/**
   * Find words for a more-like-this query former.
   *
   * @param docNum the id of the lucene document from which to find terms
   */
  private PriorityQueue<ScoreTerm> retrieveTerms(int docNum) throws IOException {
    Map<String, Int> termFreqMap = new HashMap<>();
    for (String fieldName : fieldNames) {
      final Fields vectors = ir.getTermVectors(docNum);
      final Terms vector;
      if (vectors != null) {
        vector = vectors.terms(fieldName);
      } else {
        vector = null;
      }

      // field does not store term vector info
      if (vector == null) {
        Document d = ir.document(docNum);
        IndexableField fields[] = d.getFields(fieldName);
        for (IndexableField field : fields) {
          final String stringValue = field.stringValue();
          if (stringValue != null) {
            addTermFrequencies(new FastStringReader(stringValue), termFreqMap, fieldName);
          }
        }
      } else {
        addTermFrequencies(termFreqMap, vector, fieldName);
      }
    }

    return createQueue(termFreqMap);
  }
  /**
   * Extracts termClaimsDescriptionAbstractTitles of the documents; Adds them to vector in the same
   * order
   *
   * @param hits
   * @param i
   * @param j
   * @return relevantDocsTerms docs must be in order
   * @throws java.io.IOException
   */
  public Map<TermFreqVector, String> getDocsTerms(TopDocs hits, int i, int j) throws IOException {
    Map<TermFreqVector, String> docsTerms = new HashMap<>();
    // Process each of the documents
    while (i < j && i < hits.totalHits && i >= 0) {
      ScoreDoc scoreDoc = hits.scoreDocs[i];
      if (sourceField.equals(PatentQuery.all)) {
        Terms termTitle = ir.getTermVector(scoreDoc.doc, PatentQuery.getFields()[1]);
        TermFreqVector docTermsTitle = new TermFreqVector(termTitle);
        docsTerms.put(docTermsTitle, PatentQuery.getFields()[1]);

        Terms termAbstract = ir.getTermVector(scoreDoc.doc, PatentQuery.getFields()[2]);
        TermFreqVector docTermsAbstract = new TermFreqVector(termAbstract);
        docsTerms.put(docTermsAbstract, PatentQuery.getFields()[2]);

        Terms termDescription = ir.getTermVector(scoreDoc.doc, PatentQuery.getFields()[3]);
        TermFreqVector docTermsDescription = new TermFreqVector(termDescription);
        docsTerms.put(docTermsDescription, PatentQuery.getFields()[3]);

        Terms termClaims = ir.getTermVector(scoreDoc.doc, PatentQuery.getFields()[5]);
        TermFreqVector docTermsClaims = new TermFreqVector(termClaims);
        docsTerms.put(docTermsClaims, PatentQuery.getFields()[5]);

      } else {
        Terms term = ir.getTermVector(scoreDoc.doc, sourceField); // get termvector for document
        // Create termVector and add it to vector
        TermFreqVector docTerms = new TermFreqVector(term);
        docsTerms.put(docTerms, sourceField);
      }
      i++;
    }
    return docsTerms;
  }
  /**
   * Sets boost of termClaimsDescriptionAbstractTitles. boost = weight = factor(tf*idf)
   *
   * @param vecsTerms
   * @param currentField
   * @param factor - adjustment factor ( ex. alpha or beta )
   * @param decayFactor
   * @return
   * @throws java.io.IOException
   */
  public Map<String, TermQuery> setBoost(
      Map<TermFreqVector, String> vecsTerms, String currentField, float factor, float decayFactor)
      throws IOException {
    Map<String, TermQuery> terms = new HashMap<>();
    // setBoost for each of the terms of each of the docs
    int i = 0;
    float norm = (float) 1 / vecsTerms.size();
    //        System.out.println("--------------------------");
    for (Map.Entry<TermFreqVector, String> e : vecsTerms.entrySet()) {
      // Increase decay
      String field = e.getValue();
      TermFreqVector docTerms = e.getKey();
      float decay = decayFactor * i;
      // Populate terms: with TermQuries and set boost
      for (String termTxt : docTerms.getTerms()) {
        // Create Term
        Term term = new Term(currentField, termTxt);
        // Calculate weight
        float tf = docTerms.getFreq(termTxt);
        //                float idf = ir.docFreq(termTitle);
        int docs;
        float idf;
        if (sourceField.equals(PatentQuery.all)) {
          docs = ir.getDocCount(field);
          idf = (float) Math.log10((double) docs / (ir.docFreq(new Term(field, termTxt)) + 1));
        } else {
          docs = ir.getDocCount(sourceField);
          idf =
              (float) Math.log10((double) docs / (ir.docFreq(new Term(sourceField, termTxt)) + 1));
        }
        float weight = tf * idf;

        //                System.out.println(term.text() + " -> tf= " + tf + " idf= " + idf + "
        // tfidf= " + weight);
        // Adjust weight by decay factor
        weight = weight - (weight * decay);
        // Create TermQuery and add it to the collection
        TermQuery termQuery = new TermQuery(term);
        // Calculate and set boost
        float boost;
        if (vecsTerms.size() == 1) {
          boost = factor * tf;
        } else {
          boost = factor;
        }

        if (boost != 0) {
          termQuery.setBoost(boost * norm);
          if (terms.containsKey(termTxt)) {
            TermQuery tq = terms.get(termTxt);
            tq.setBoost(tq.getBoost() + termQuery.getBoost());
          } else {
            terms.put(termTxt, termQuery);
          }
        }
      }
      i++;
    }
    return terms;
  }
  /*
   *  listTermVectors displays the term vectors for all of the fields
   *  in a document in an index (specified by reader).
   */
  static void listTermVectors(IndexReader reader, String docidString) throws IOException {

    System.out.println("\nTermVector:  docid " + docidString);

    int docid = Integer.parseInt(docidString);

    if ((docid < 0) || (docid >= reader.numDocs())) {
      System.out.println("ERROR:  " + docidString + " is a bad document id.");
      return;
    }
    ;

    /*
     *  Iterate over the fields in this document.
     */
    Fields fields = reader.getTermVectors(docid);
    Iterator<String> fieldIterator = fields.iterator();

    while (fieldIterator.hasNext()) {
      String fieldName = fieldIterator.next();
      System.out.println("  Field: " + fieldName);

      Terms terms = fields.terms(fieldName);
      termVectorDisplay(terms);
    }
    ;
  }
  private void doTest(Random random, PrintWriter out, boolean useCompoundFiles, int MAX_DOCS)
      throws Exception {
    Directory directory = newDirectory();
    Analyzer analyzer = new MockAnalyzer(random);
    IndexWriterConfig conf = newIndexWriterConfig(analyzer);
    final MergePolicy mp = conf.getMergePolicy();
    mp.setNoCFSRatio(useCompoundFiles ? 1.0 : 0.0);
    IndexWriter writer = new IndexWriter(directory, conf);
    if (VERBOSE) {
      System.out.println("TEST: now build index MAX_DOCS=" + MAX_DOCS);
    }

    for (int j = 0; j < MAX_DOCS; j++) {
      Document d = new Document();
      d.add(newTextField(PRIORITY_FIELD, HIGH_PRIORITY, Field.Store.YES));
      d.add(newTextField(ID_FIELD, Integer.toString(j), Field.Store.YES));
      writer.addDocument(d);
    }
    writer.close();

    // try a search without OR
    IndexReader reader = DirectoryReader.open(directory);
    IndexSearcher searcher = newSearcher(reader);

