links" ];
std::string outputFile = parameters[ "output" ];
std::string indexPath = parameters.get("index", "");
UINT64 colLen = 0;
indri::parse::PageRank *pr = 0;
if (indexPath.size() > 0) {
int maxIters = parameters.get( "iters", 100 );
double c = parameters.get( "c", 0.85 );
pr = new indri::parse::PageRank ( corpusPath, linkPath, indexPath );
pr->indexPageRank(outputFile, maxIters, c);
} else {
int docsPerIter = parameters.get( "docs", 1000 );
int maxIters = parameters.get( "iters", 10 );
double c = parameters.get( "c", 0.5 );
pr = new indri::parse::PageRank( corpusPath, linkPath, colLen );
pr->computePageRank( outputFile, maxIters, docsPerIter, c );
}
if( parameters.get( "writeRaw", false ) ) {
std::string rawFile = outputFile + ".raw";
pr->writeRaw( outputFile, rawFile );
}
// default is to produce a prior file for makeprior.
if( parameters.get( "writePriors", true ) ) {
std::string priorFile = outputFile + ".prior";
pr->writePriors( outputFile, priorFile );
}
if( parameters.get( "writeRanks", false ) ) {
std::string ranksFile = outputFile + ".ranks";
pr->writeRanks( outputFile, ranksFile );
}
// don't really need the outputFile
::remove(outputFile.c_str());
delete pr;
}
catch( lemur::api::Exception& e ) {
LEMUR_ABORT(e);
}
}
int main(int argc, char * argv[]) {
try {
indri::api::Parameters& parameters = indri::api::Parameters::instance();
parameters.loadCommandLine( argc, argv );
require_parameter( "corpus", parameters );
require_parameter( "index", parameters );
StatusMonitor monitor;
indri::api::IndexEnvironment env;
std::string repositoryPath = parameters["index"];
buildindex_start_time();
if( parameters.get( "version", 0 ) ) {
std::cout << INDRI_DISTRIBUTION << std::endl;
}
env.setMemory( parameters.get("memory", INT64(1024*1024*1024)) );
env.setNormalization( parameters.get("normalize", true));
env.setInjectURL( parameters.get("injectURL", true));
env.setStoreDocs( parameters.get("storeDocs", true));
std::string blackList = parameters.get("blacklist", "");
if( blackList.length() ) {
int count = env.setBlackList(blackList);
std::cout << "Added to blacklist: "<< count << std::endl;
std::cout.flush();
}
std::string offsetAnnotationHint=parameters.get("offsetannotationhint", "default");
if (offsetAnnotationHint=="ordered") {
env.setOffsetAnnotationIndexHint(indri::parse::OAHintOrderedAnnotations);
} if (offsetAnnotationHint=="unordered") {
env.setOffsetAnnotationIndexHint(indri::parse::OAHintSizeBuffers);
} else {
env.setOffsetAnnotationIndexHint(indri::parse::OAHintDefault);
}
std::string stemmerName = parameters.get("stemmer.name", "");
if( stemmerName.length() )
env.setStemmer(stemmerName);
std::vector<std::string> stopwords;
if( copy_parameters_to_string_vector( stopwords, parameters, "stopper.word" ) )
env.setStopwords(stopwords);
// fields to include as metadata (unindexed)
std::vector<std::string> metadata;
// metadata fields that should have a forward lookup table.
std::vector<std::string> metadataForward;
// metadata fields that should have a backward lookup table.
std::vector<std::string> metadataBackward;
copy_parameters_to_string_vector( metadata, parameters, "metadata.field" );
downcase_string_vector(metadata);
copy_parameters_to_string_vector( metadataForward, parameters, "metadata.forward" );
downcase_string_vector(metadataForward);
copy_parameters_to_string_vector( metadataBackward, parameters, "metadata.backward" );
downcase_string_vector(metadataBackward);
// docno is a special field, automagically add it as forward and backward.
std::string docno = "docno";
if( std::find( metadataForward.begin(),
metadataForward.end(),
docno ) == metadataForward.end() )
metadataForward.push_back(docno);
if( std::find( metadataBackward.begin(),
metadataBackward.end(),
docno ) == metadataBackward.end() )
metadataBackward.push_back(docno);
env.setMetadataIndexedFields( metadataForward, metadataBackward );
#if 0
// "document" is a special field.
