/**
* Creates the {@link Document} for {@link KnowledgeBasePathIndex}
*
* @param kb the knowledge base to index
* @throws IOException
*/
public static void createIndex(KnowledgeBase kb) throws IOException {
File txtFileDir =
new File(
KnowledgeConfiguration.getInstance().getKnowledgeBasePathIndexDir(kb) + TXT_FILE_DIR);
File indexDir =
new File(KnowledgeConfiguration.getInstance().getKnowledgeBasePathIndexDir(kb) + INDEX_DIR);
createIndex(txtFileDir, indexDir);
}
/**
* Creates the files containing all paths for all concepts up to a certain depth
*
* @param kb
* @param concepts
* @throws IOException
*/
public static void createIndexFiles(KnowledgeBase kb, Collection<String> concepts)
throws IOException {
Iterator<String> conceptIterator = kb.getConceptIterator();
String outputDir =
KnowledgeConfiguration.getInstance().getKnowledgeBasePathIndexDir(kb) + TXT_FILE_DIR;
if (!Files.isDirectoryEmpty(new File(outputDir))) {
log.warn(outputDir + " IS NOT EMPTY, BAILING OUT ...");
return;
}
int maxSearchDepth = getMaxSearchDepth();
DirectoryFileManager dirManager = new DirectoryFileManager(outputDir, 1000);
log.info("CREATING THE INDEXABLE PATH FILES FOR: " + kb.name() + " INTO " + outputDir);
int idx = 0;
List<String> goodConcepts = new ArrayList<String>();
while (conceptIterator.hasNext()) {
String concept = conceptIterator.next();
if (concepts.isEmpty() || concepts.contains(concept)) goodConcepts.add(concept);
idx++;
if ((idx % 100000) == 0) log.info("ITERATED THROUGH " + idx + " CONCEPTS SO FAR ...");
}
List<List<String>> splitConcepts =
Collections.split(goodConcepts, goodConcepts.size() / NSPLITS);
int nThreads = splitConcepts.size();
ExecutorService threadExecutor = Executors.newFixedThreadPool(nThreads);
CountDownLatch doneSignal = new CountDownLatch(nThreads);
for (int i = 0; i < nThreads; i++) {
ConceptIndexer worker =
new ConceptIndexer(splitConcepts.get(i), i, dirManager, kb, maxSearchDepth, doneSignal);
threadExecutor.execute(worker);
}
Timer timer = new Timer();
try {
doneSignal.await();
threadExecutor.shutdown();
} catch (InterruptedException ie) {
ie.printStackTrace();
}
timer.tick("TO CREATE INDEXABLE FILES");
}
/**
* Creates the {@link Document} for {@link KnowledgeBasePathIndex} for a bunch of input concepts
* only
*
* @param kb
* @throws IOException
*/
public static void createIndex(KnowledgeBase kb, Collection<String> concepts) throws IOException {
File txtFileDir =
new File(
KnowledgeConfiguration.getInstance().getKnowledgeBasePathIndexDir(kb) + TXT_FILE_DIR);
// init the index dirs and the index writer
File indexDir =
new File(KnowledgeConfiguration.getInstance().getKnowledgeBasePathIndexDir(kb) + INDEX_DIR);
if (!indexDir.exists()) indexDir.mkdirs();
else if (!Files.isDirectoryEmpty(indexDir)) {
log.warn(indexDir + " IS NOT EMPTY, BAILING OUT ...");
return;
}
IndexWriter indexWriter =
new IndexWriter(
new SimpleFSDirectory(indexDir),
new StandardAnalyzer(Version.LUCENE_29),
true,
IndexWriter.MaxFieldLength.UNLIMITED);
int counter = 0;
FilenameFilter filter =
new FilenameFilter() {
public boolean accept(File dir, String name) {
return name.endsWith(".txt");
}
};
Collection<File> files = Files.listFiles(txtFileDir, filter, true);
// the intermediate concepts @depth 2 from the starting nodes
Set<String> intermediate = new HashSet<String>();
// PHASE 1. index all paths starting with concepts of interest
for (File file : files) {
String fileName = file.getName();
String concept = fileName.substring(0, fileName.indexOf("."));
if (!concepts.contains(concept)) continue;
BufferedReader reader = new BufferedReader(new FileReader(file));
Set<List<String>> paths = new HashSet<List<String>>();
while (reader.ready()) {
String line = reader.readLine();
String[] path = line.split(" ");
paths.add(Arrays.asList(path));
intermediate.add(path[path.length - 1]);
}
reader.close();
List<Document> docs = getRecords(paths);
for (Document doc : docs) indexWriter.addDocument(doc);
if ((counter % 1000) == 0)
log.info("\tINDEXED STARTING PATHS FOR " + counter + " CONCEPTS SO FAR ... ");
counter++;
}
log.info("\tINDEXED " + counter + " STARTING CONCEPTS, DONE!");
log.info("NOW INDEXING ENDING PATHS FOR " + intermediate.size() + " INTERMEDIATE CONCEPTS");
// PHASE 2. index all paths ending with concepts of interest
counter = 0;
for (File file : files) {
String fileName = file.getName();
String concept = fileName.substring(0, fileName.indexOf("."));
if (concepts.contains(concept)) continue; // already seen
if (!intermediate.contains(concept)) continue; // interested only
// in intermediate
BufferedReader reader = new BufferedReader(new FileReader(file));
Set<List<String>> paths = new HashSet<List<String>>();
while (reader.ready()) {
String line = reader.readLine();
String[] path = line.split(" ");
if (concepts.contains(path[path.length - 1])) paths.add(Arrays.asList(path));
}
reader.close();
List<Document> docs = getRecords(paths);
for (Document doc : docs) indexWriter.addDocument(doc);
if ((counter % 1000) == 0)
log.info("\tINDEXED ENDING PATHS FOR " + counter + " CONCEPTS SO FAR ... ");
counter++;
}
log.info("\tINDEXED " + counter + " ENDING CONCEPTS, DONE!");
// closes the index writers
indexWriter.optimize();
indexWriter.close();
}