    Query query = new TermQuery(new Term(PRIORITY_FIELD, HIGH_PRIORITY));
    out.println("Query: " + query.toString(PRIORITY_FIELD));
    if (VERBOSE) {
      System.out.println("TEST: search query=" + query);
    }

    final Sort sort = new Sort(SortField.FIELD_SCORE, new SortField(ID_FIELD, SortField.Type.INT));

    ScoreDoc[] hits = searcher.search(query, null, MAX_DOCS, sort).scoreDocs;
    printHits(out, hits, searcher);
    checkHits(hits, MAX_DOCS, searcher);

    // try a new search with OR
    searcher = newSearcher(reader);
    hits = null;

    BooleanQuery booleanQuery = new BooleanQuery();
    booleanQuery.add(
        new TermQuery(new Term(PRIORITY_FIELD, HIGH_PRIORITY)), BooleanClause.Occur.SHOULD);
    booleanQuery.add(
        new TermQuery(new Term(PRIORITY_FIELD, MED_PRIORITY)), BooleanClause.Occur.SHOULD);
    out.println("Query: " + booleanQuery.toString(PRIORITY_FIELD));

    hits = searcher.search(booleanQuery, null, MAX_DOCS, sort).scoreDocs;
    printHits(out, hits, searcher);
    checkHits(hits, MAX_DOCS, searcher);

    reader.close();
    directory.close();
  }
  /*
   *  listPostings displays the first n postings for a term in a
   *  field in an index (specified by reader).  Set n to MAX_VALUE
   *  to display all postings.
   */
  static void listPostings(IndexReader reader, String termString, String field, Integer n)
      throws IOException {

    System.out.println("\nPostings:  " + termString + " " + field);

    /*
     *  Prepare to access the index.
     */
    BytesRef termBytes = new BytesRef(termString);
    Term term = new Term(field, termBytes);
    Bits liveDocs = MultiFields.getLiveDocs(reader);

    /*
     *  Lookup the collection term frequency (ctf).
     */
    long df = reader.docFreq(term);
    System.out.println("\tdf:  " + df);

    long ctf = reader.totalTermFreq(term);
    System.out.println("\tctf:  " + ctf);

    if (df < 1) return;

    /*
     *  Lookup the inverted list.
     */
    DocsAndPositionsEnum postings =
        MultiFields.getTermPositionsEnum(reader, liveDocs, field, termBytes);

    /*
     *  Iterate through the first n postings.
     */
    long count = 0;

    while ((count < n) && (postings.nextDoc() != DocIdSetIterator.NO_MORE_DOCS)) {

      System.out.println("\tdocid: " + postings.docID());
      int tf = postings.freq();
      System.out.println("\ttf: " + tf);
      System.out.print("\tPositions: ");

      for (int j = 0; j < tf; j++) {
        int pos = postings.nextPosition();
        System.out.print(pos + " ");
      }

      System.out.println("");

      count++;
    }
    ;

    return;
  }
  @Test
  public void testSimple() throws Exception {
    Directory directory = newDirectory();
    RandomIndexWriter indexWriter = new RandomIndexWriter(random(), directory);

    for (int parent = 1; parent <= 5; parent++) {
      Document document = new Document();
      document.add(
          new StringField(
              UidFieldMapper.NAME,
              Uid.createUid("parent", Integer.toString(parent)),
              Field.Store.NO));
      document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
      indexWriter.addDocument(document);

      for (int child = 1; child <= 3; child++) {
        document = new Document();
        document.add(
            new StringField(
                UidFieldMapper.NAME,
                Uid.createUid("child", Integer.toString(parent * 3 + child)),
                Field.Store.NO));
        document.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
        document.add(
            new StringField(
                ParentFieldMapper.NAME,
                Uid.createUid("parent", Integer.toString(parent)),
                Field.Store.NO));
        document.add(new StringField("field1", "value" + child, Field.Store.NO));
        indexWriter.addDocument(document);
      }
    }

    IndexReader indexReader = DirectoryReader.open(indexWriter.w, false);
    IndexSearcher searcher = new IndexSearcher(indexReader);

    TermQuery childQuery = new TermQuery(new Term("field1", "value" + (1 + random().nextInt(3))));
    TermFilter parentFilter = new TermFilter(new Term(TypeFieldMapper.NAME, "parent"));
    int shortCircuitParentDocSet = random().nextInt(5);
    ChildrenConstantScoreQuery query =
        new ChildrenConstantScoreQuery(
            childQuery, "parent", "child", parentFilter, shortCircuitParentDocSet, null);

    BitSetCollector collector = new BitSetCollector(indexReader.maxDoc());
    searcher.search(query, collector);
    FixedBitSet actualResult = collector.getResult();

    assertThat(actualResult.cardinality(), equalTo(5));

    indexWriter.close();
    indexReader.close();
    directory.close();
  }
  /**
   * Create a PriorityQueue from a word->tf map.
   *
   * @param words a map of words keyed on the word(String) with Int objects as the values.
   * @param fieldNames an array of field names to override defaults.
   */
  private PriorityQueue<ScoreTerm> createQueue(Map<String, Int> words, String... fieldNames)
      throws IOException {
    // have collected all words in doc and their freqs
    int numDocs = ir.numDocs();
    final int limit = Math.min(maxQueryTerms, words.size());
    FreqQ queue = new FreqQ(limit); // will order words by score

    for (String word : words.keySet()) { // for every word
      int tf = words.get(word).x; // term freq in the source doc
      if (minTermFreq > 0 && tf < minTermFreq) {
        continue; // filter out words that don't occur enough times in the source
      }

      // go through all the fields and find the largest document frequency
      String topField = fieldNames[0];
      int docFreq = 0;
      for (String fieldName : fieldNames) {
        int freq = ir.docFreq(new Term(fieldName, word));
        topField = (freq > docFreq) ? fieldName : topField;
        docFreq = (freq > docFreq) ? freq : docFreq;
      }

      if (minDocFreq > 0 && docFreq < minDocFreq) {
        continue; // filter out words that don't occur in enough docs
      }

      if (docFreq > maxDocFreq) {
        continue; // filter out words that occur in too many docs
      }

      if (docFreq == 0) {
        continue; // index update problem?
      }

      float idf = similarity.idf(docFreq, numDocs);
      float score = tf * idf;

      if (queue.size() < limit) {
        // there is still space in the queue
        queue.add(new ScoreTerm(word, topField, score, idf, docFreq, tf));
      } else {
        ScoreTerm term = queue.top();
        if (term.score < score) { // update the smallest in the queue in place and update the queue.
          term.update(word, topField, score, idf, docFreq, tf);
          queue.updateTop();
        }
      }
    }
    return queue;
  }
  /*
   *  listTermVectorField displays the term vector for a field in
   *  a document in an index (specified by reader).
   */
  static void listTermVectorField(IndexReader reader, String docidString, String field)
      throws IOException {

    System.out.println("\nTermVector:  docid " + docidString + ", field " + field);

    int docid = Integer.parseInt(docidString);

    if ((docid < 0) || (docid >= reader.numDocs())) {
      System.out.println("ERROR:  " + docidString + " is a bad document id.");
      return;
    }
    ;