// automagically add it as an indexed field.
indri::api::Parameters field = parameters.append("field");
field.set( "name", "document" );
field.set( "ordinal", true );
field.set("parental", true);
#endif
std::vector<std::string> fields;
std::string subName = "name";
if( copy_parameters_to_string_vector( fields, parameters, "field", &subName ) ) {
downcase_string_vector(fields);
env.setIndexedFields(fields);
process_numeric_fields( parameters, env );
process_ordinal_fields( parameters, env );
process_parental_fields( parameters, env ); //pto
}
if( indri::collection::Repository::exists( repositoryPath ) ) {
// check if the repository was corrupted by an indexing crash
// if so, recover it and continue.
if (_recoverRepository(repositoryPath)) {
env.open( repositoryPath, &monitor );
buildindex_print_event( std::string() + "Opened repository " + repositoryPath );
} else {
// failed to open it, needs to be created from scratch.
// create will remove any cruft.
env.create( repositoryPath, &monitor );
buildindex_print_event( std::string() + "Created repository " + repositoryPath );
}
} else {
env.create( repositoryPath, &monitor );
buildindex_print_event( std::string() + "Created repository " + repositoryPath );
}
indri::api::Parameters corpus = parameters["corpus"];
for( unsigned int i=0; i<corpus.size(); i++ ) {
indri::api::Parameters thisCorpus = corpus[i];
require_parameter( "path", thisCorpus );
std::string corpusPath = thisCorpus["path"];
std::string fileClass = thisCorpus.get("class", "");
// augment field/metadata tags in the environment if needed.
if( fileClass.length() ) {
indri::parse::FileClassEnvironmentFactory::Specification *spec = env.getFileClassSpec(fileClass);
if( spec ) {
// add fields if necessary, only update if changed.
if( augmentSpec( spec, fields, metadata, metadataForward, metadataBackward ) )
env.addFileClass(*spec);
delete(spec);
}
}
bool isDirectory = indri::file::Path::isDirectory( corpusPath );
// First record the document root, and then the paths to any annotator inputs
env.setDocumentRoot( corpusPath );
// Support for anchor text
std::string anchorText = thisCorpus.get("inlink", "");
env.setAnchorTextPath( anchorText );
// Support for offset annotations
std::string offsetAnnotationsPath = thisCorpus.get( "annotations", "" );
env.setOffsetAnnotationsPath( offsetAnnotationsPath );
// Support for offset metadata file
std::string offsetMetadataPath = thisCorpus.get( "metadata", "" );
env.setOffsetMetadataPath( offsetMetadataPath );
if( isDirectory ) {
indri::file::FileTreeIterator files( corpusPath );
for( ; files != indri::file::FileTreeIterator::end(); files++ ) {
if( fileClass.length() )
env.addFile( *files, fileClass );
else {
std::string extension = indri::file::Path::extension( *files );
indri::parse::FileClassEnvironmentFactory::Specification *spec = env.getFileClassSpec(extension);
if( spec ) {
// add fields if necessary, only update if changed.
if( augmentSpec( spec, fields, metadata, metadataForward, metadataBackward ) )
env.addFileClass(*spec);
delete(spec);
}
env.addFile( *files );
}
}
} else {
if( fileClass.length() )
env.addFile( corpusPath, fileClass );
else {
std::string extension = indri::file::Path::extension( corpusPath );
indri::parse::FileClassEnvironmentFactory::Specification *spec = env.getFileClassSpec(extension);
if( spec ) {
// add fields if necessary, only update if changed.
if( augmentSpec( spec, fields, metadata, metadataForward, metadataBackward ) )
env.addFileClass(*spec);
delete(spec);
}
env.addFile( corpusPath );
}
}
}
buildindex_print_event( "Closing index" );
env.close();
buildindex_print_event( "Finished" );
} catch( lemur::api::Exception& e ) {
LEMUR_ABORT(e);
}
return 0;
}
