Exemple #1
0
 /** Get an alphabetically sorted list of field names. */
 public List<String> getFieldNames() {
   HashSet<String> names = new HashSet<String>();
   names.addAll(storedFields.keySet());
   names.addAll(reconstructedFields.keySet());
   ArrayList<String> res = new ArrayList<String>(names.size());
   res.addAll(names);
   Collections.sort(res);
   return res;
 }
Exemple #2
0
 /**
  * Reconstruct document fields.
  *
  * @param docNum document number. If this document is deleted, but the index is not optimized yet,
  *     the reconstruction process may still yield the reconstructed field content even from
  *     deleted documents.
  * @return reconstructed document
  * @throws Exception
  */
 public Reconstructed reconstruct(int docNum) throws Exception {
   if (docNum < 0 || docNum > reader.maxDoc()) {
     throw new Exception("Document number outside of valid range.");
   }
   Reconstructed res = new Reconstructed();
   if (deleted != null && deleted.get(docNum)) {
     throw new Exception("Document is deleted.");
   } else {
     Document doc = reader.document(docNum);
     for (int i = 0; i < fieldNames.length; i++) {
       Field[] fs = doc.getFields(fieldNames[i]);
       if (fs != null && fs.length > 0) {
         res.getStoredFields().put(fieldNames[i], fs);
       }
     }
   }
   // collect values from unstored fields
   HashSet<String> fields = new HashSet<String>(Arrays.asList(fieldNames));
   // try to use term vectors if available
   progress.maxValue = fieldNames.length;
   progress.curValue = 0;
   progress.minValue = 0;
   for (int i = 0; i < fieldNames.length; i++) {
     TermFreqVector tvf = reader.getTermFreqVector(docNum, fieldNames[i]);
     if (tvf != null && tvf.size() > 0 && (tvf instanceof TermPositionVector)) {
       TermPositionVector tpv = (TermPositionVector) tvf;
       progress.message = "Reading term vectors ...";
       progress.curValue = i;
       setChanged();
       notifyObservers(progress);
       BytesRef[] tv = tpv.getTerms();
       for (int k = 0; k < tv.length; k++) {
         // do we have positions?
         int[] posArr = tpv.getTermPositions(k);
         if (posArr == null) {
           // only offsets
           TermVectorOffsetInfo[] offsets = tpv.getOffsets(k);
           if (offsets.length == 0) {
             continue;
           }
           // convert offsets into positions
           posArr = convertOffsets(offsets);
         }
         GrowableStringArray gsa = res.getReconstructedFields().get(fieldNames[i]);
         if (gsa == null) {
           gsa = new GrowableStringArray();
           res.getReconstructedFields().put(fieldNames[i], gsa);
         }
         for (int m = 0; m < posArr.length; m++) {
           gsa.append(posArr[m], "|", tv[k].utf8ToString());
         }
       }
       fields.remove(fieldNames[i]); // got what we wanted
     }
   }
   // this loop collects data only from left-over fields
   // not yet collected through term vectors
   progress.maxValue = fields.size();
   progress.curValue = 0;
   progress.minValue = 0;
   for (String fld : fields) {
     progress.message = "Collecting terms in " + fld + " ...";
     progress.curValue++;
     setChanged();
     notifyObservers(progress);
     Terms terms = MultiFields.getTerms(reader, fld);
     if (terms == null) { // no terms in this field
       continue;
     }
     TermsEnum te = terms.iterator();
     while (te.next() != null) {
       DocsAndPositionsEnum dpe = te.docsAndPositions(deleted, null);
       if (dpe == null) { // no position info for this field
         break;
       }
       int num = dpe.advance(docNum);
       if (num != docNum) { // either greater than or NO_MORE_DOCS
         continue; // no data for this term in this doc
       }
       String term = te.term().utf8ToString();
       GrowableStringArray gsa = (GrowableStringArray) res.getReconstructedFields().get(fld);
       if (gsa == null) {
         gsa = new GrowableStringArray();
         res.getReconstructedFields().put(fld, gsa);
       }
       for (int k = 0; k < dpe.freq(); k++) {
         int pos = dpe.nextPosition();
         gsa.append(pos, "|", term);
       }
     }
   }
   progress.message = "Done.";
   progress.curValue = 100;
   setChanged();
   notifyObservers(progress);
   return res;
 }
  /**
   * Creates a set of reference objects and stores it in a new index (name "<indexPath>-ro"). Then
   * creates ordered lists of reference object positions for each data item in the index with given
   * feature. Finally a new index (name "<indexPath>-ms") is created where all the original
   * documents as well as the new data are stored.
   *
   * @param indexPath the path to the original index
   * @throws IOException
   */
  public void createIndex(String indexPath) throws IOException {
    IndexReader reader = DirectoryReader.open(FSDirectory.open(new File(indexPath)));
    int numDocs = reader.numDocs();