    Terms terms = reader.getTermVector(docid, field);
    termVectorDisplay(terms);
  }
Esempio n. 10
0
 /** test that when freqs are omitted, that totalTermFreq and sumTotalTermFreq are -1 */
 public void testStats() throws Exception {
   Directory dir = newDirectory();
   RandomIndexWriter iw =
       new RandomIndexWriter(
           random(), dir, newIndexWriterConfig(TEST_VERSION_CURRENT, new MockAnalyzer(random())));
   Document doc = new Document();
   FieldType ft = new FieldType(TextField.TYPE_NOT_STORED);
   ft.setIndexOptions(IndexOptions.DOCS_ONLY);
   ft.freeze();
   Field f = newField("foo", "bar", ft);
   doc.add(f);
   iw.addDocument(doc);
   IndexReader ir = iw.getReader();
   iw.close();
   assertEquals(-1, ir.totalTermFreq(new Term("foo", new BytesRef("bar"))));
   assertEquals(-1, ir.getSumTotalTermFreq("foo"));
   ir.close();
   dir.close();
 }
Esempio n. 11
0
 public static int docId(IndexReader reader, Term term) throws IOException {
   TermDocs termDocs = reader.termDocs(term);
   try {
     if (termDocs.next()) {
       return termDocs.doc();
     }
     return NO_DOC;
   } finally {
     termDocs.close();
   }
 }
    // public SortingReader(IndexReader oldReader, int[] oldToNew) { // TODO MC
    public SortingReader(IndexReader oldReader, DocScore[] newToOld) { // TODO MC
      super(oldReader);
      this.newToOld = newToOld;

      this.oldToNew = new int[oldReader.maxDoc()];
      int newDoc = 0;
      while (newDoc < newToOld.length) {
        int oldDoc = newToOld[newDoc].doc;
        oldToNew[oldDoc] = newDoc;
        newDoc++;
      }
    }
  public void sort(File directory) throws IOException {
    LOG.info("IndexSorter: starting.");
    Date start = new Date();
    int termIndexInterval = getConf().getInt("indexer.termIndexInterval", 128);
    IndexReader reader = IndexReader.open(new File(directory, "index"));
    Searcher searcher =
        new IndexSearcher(new File(directory, "index").getAbsolutePath()); // TODO MC

    SortingReader sorter = new SortingReader(reader, newToOld(reader, searcher)); // TODO MC
    IndexWriter writer = new IndexWriter(new File(directory, "index-sorted"), null, true);
    writer.setTermIndexInterval(termIndexInterval);
    writer.setUseCompoundFile(false);
    writer.addIndexes(new IndexReader[] {sorter});
    writer.close();
    Date end = new Date();
    LOG.info("IndexSorter: done, " + (end.getTime() - start.getTime()) + " total milliseconds");
  }
  @Test
  public void testRandom() throws Exception {
    Directory directory = newDirectory();
    RandomIndexWriter indexWriter = new RandomIndexWriter(random(), directory);
    int numUniqueChildValues = 1 + random().nextInt(TEST_NIGHTLY ? 10000 : 1000);
    String[] childValues = new String[numUniqueChildValues];
    for (int i = 0; i < numUniqueChildValues; i++) {
      childValues[i] = Integer.toString(i);
    }

    IntOpenHashSet filteredOrDeletedDocs = new IntOpenHashSet();
    int childDocId = 0;
    int numParentDocs = 1 + random().nextInt(TEST_NIGHTLY ? 20000 : 1000);
    ObjectObjectOpenHashMap<String, NavigableSet<String>> childValueToParentIds =
        new ObjectObjectOpenHashMap<String, NavigableSet<String>>();
    for (int parentDocId = 0; parentDocId < numParentDocs; parentDocId++) {
      boolean markParentAsDeleted = rarely();
      boolean filterMe = rarely();
      String parent = Integer.toString(parentDocId);
      Document document = new Document();
      document.add(
          new StringField(UidFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.YES));
      document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
      if (markParentAsDeleted) {
        filteredOrDeletedDocs.add(parentDocId);
        document.add(new StringField("delete", "me", Field.Store.NO));
      }
      if (filterMe) {
        filteredOrDeletedDocs.add(parentDocId);
        document.add(new StringField("filter", "me", Field.Store.NO));
      }
      indexWriter.addDocument(document);

      int numChildDocs;
      if (rarely()) {
        numChildDocs = random().nextInt(TEST_NIGHTLY ? 100 : 25);
      } else {
        numChildDocs = random().nextInt(TEST_NIGHTLY ? 40 : 10);
      }
      for (int i = 0; i < numChildDocs; i++) {
        boolean markChildAsDeleted = rarely();
        String childValue = childValues[random().nextInt(childValues.length)];

        document = new Document();
        document.add(
            new StringField(
                UidFieldMapper.NAME,
                Uid.createUid("child", Integer.toString(childDocId)),
                Field.Store.NO));
        document.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
        document.add(
            new StringField(
                ParentFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.NO));
        document.add(new StringField("field1", childValue, Field.Store.NO));
        if (markChildAsDeleted) {
          document.add(new StringField("delete", "me", Field.Store.NO));
        }
        indexWriter.addDocument(document);

        if (!markChildAsDeleted) {
          NavigableSet<String> parentIds;
          if (childValueToParentIds.containsKey(childValue)) {
            parentIds = childValueToParentIds.lget();
          } else {
            childValueToParentIds.put(childValue, parentIds = new TreeSet<String>());
          }
          if (!markParentAsDeleted && !filterMe) {
            parentIds.add(parent);
          }
        }
      }
    }

    // Delete docs that are marked to be deleted.
    indexWriter.deleteDocuments(new Term("delete", "me"));

    indexWriter.commit();
    IndexReader indexReader = DirectoryReader.open(directory);
    IndexSearcher searcher = new IndexSearcher(indexReader);
    Engine.Searcher engineSearcher =
        new Engine.SimpleSearcher(ChildrenConstantScoreQueryTests.class.getSimpleName(), searcher);
    ((TestSearchContext) SearchContext.current())
        .setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));

    Filter rawParentFilter = new TermFilter(new Term(TypeFieldMapper.NAME, "parent"));
    Filter rawFilterMe = new NotFilter(new TermFilter(new Term("filter", "me")));
    int max = numUniqueChildValues / 4;
    for (int i = 0; i < max; i++) {
      // Randomly pick a cached version: there is specific logic inside ChildrenQuery that deals
      // with the fact
      // that deletes are applied at the top level when filters are cached.
      Filter parentFilter;
      if (random().nextBoolean()) {
        parentFilter = SearchContext.current().filterCache().cache(rawParentFilter);
      } else {
        parentFilter = rawParentFilter;
      }

      // Using this in FQ, will invoke / test the Scorer#advance(..) and also let the Weight#scorer
      // not get live docs as acceptedDocs
      Filter filterMe;
      if (random().nextBoolean()) {
        filterMe = SearchContext.current().filterCache().cache(rawFilterMe);
      } else {
        filterMe = rawFilterMe;
      }

      // Simulate a parent update
      if (random().nextBoolean()) {
        int numberOfUpdates = 1 + random().nextInt(TEST_NIGHTLY ? 25 : 5);
        for (int j = 0; j < numberOfUpdates; j++) {
          int parentId;
          do {
            parentId = random().nextInt(numParentDocs);
          } while (filteredOrDeletedDocs.contains(parentId));

          String parentUid = Uid.createUid("parent", Integer.toString(parentId));
          indexWriter.deleteDocuments(new Term(UidFieldMapper.NAME, parentUid));

          Document document = new Document();
          document.add(new StringField(UidFieldMapper.NAME, parentUid, Field.Store.YES));
          document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
          indexWriter.addDocument(document);
        }

        indexReader.close();
        indexReader = DirectoryReader.open(indexWriter.w, true);
        searcher = new IndexSearcher(indexReader);
        engineSearcher =
            new Engine.SimpleSearcher(
                ChildrenConstantScoreQueryTests.class.getSimpleName(), searcher);
        ((TestSearchContext) SearchContext.current())
            .setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));
      }

      String childValue = childValues[random().nextInt(numUniqueChildValues)];
      TermQuery childQuery = new TermQuery(new Term("field1", childValue));
      int shortCircuitParentDocSet = random().nextInt(numParentDocs);
      Filter nonNestedDocsFilter = random().nextBoolean() ? NonNestedDocsFilter.INSTANCE : null;
      Query query;
      if (random().nextBoolean()) {
        // Usage in HasChildQueryParser
        query =
            new ChildrenConstantScoreQuery(
                childQuery,
                "parent",
                "child",
                parentFilter,
                shortCircuitParentDocSet,
                nonNestedDocsFilter);
      } else {
        // Usage in HasChildFilterParser
        query =
            new XConstantScoreQuery(
                new CustomQueryWrappingFilter(
                    new ChildrenConstantScoreQuery(
                        childQuery,
                        "parent",
                        "child",
                        parentFilter,
                        shortCircuitParentDocSet,
                        nonNestedDocsFilter)));
      }
      query = new XFilteredQuery(query, filterMe);
      BitSetCollector collector = new BitSetCollector(indexReader.maxDoc());
      searcher.search(query, collector);
      FixedBitSet actualResult = collector.getResult();

      FixedBitSet expectedResult = new FixedBitSet(indexReader.maxDoc());
      if (childValueToParentIds.containsKey(childValue)) {
        AtomicReader slowAtomicReader = SlowCompositeReaderWrapper.wrap(indexReader);
        Terms terms = slowAtomicReader.terms(UidFieldMapper.NAME);
        if (terms != null) {
          NavigableSet<String> parentIds = childValueToParentIds.lget();
          TermsEnum termsEnum = terms.iterator(null);
          DocsEnum docsEnum = null;
          for (String id : parentIds) {
            TermsEnum.SeekStatus seekStatus =
                termsEnum.seekCeil(Uid.createUidAsBytes("parent", id));
            if (seekStatus == TermsEnum.SeekStatus.FOUND) {
              docsEnum =
                  termsEnum.docs(slowAtomicReader.getLiveDocs(), docsEnum, DocsEnum.FLAG_NONE);
              expectedResult.set(docsEnum.nextDoc());
            } else if (seekStatus == TermsEnum.SeekStatus.END) {
              break;
            }
          }
        }
      }

      assertBitSet(actualResult, expectedResult, searcher);
    }

    indexWriter.close();
    indexReader.close();
    directory.close();
  }
  /**
   * Assert that the {@code scoreType} operates as expected and parents are found in the expected
   * order.
   *
   * <p>This will use the test index's parent/child types to create parents with multiple children.
   * Each child will have a randomly generated scored stored in {@link #CHILD_SCORE_NAME}, which is
   * used to score based on the {@code scoreType} by using a {@link MockScorer} to determine the
   * expected scores.
   *
   * @param scoreType The score type to use within the query to score parents relative to their
   *     children.
   * @throws IOException if any unexpected error occurs
   */
  private void assertScoreType(ScoreType scoreType) throws IOException {
    SearchContext context = SearchContext.current();
    Directory directory = newDirectory();
    IndexWriter writer =
        new IndexWriter(directory, newIndexWriterConfig(new MockAnalyzer(random())));

    // calculates the expected score per parent
    MockScorer scorer = new MockScorer(scoreType);
    scorer.scores = new FloatArrayList(10);

    // number of parents to generate
    int parentDocs = scaledRandomIntBetween(2, 10);
    // unique child ID
    int childDocId = 0;

    // Parent ID to expected score
    Map<String, Float> parentScores = new TreeMap<>();

    // Add a few random parents to ensure that the children's score is appropriately taken into
    // account
    for (int parentDocId = 0; parentDocId < parentDocs; ++parentDocId) {
      String parent = Integer.toString(parentDocId);

      // Create the parent
      Document parentDocument = new Document();

      parentDocument.add(
          new StringField(UidFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.YES));
      parentDocument.add(new StringField(IdFieldMapper.NAME, parent, Field.Store.YES));
      parentDocument.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));

      // add the parent to the index
      writer.addDocument(parentDocument);

      int numChildDocs = scaledRandomIntBetween(1, 10);

      // forget any parent's previous scores
      scorer.scores.clear();

      // associate children with the parent
      for (int i = 0; i < numChildDocs; ++i) {
        int childScore = random().nextInt(128);

        Document childDocument = new Document();

        childDocument.add(
            new StringField(
                UidFieldMapper.NAME,
                Uid.createUid("child", Integer.toString(childDocId++)),
                Field.Store.NO));
        childDocument.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
        // parent association:
        childDocument.add(
            new StringField(
                ParentFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.NO));
        childDocument.add(new DoubleField(CHILD_SCORE_NAME, childScore, Field.Store.NO));

        // remember the score to be calculated
        scorer.scores.add(childScore);

        // add the associated child to the index
        writer.addDocument(childDocument);
      }

      // this score that should be returned for this parent
      parentScores.put(parent, scorer.score());
    }

    writer.commit();

    IndexReader reader = DirectoryReader.open(writer, true);
    IndexSearcher searcher = new IndexSearcher(reader);

    // setup to read the parent/child map
    Engine.Searcher engineSearcher =
        new Engine.Searcher(ChildrenQueryTests.class.getSimpleName(), searcher);
    ((TestSearchContext) context).setSearcher(new ContextIndexSearcher(context, engineSearcher));

    // child query that returns the score as the value of "childScore" for each child document, with
    // the parent's score determined by the score type
    QueryBuilder childQueryBuilder =
        functionScoreQuery(typeFilter("child"))
            .add(new FieldValueFactorFunctionBuilder(CHILD_SCORE_NAME));
    QueryBuilder queryBuilder =
        hasChildQuery("child", childQueryBuilder)
            .scoreType(scoreType.name().toLowerCase(Locale.ENGLISH))
            .setShortCircuitCutoff(parentDocs);

    // Perform the search for the documents using the selected score type
    TopDocs docs = searcher.search(parseQuery(queryBuilder), parentDocs);
    assertThat("Expected all parents", docs.totalHits, is(parentDocs));

    // score should be descending (just a sanity check)
    float topScore = docs.scoreDocs[0].score;

    // ensure each score is returned as expected
    for (int i = 0; i < parentDocs; ++i) {
      ScoreDoc scoreDoc = docs.scoreDocs[i];
      // get the ID from the document to get its expected score; remove it so we cannot double-count
      // it
      float score = parentScores.remove(reader.document(scoreDoc.doc).get(IdFieldMapper.NAME));

      // expect exact match
      assertThat("Unexpected score", scoreDoc.score, is(score));
      assertThat("Not descending", score, lessThanOrEqualTo(topScore));

      // it had better keep descending
      topScore = score;
    }

    reader.close();
    writer.close();
    directory.close();
  }
  @Test
  public void testRandom() throws Exception {
    Directory directory = newDirectory();
    final Random r = random();
    final IndexWriterConfig iwc =
        LuceneTestCase.newIndexWriterConfig(r, new MockAnalyzer(r))
            .setMaxBufferedDocs(IndexWriterConfig.DISABLE_AUTO_FLUSH)
            .setRAMBufferSizeMB(
                scaledRandomIntBetween(16, 64)); // we might index a lot - don't go crazy here
    RandomIndexWriter indexWriter = new RandomIndexWriter(r, directory, iwc);
    int numUniqueChildValues = scaledRandomIntBetween(100, 2000);
    String[] childValues = new String[numUniqueChildValues];
    for (int i = 0; i < numUniqueChildValues; i++) {
      childValues[i] = Integer.toString(i);
    }