    if (numDocs < numReferenceObjects) {
      throw new UnsupportedOperationException("Too few documents in index.");
    }

    // progress report
    progress.setNumDocsAll(numDocs);
    progress.setCurrentState(State.RoSelection);

    boolean hasDeletions = reader.hasDeletions();

    // init reference objects:
    IndexWriter iw = LuceneUtils.createIndexWriter(indexPath + "-ro", true);
    HashSet<Integer> referenceObjsIds = new HashSet<Integer>(numReferenceObjects);

    double numDocsDouble = (double) numDocs;
    while (referenceObjsIds.size() < numReferenceObjects) {
      referenceObjsIds.add((int) (numDocsDouble * Math.random()));
    }
    int count = 0;

    if (hasDeletions) {
      System.err.println(
          "WARNING: There are deleted docs in your index. You should "
              + "optimize your index before using this method.");
    }

    // progress report
    progress.setCurrentState(State.RoIndexing);

    // find them in the index and put them into a separate index:
    for (int i : referenceObjsIds) {
      count++;
      Document document = reader.document(i);
      document.add(new Field("ro-id", count + "", StringField.TYPE_STORED));
      iw.addDocument(document);
    }
    iw.commit();
    iw.close();

    // progress report
    progress.setCurrentState(State.Indexing);

    // now find the reference objects for each entry ;)
    IndexReader readerRo = DirectoryReader.open(FSDirectory.open(new File(indexPath + "-ro")));
    ImageSearcher searcher =
        new GenericImageSearcher(numReferenceObjectsUsed, featureClass, featureFieldName);
    Map<String, Analyzer> analyzerPerField = new HashMap<String, Analyzer>();
    analyzerPerField.put("ro-order", new WhitespaceAnalyzer(LuceneUtils.LUCENE_VERSION));
    PerFieldAnalyzerWrapper aWrapper =
        new PerFieldAnalyzerWrapper(
            new SimpleAnalyzer(LuceneUtils.LUCENE_VERSION), analyzerPerField);

    iw =
        new IndexWriter(
            FSDirectory.open(new File(indexPath)),
            new IndexWriterConfig(LuceneUtils.LUCENE_VERSION, aWrapper)
                .setOpenMode(IndexWriterConfig.OpenMode.CREATE));
    StringBuilder sb = new StringBuilder(256);
    // Needed for check whether the document is deleted.
    Bits liveDocs = MultiFields.getLiveDocs(reader);

    for (int i = 0; i < numDocs; i++) {
      if (reader.hasDeletions() && !liveDocs.get(i)) continue; // if it is deleted, just ignore it.
      Document document = reader.document(i);
      ImageSearchHits hits = searcher.search(document, readerRo);
      sb.delete(0, sb.length());
      for (int j = 0; j < numReferenceObjectsUsed; j++) {
        sb.append(hits.doc(j).getValues("ro-id")[0]);
        sb.append(' ');
      }
      // System.out.println(sb.toString());
      document.add(new TextField("ro-order", sb.toString(), Field.Store.YES));
      iw.updateDocument(
          new Term(
              DocumentBuilder.FIELD_NAME_IDENTIFIER,
              document.getValues(DocumentBuilder.FIELD_NAME_IDENTIFIER)[0]),
          document);

      // progress report
      progress.setNumDocsProcessed(progress.getNumDocsProcessed() + 1);
    }
    iw.commit();
    iw.close();

    // progress report
    progress.setCurrentState(State.Idle);
  }