    IntOpenHashSet filteredOrDeletedDocs = new IntOpenHashSet();

    int childDocId = 0;
    int numParentDocs = scaledRandomIntBetween(1, numUniqueChildValues);
    ObjectObjectOpenHashMap<String, NavigableMap<String, FloatArrayList>> childValueToParentIds =
        new ObjectObjectOpenHashMap<>();
    for (int parentDocId = 0; parentDocId < numParentDocs; parentDocId++) {
      boolean markParentAsDeleted = rarely();
      boolean filterMe = rarely();
      String parent = Integer.toString(parentDocId);
      Document document = new Document();
      document.add(
          new StringField(UidFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.YES));
      document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
      if (markParentAsDeleted) {
        filteredOrDeletedDocs.add(parentDocId);
        document.add(new StringField("delete", "me", Field.Store.NO));
      }
      if (filterMe) {
        filteredOrDeletedDocs.add(parentDocId);
        document.add(new StringField("filter", "me", Field.Store.NO));
      }
      indexWriter.addDocument(document);

      int numChildDocs = scaledRandomIntBetween(0, 100);
      for (int i = 0; i < numChildDocs; i++) {
        boolean markChildAsDeleted = rarely();
        String childValue = childValues[random().nextInt(childValues.length)];

        document = new Document();
        document.add(
            new StringField(
                UidFieldMapper.NAME,
                Uid.createUid("child", Integer.toString(childDocId++)),
                Field.Store.NO));
        document.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
        document.add(
            new StringField(
                ParentFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.NO));
        document.add(new StringField("field1", childValue, Field.Store.NO));
        if (markChildAsDeleted) {
          document.add(new StringField("delete", "me", Field.Store.NO));
        }
        indexWriter.addDocument(document);

        if (!markChildAsDeleted) {
          NavigableMap<String, FloatArrayList> parentIdToChildScores;
          if (childValueToParentIds.containsKey(childValue)) {
            parentIdToChildScores = childValueToParentIds.lget();
          } else {
            childValueToParentIds.put(childValue, parentIdToChildScores = new TreeMap<>());
          }
          if (!markParentAsDeleted && !filterMe) {
            FloatArrayList childScores = parentIdToChildScores.get(parent);
            if (childScores == null) {
              parentIdToChildScores.put(parent, childScores = new FloatArrayList());
            }
            childScores.add(1f);
          }
        }
      }
    }

    // Delete docs that are marked to be deleted.
    indexWriter.deleteDocuments(new Term("delete", "me"));
    indexWriter.commit();

    IndexReader indexReader = DirectoryReader.open(directory);
    IndexSearcher searcher = new IndexSearcher(indexReader);
    Engine.Searcher engineSearcher =
        new Engine.Searcher(ChildrenQueryTests.class.getSimpleName(), searcher);
    ((TestSearchContext) SearchContext.current())
        .setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));

    int max = numUniqueChildValues / 4;
    for (int i = 0; i < max; i++) {
      // Simulate a parent update
      if (random().nextBoolean()) {
        final int numberOfUpdatableParents = numParentDocs - filteredOrDeletedDocs.size();
        int numberOfUpdates =
            RandomInts.randomIntBetween(
                random(), 0, Math.min(numberOfUpdatableParents, TEST_NIGHTLY ? 25 : 5));
        for (int j = 0; j < numberOfUpdates; j++) {
          int parentId;
          do {
            parentId = random().nextInt(numParentDocs);
          } while (filteredOrDeletedDocs.contains(parentId));

          String parentUid = Uid.createUid("parent", Integer.toString(parentId));
          indexWriter.deleteDocuments(new Term(UidFieldMapper.NAME, parentUid));

          Document document = new Document();
          document.add(new StringField(UidFieldMapper.NAME, parentUid, Field.Store.YES));
          document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
          indexWriter.addDocument(document);
        }

        indexReader.close();
        indexReader = DirectoryReader.open(indexWriter.w, true);
        searcher = new IndexSearcher(indexReader);
        engineSearcher =
            new Engine.Searcher(ChildrenConstantScoreQueryTests.class.getSimpleName(), searcher);
        ((TestSearchContext) SearchContext.current())
            .setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));
      }

      String childValue = childValues[random().nextInt(numUniqueChildValues)];
      int shortCircuitParentDocSet = random().nextInt(numParentDocs);
      ScoreType scoreType = ScoreType.values()[random().nextInt(ScoreType.values().length)];
      // leave min/max set to 0 half the time
      int minChildren = random().nextInt(2) * scaledRandomIntBetween(0, 110);
      int maxChildren = random().nextInt(2) * scaledRandomIntBetween(minChildren, 110);

      QueryBuilder queryBuilder =
          hasChildQuery("child", constantScoreQuery(termQuery("field1", childValue)))
              .scoreType(scoreType.name().toLowerCase(Locale.ENGLISH))
              .minChildren(minChildren)
              .maxChildren(maxChildren)
              .setShortCircuitCutoff(shortCircuitParentDocSet);
      // Using a FQ, will invoke / test the Scorer#advance(..) and also let the Weight#scorer not
      // get live docs as acceptedDocs
      queryBuilder = filteredQuery(queryBuilder, notFilter(termFilter("filter", "me")));
      Query query = parseQuery(queryBuilder);
      BitSetCollector collector = new BitSetCollector(indexReader.maxDoc());
      int numHits = 1 + random().nextInt(25);
      TopScoreDocCollector actualTopDocsCollector = TopScoreDocCollector.create(numHits);
      searcher.search(query, MultiCollector.wrap(collector, actualTopDocsCollector));
      FixedBitSet actualResult = collector.getResult();

      FixedBitSet expectedResult = new FixedBitSet(indexReader.maxDoc());
      TopScoreDocCollector expectedTopDocsCollector = TopScoreDocCollector.create(numHits);
      if (childValueToParentIds.containsKey(childValue)) {
        LeafReader slowLeafReader = SlowCompositeReaderWrapper.wrap(indexReader);
        final FloatArrayList[] scores = new FloatArrayList[slowLeafReader.maxDoc()];
        Terms terms = slowLeafReader.terms(UidFieldMapper.NAME);
        if (terms != null) {
          NavigableMap<String, FloatArrayList> parentIdToChildScores = childValueToParentIds.lget();
          TermsEnum termsEnum = terms.iterator(null);
          DocsEnum docsEnum = null;
          for (Map.Entry<String, FloatArrayList> entry : parentIdToChildScores.entrySet()) {
            int count = entry.getValue().elementsCount;
            if (count >= minChildren && (maxChildren == 0 || count <= maxChildren)) {
              TermsEnum.SeekStatus seekStatus =
                  termsEnum.seekCeil(Uid.createUidAsBytes("parent", entry.getKey()));
              if (seekStatus == TermsEnum.SeekStatus.FOUND) {
                docsEnum =
                    termsEnum.docs(slowLeafReader.getLiveDocs(), docsEnum, DocsEnum.FLAG_NONE);
                expectedResult.set(docsEnum.nextDoc());
                scores[docsEnum.docID()] = new FloatArrayList(entry.getValue());
              } else if (seekStatus == TermsEnum.SeekStatus.END) {
                break;
              }
            }
          }
        }
        MockScorer mockScorer = new MockScorer(scoreType);
        final LeafCollector leafCollector =
            expectedTopDocsCollector.getLeafCollector(slowLeafReader.getContext());
        leafCollector.setScorer(mockScorer);
        for (int doc = expectedResult.nextSetBit(0);
            doc < slowLeafReader.maxDoc();
            doc =
                doc + 1 >= expectedResult.length()
                    ? DocIdSetIterator.NO_MORE_DOCS
                    : expectedResult.nextSetBit(doc + 1)) {
          mockScorer.scores = scores[doc];
          leafCollector.collect(doc);
        }
      }

      assertBitSet(actualResult, expectedResult, searcher);
      assertTopDocs(actualTopDocsCollector.topDocs(), expectedTopDocsCollector.topDocs());
    }

    indexWriter.close();
    indexReader.close();
    directory.close();
  }
Esempio n. 17
0
  // Test scores with one field with Term Freqs and one without, otherwise with equal content
  public void testBasic() throws Exception {
    Directory dir = newDirectory();
    Analyzer analyzer = new MockAnalyzer(random());
    IndexWriter writer =
        new IndexWriter(
            dir,
            newIndexWriterConfig(TEST_VERSION_CURRENT, analyzer)
                .setMaxBufferedDocs(2)
                .setSimilarity(new SimpleSimilarity())
                .setMergePolicy(newLogMergePolicy(2)));

    StringBuilder sb = new StringBuilder(265);
    String term = "term";
    for (int i = 0; i < 30; i++) {
      Document d = new Document();
      sb.append(term).append(" ");
      String content = sb.toString();
      Field noTf = newField("noTf", content + (i % 2 == 0 ? "" : " notf"), omitType);
      d.add(noTf);

      Field tf = newField("tf", content + (i % 2 == 0 ? " tf" : ""), normalType);
      d.add(tf);

      writer.addDocument(d);
      // System.out.println(d);
    }

    writer.forceMerge(1);
    // flush
    writer.close();

    /*
     * Verify the index
     */
    IndexReader reader = DirectoryReader.open(dir);
    IndexSearcher searcher = newSearcher(reader);
    searcher.setSimilarity(new SimpleSimilarity());

    Term a = new Term("noTf", term);
    Term b = new Term("tf", term);
    Term c = new Term("noTf", "notf");
    Term d = new Term("tf", "tf");
    TermQuery q1 = new TermQuery(a);
    TermQuery q2 = new TermQuery(b);
    TermQuery q3 = new TermQuery(c);
    TermQuery q4 = new TermQuery(d);

    PhraseQuery pq = new PhraseQuery();
    pq.add(a);
    pq.add(c);
    try {
      searcher.search(pq, 10);
      fail("did not hit expected exception");
    } catch (Exception e) {
      Throwable cause = e;
      // If the searcher uses an executor service, the IAE is wrapped into other exceptions
      while (cause.getCause() != null) {
        cause = cause.getCause();
      }
      assertTrue("Expected an IAE, got " + cause, cause instanceof IllegalStateException);
    }

    searcher.search(
        q1,
        new CountingHitCollector() {
          private Scorer scorer;

          @Override
          public final void setScorer(Scorer scorer) {
            this.scorer = scorer;
          }

          @Override
          public final void collect(int doc) throws IOException {
            // System.out.println("Q1: Doc=" + doc + " score=" + score);
            float score = scorer.score();
            assertTrue("got score=" + score, score == 1.0f);
            super.collect(doc);
          }
        });
    // System.out.println(CountingHitCollector.getCount());

    searcher.search(
        q2,
        new CountingHitCollector() {
          private Scorer scorer;

          @Override
          public final void setScorer(Scorer scorer) {
            this.scorer = scorer;
          }

          @Override
          public final void collect(int doc) throws IOException {
            // System.out.println("Q2: Doc=" + doc + " score=" + score);
            float score = scorer.score();
            assertEquals(1.0f + doc, score, 0.00001f);
            super.collect(doc);
          }
        });
    // System.out.println(CountingHitCollector.getCount());

    searcher.search(
        q3,
        new CountingHitCollector() {
          private Scorer scorer;

          @Override
          public final void setScorer(Scorer scorer) {
            this.scorer = scorer;
          }

          @Override
          public final void collect(int doc) throws IOException {
            // System.out.println("Q1: Doc=" + doc + " score=" + score);
            float score = scorer.score();
            assertTrue(score == 1.0f);
            assertFalse(doc % 2 == 0);
            super.collect(doc);
          }
        });
    // System.out.println(CountingHitCollector.getCount());

    searcher.search(
        q4,
        new CountingHitCollector() {
          private Scorer scorer;

          @Override
          public final void setScorer(Scorer scorer) {
            this.scorer = scorer;
          }

          @Override
          public final void collect(int doc) throws IOException {
            float score = scorer.score();
            // System.out.println("Q1: Doc=" + doc + " score=" + score);
            assertTrue(score == 1.0f);
            assertTrue(doc % 2 == 0);
            super.collect(doc);
          }
        });
    // System.out.println(CountingHitCollector.getCount());

    BooleanQuery bq = new BooleanQuery();
    bq.add(q1, Occur.MUST);
    bq.add(q4, Occur.MUST);

    searcher.search(
        bq,
        new CountingHitCollector() {
          @Override
          public final void collect(int doc) throws IOException {
            // System.out.println("BQ: Doc=" + doc + " score=" + score);
            super.collect(doc);
          }
        });
    assertEquals(15, CountingHitCollector.getCount());

    reader.close();
    dir.close();
  }
Esempio n. 18
0
 /**
  * Set the maximum percentage in which words may still appear. Words that appear in more than this
  * many percent of all docs will be ignored.
  *
  * @param maxPercentage the maximum percentage of documents (0-100) that a term may appear in to
  *     be still considered relevant
  */
 public void setMaxDocFreqPct(int maxPercentage) {
   this.maxDocFreq = maxPercentage * ir.numDocs() / 100;
 }
  public static void main(String[] args) throws IOException {

    IndexReader reader = null;

    /*
     *  Opening the index first simplifies the processing of the
     *  rest of the command line arguments.
     */
    for (int i = 0; i < args.length; i++) {
      if (("-index".equals(args[i])) && ((i + 1) < args.length)) {
        reader = DirectoryReader.open(FSDirectory.open(new File(args[i + 1])));

        if (reader == null) {
          System.err.println("Error:  Can't open index " + args[i + 1]);
          System.exit(1);
        }
        ;

        break;
      }
      ;
    }
    ;

    if (reader == null) {
      System.err.println(usage);
      System.exit(1);
    }
    ;

    /*
     *  Process the command line arguments sequentially.
     */
    for (int i = 0; i < args.length; i++) {

      if ("-index".equals(args[i])) {

        /*
         *  Handled in the previous loop, so just skip the argument.
         */
        i++;

      } else if ("-list-edocid".equals(args[i])) {

        System.out.println("-list-edocid:");

        if ((i + 1) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        Document d = reader.document(Integer.parseInt(args[i + 1]));

        System.out.println(
            "Internal docid --> External docid: " + args[i + 1] + " --> " + d.get("externalId"));

        i += 1;
      } else if ("-list-docids".equals(args[i])) {

        System.out.println("-list-docids:");

        for (int j = 0; j < reader.numDocs(); j++) {
          Document d = reader.document(j);
          System.out.println("Internal --> external docid: " + j + " --> " + d.get("externalId"));
        }
        ;

      } else if ("-list-fields".equals(args[i])) {

        Fields fields = MultiFields.getFields(reader);

        System.out.print("\nNumber of fields:  ");

        if (fields == null) System.out.println("0");
        else {
          System.out.println(fields.size());

          Iterator<String> is = fields.iterator();

          while (is.hasNext()) {
            System.out.println("\t" + is.next());
          }
          ;
        }
        ;

      } else if ("-list-postings".equals(args[i])) {

        if ((i + 2) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listPostings(reader, args[i + 1], args[i + 2], Integer.MAX_VALUE);
        i += 2;

      } else if ("-list-postings-sample".equals(args[i])) {

        if ((i + 2) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listPostings(reader, args[i + 1], args[i + 2], 5);
        i += 2;

      } else if ("-list-stats".equals(args[i])) {

        System.out.println("Corpus statistics:");
        System.out.println("\tnumdocs\t\t" + reader.numDocs());
        System.out.println(
            "\turl:\t"
                + "\tnumdocs="
                + reader.getDocCount("url")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("url")
                + "\tavglen="
                + reader.getSumTotalTermFreq("url") / (float) reader.getDocCount("url"));

        System.out.println(
            "\tkeywords:"
                + "\tnumdocs="
                + reader.getDocCount("keywords")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("keywords")
                + "\tavglen="
                + reader.getSumTotalTermFreq("keywords") / (float) reader.getDocCount("keywords"));

        System.out.println(
            "\ttitle:\t"
                + "\tnumdocs="
                + reader.getDocCount("title")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("title")
                + "\tavglen="
                + reader.getSumTotalTermFreq("title") / (float) reader.getDocCount("title"));

        System.out.println(
            "\tbody:\t"
                + "\tnumdocs="
                + reader.getDocCount("body")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("body")
                + "\tavglen="
                + reader.getSumTotalTermFreq("body") / (float) reader.getDocCount("body"));

        System.out.println(
            "\tinlink:\t"
                + "\tnumdocs="
                + reader.getDocCount("inlink")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("inlink")
                + "\tavglen="
                + reader.getSumTotalTermFreq("inlink") / (float) reader.getDocCount("inlink"));

      } else if ("-list-terms".equals(args[i])) {

        if ((i + 1) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listTermDictionary(reader, args[i + 1]);
        i += 1;

      } else if ("-list-termvector".equals(args[i])) {

        if ((i + 1) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listTermVectors(reader, args[i + 1]);
        i += 1;

      } else if ("-list-termvector-field".equals(args[i])) {

        if ((i + 2) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listTermVectorField(reader, args[i + 1], args[i + 2]);
        i += 2;

      } else System.err.println("\nWarning:  Unknown argument " + args[i] + " ignored.");
    }
    ;

    /*
     *  Close the index and exit gracefully.
     */
    reader.close();
  }
  @Test
  public void testRandom() throws Exception {
    Directory directory = newDirectory();
    RandomIndexWriter indexWriter = new RandomIndexWriter(random(), directory);
    int numUniqueChildValues = 1 + random().nextInt(TEST_NIGHTLY ? 6000 : 600);
    String[] childValues = new String[numUniqueChildValues];
    for (int i = 0; i < numUniqueChildValues; i++) {
      childValues[i] = Integer.toString(i);
    }

    int childDocId = 0;
    int numParentDocs = 1 + random().nextInt(TEST_NIGHTLY ? 20000 : 1000);
    ObjectObjectOpenHashMap<String, NavigableMap<String, FloatArrayList>> childValueToParentIds =
        new ObjectObjectOpenHashMap<String, NavigableMap<String, FloatArrayList>>();
    for (int parentDocId = 0; parentDocId < numParentDocs; parentDocId++) {
      boolean markParentAsDeleted = rarely();
      String parent = Integer.toString(parentDocId);
      Document document = new Document();
      document.add(
          new StringField(UidFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.YES));
      document.add(new StringField(TypeFieldMapper.NAME, "parent", Field.Store.NO));
      if (markParentAsDeleted) {
        document.add(new StringField("delete", "me", Field.Store.NO));
      }
      indexWriter.addDocument(document);

      int numChildDocs = random().nextInt(TEST_NIGHTLY ? 100 : 25);
      for (int i = 0; i < numChildDocs; i++) {
        boolean markChildAsDeleted = rarely();
        String childValue = childValues[random().nextInt(childValues.length)];

        document = new Document();
        document.add(
            new StringField(
                UidFieldMapper.NAME,
                Uid.createUid("child", Integer.toString(childDocId)),
                Field.Store.NO));
        document.add(new StringField(TypeFieldMapper.NAME, "child", Field.Store.NO));
        document.add(
            new StringField(
                ParentFieldMapper.NAME, Uid.createUid("parent", parent), Field.Store.NO));
        document.add(new StringField("field1", childValue, Field.Store.NO));
        if (markChildAsDeleted) {
          document.add(new StringField("delete", "me", Field.Store.NO));
        }
        indexWriter.addDocument(document);

        if (!markChildAsDeleted) {
          NavigableMap<String, FloatArrayList> parentIdToChildScores;
          if (childValueToParentIds.containsKey(childValue)) {
            parentIdToChildScores = childValueToParentIds.lget();
          } else {
            childValueToParentIds.put(
                childValue, parentIdToChildScores = new TreeMap<String, FloatArrayList>());
          }
          if (!markParentAsDeleted) {
            FloatArrayList childScores = parentIdToChildScores.get(parent);
            if (childScores == null) {
              parentIdToChildScores.put(parent, childScores = new FloatArrayList());
            }
            childScores.add(1f);
          }
        }
      }
    }

    // Delete docs that are marked to be deleted.
    indexWriter.deleteDocuments(new Term("delete", "me"));

    indexWriter.close();
    IndexReader indexReader = DirectoryReader.open(directory);
    IndexSearcher searcher = new IndexSearcher(indexReader);
    Engine.Searcher engineSearcher =
        new Engine.SimpleSearcher(ChildrenQueryTests.class.getSimpleName(), searcher);
    ((TestSearchContext) SearchContext.current())
        .setSearcher(new ContextIndexSearcher(SearchContext.current(), engineSearcher));

    TermFilter parentFilter = new TermFilter(new Term(TypeFieldMapper.NAME, "parent"));
    for (String childValue : childValues) {
      Query childQuery = new ConstantScoreQuery(new TermQuery(new Term("field1", childValue)));
      int shortCircuitParentDocSet = random().nextInt(numParentDocs);
      ScoreType scoreType = ScoreType.values()[random().nextInt(ScoreType.values().length)];
      Query query =
          new ChildrenQuery(
              "parent", "child", parentFilter, childQuery, scoreType, shortCircuitParentDocSet);
      BitSetCollector collector = new BitSetCollector(indexReader.maxDoc());
      int numHits = 1 + random().nextInt(25);
      TopScoreDocCollector actualTopDocsCollector = TopScoreDocCollector.create(numHits, false);
      searcher.search(query, MultiCollector.wrap(collector, actualTopDocsCollector));
      FixedBitSet actualResult = collector.getResult();

      FixedBitSet expectedResult = new FixedBitSet(indexReader.maxDoc());
      MockScorer mockScorer = new MockScorer(scoreType);
      TopScoreDocCollector expectedTopDocsCollector = TopScoreDocCollector.create(numHits, false);
      expectedTopDocsCollector.setScorer(mockScorer);
      if (childValueToParentIds.containsKey(childValue)) {
        AtomicReader slowAtomicReader = SlowCompositeReaderWrapper.wrap(indexReader);
        Terms terms = slowAtomicReader.terms(UidFieldMapper.NAME);
        if (terms != null) {
          NavigableMap<String, FloatArrayList> parentIdToChildScores = childValueToParentIds.lget();
          TermsEnum termsEnum = terms.iterator(null);
          DocsEnum docsEnum = null;
          for (Map.Entry<String, FloatArrayList> entry : parentIdToChildScores.entrySet()) {
            TermsEnum.SeekStatus seekStatus =
                termsEnum.seekCeil(Uid.createUidAsBytes("parent", entry.getKey()));
            if (seekStatus == TermsEnum.SeekStatus.FOUND) {
              docsEnum =
                  termsEnum.docs(slowAtomicReader.getLiveDocs(), docsEnum, DocsEnum.FLAG_NONE);
              expectedResult.set(docsEnum.nextDoc());
              mockScorer.scores = entry.getValue();
              expectedTopDocsCollector.collect(docsEnum.docID());
            } else if (seekStatus == TermsEnum.SeekStatus.END) {
              break;
            }
          }
        }
      }

      assertBitSet(actualResult, expectedResult, searcher);
      assertTopDocs(actualTopDocsCollector.topDocs(), expectedTopDocsCollector.topDocs());
    }

    indexReader.close();
    directory.close();
  }
Esempio n. 21
0
  /**
   * Searches pages using a particular combination of flags.
   *
   * @param query The query to perform in Lucene query language
   * @param flags A set of flags
   * @return A Collection of SearchResult instances
   * @throws ProviderException if there is a problem with the backend
   */
  public Collection findPages(String query, int flags) throws ProviderException {
    IndexSearcher searcher = null;
    ArrayList<SearchResult> list = null;
    Highlighter highlighter = null;

    try {
      String[] queryfields = {
        LUCENE_PAGE_CONTENTS, LUCENE_PAGE_NAME, LUCENE_AUTHOR, LUCENE_ATTACHMENTS
      };
      QueryParser qp =
          new MultiFieldQueryParser(Version.LUCENE_36, queryfields, getLuceneAnalyzer());

      // QueryParser qp = new QueryParser( LUCENE_PAGE_CONTENTS, getLuceneAnalyzer() );
      Query luceneQuery = qp.parse(query);

      if ((flags & FLAG_CONTEXTS) != 0) {
        highlighter =
            new Highlighter(
                new SimpleHTMLFormatter("<span class=\"searchmatch\">", "</span>"),
                new SimpleHTMLEncoder(),
                new QueryScorer(luceneQuery));
      }

      try {
        File dir = new File(m_luceneDirectory);
        Directory luceneDir = new SimpleFSDirectory(dir, null);
        IndexReader reader = IndexReader.open(luceneDir);
        searcher = new IndexSearcher(reader);
      } catch (Exception ex) {
        log.info("Lucene not yet ready; indexing not started", ex);
        return null;
      }

      ScoreDoc[] hits = searcher.search(luceneQuery, MAX_SEARCH_HITS).scoreDocs;

      list = new ArrayList<SearchResult>(hits.length);
      for (int curr = 0; curr < hits.length; curr++) {
        int docID = hits[curr].doc;
        Document doc = searcher.doc(docID);
        String pageName = doc.get(LUCENE_ID);
        WikiPage page = m_engine.getPage(pageName, WikiPageProvider.LATEST_VERSION);

        if (page != null) {
          if (page instanceof Attachment) {
            // Currently attachments don't look nice on the search-results page
            // When the search-results are cleaned up this can be enabled again.
          }

          int score = (int) (hits[curr].score * 100);

          // Get highlighted search contexts
          String text = doc.get(LUCENE_PAGE_CONTENTS);

          String[] fragments = new String[0];
          if (text != null && highlighter != null) {
            TokenStream tokenStream =
                getLuceneAnalyzer().tokenStream(LUCENE_PAGE_CONTENTS, new StringReader(text));
            fragments = highlighter.getBestFragments(tokenStream, text, MAX_FRAGMENTS);
          }

          SearchResult result = new SearchResultImpl(page, score, fragments);
          list.add(result);
        } else {
          log.error(
              "Lucene found a result page '"
                  + pageName
                  + "' that could not be loaded, removing from Lucene cache");
          pageRemoved(new WikiPage(m_engine, pageName));
        }
      }
    } catch (IOException e) {
      log.error("Failed during lucene search", e);
    } catch (ParseException e) {
      log.info("Broken query; cannot parse query ", e);

      throw new ProviderException(
          "You have entered a query Lucene cannot process: " + e.getMessage());
    } catch (InvalidTokenOffsetsException e) {
      log.error("Tokens are incompatible with provided text ", e);
    } finally {
      if (searcher != null) {
        try {
          searcher.close();
        } catch (IOException e) {
          log.error(e);
        }
      }
    }

    return list;
  }
  // private static int[] oldToNew(IndexReader reader, Searcher searcher) throws IOException {
  private static DocScore[] newToOld(IndexReader reader, Searcher searcher) throws IOException {
    int readerMax = reader.maxDoc();
    DocScore[] newToOld = new DocScore[readerMax];

    // use site, an indexed, un-tokenized field to get boost
    // byte[] boosts = reader.norms("site"); TODO MC
    /* TODO MC */
    Document docMeta;
    Pattern includes = Pattern.compile("\\|");
    String value = NutchConfiguration.create().get(INCLUDE_EXTENSIONS_KEY, "");
    String includeExtensions[] = includes.split(value);
    Hashtable<String, Boolean> validExtensions = new Hashtable<String, Boolean>();
    for (int i = 0; i < includeExtensions.length; i++) {
      validExtensions.put(includeExtensions[i], true);
      System.out.println("extension boosted " + includeExtensions[i]);
    }
    /* TODO MC */

    for (int oldDoc = 0; oldDoc < readerMax; oldDoc++) {
      float score;
      if (reader.isDeleted(oldDoc)) {
        // score = 0.0f;
        score = -1f; // TODO MC
      } else {
        // score = Similarity.decodeNorm(boosts[oldDoc]); TODO MC
        /* TODO MC */
        docMeta = searcher.doc(oldDoc);
        if (validExtensions.get(docMeta.get("subType"))
            == null) { // searched extensions will have higher scores
          score = -0.5f;
        } else {
          score = Integer.parseInt(docMeta.get("inlinks"));
          /*
          if (score==0) {
          	score=0.001f; // TODO MC - to not erase
          }
          */
        }
        /* TODO MC */
        // System.out.println("Score for old document "+oldDoc+" is "+score+" and type
        // "+docMeta.get("subType")); // TODO MC debug remove
      }
      DocScore docScore = new DocScore();
      docScore.doc = oldDoc;
      docScore.score = score;
      newToOld[oldDoc] = docScore;
    }

    System.out.println("Sorting " + newToOld.length + " documents.");
    Arrays.sort(newToOld);
    // HeapSorter.sort(newToOld); // TODO MC - due to the lack of space

    /* TODO MC
    int[] oldToNew = new int[readerMax];
    for (int newDoc = 0; newDoc < readerMax; newDoc++) {
      DocScore docScore = newToOld[newDoc];
      //oldToNew[docScore.oldDoc] = docScore.score > 0.0f ? newDoc : -1; // TODO MC
      oldToNew[docScore.oldDoc] = newDoc; // TODO MC
    }
    */

    /* TODO MC *
    for (int newDoc = 0; newDoc < readerMax; newDoc++) {
    	DocScore docScore = newToOld[newDoc];
    	System.out.println("Score for new document "+newDoc+" is "+docScore.score); // TODO MC debug remove
    }
    * TODO MC */

    // return oldToNew; TODO MC
    return newToOld; // TODO MC
  }