예제 #1
0
  /** accumulates per-segment multi-valued facet counts, mapping to global ordinal space */
  static void accumMulti(
      int counts[],
      int startTermIndex,
      SortedSetDocValues si,
      DocIdSetIterator disi,
      int subIndex,
      OrdinalMap map)
      throws IOException {
    int doc;
    while ((doc = disi.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) {
      si.setDocument(doc);
      // strange do-while to collect the missing count (first ord is NO_MORE_ORDS)
      int term = (int) si.nextOrd();
      if (term < 0) {
        if (startTermIndex == -1) {
          counts[0]++; // missing count
        }
        continue;
      }

      do {
        if (map != null) {
          term = (int) map.getGlobalOrd(subIndex, term);
        }
        int arrIdx = term - startTermIndex;
        if (arrIdx >= 0 && arrIdx < counts.length) counts[arrIdx]++;
      } while ((term = (int) si.nextOrd()) >= 0);
    }
  }
예제 #2
0
  /**
   * "typical" multi-valued faceting: not too many unique values, no prefixing. maps to global
   * ordinals as a separate step
   */
  static void accumMultiSeg(
      int counts[], SortedSetDocValues si, DocIdSetIterator disi, int subIndex, OrdinalMap map)
      throws IOException {
    // First count in seg-ord space:
    final int segCounts[];
    if (map == null) {
      segCounts = counts;
    } else {
      segCounts = new int[1 + (int) si.getValueCount()];
    }

    int doc;
    while ((doc = disi.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) {
      si.setDocument(doc);
      int term = (int) si.nextOrd();
      if (term < 0) {
        counts[0]++; // missing
      } else {
        do {
          segCounts[1 + term]++;
        } while ((term = (int) si.nextOrd()) >= 0);
      }
    }

    // migrate to global ords (if necessary)
    if (map != null) {
      migrateGlobal(counts, segCounts, subIndex, map);
    }
  }
 @Override
 public void collect(int doc) throws IOException {
   docTermOrds.setDocument(doc);
   long ord;
   while ((ord = docTermOrds.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
     final BytesRef term = docTermOrds.lookupOrd(ord);
     collectorTerms.add(term);
   }
 }
    @Override
    public void collect(int doc) throws IOException {
      if (doc > groupFieldTermsIndex.docID()) {
        groupFieldTermsIndex.advance(doc);
      }

      int groupOrd;
      if (doc == groupFieldTermsIndex.docID()) {
        groupOrd = groupFieldTermsIndex.ordValue();
      } else {
        groupOrd = -1;
      }

      if (facetFieldNumTerms == 0) {
        int segmentGroupedFacetsIndex = groupOrd * (facetFieldNumTerms + 1);
        if (facetPrefix != null || segmentGroupedFacetHits.exists(segmentGroupedFacetsIndex)) {
          return;
        }

        segmentTotalCount++;
        segmentFacetCounts[facetFieldNumTerms]++;

        segmentGroupedFacetHits.put(segmentGroupedFacetsIndex);
        BytesRef groupKey;
        if (groupOrd == -1) {
          groupKey = null;
        } else {
          groupKey = BytesRef.deepCopyOf(groupFieldTermsIndex.lookupOrd(groupOrd));
        }
        groupedFacetHits.add(new GroupedFacetHit(groupKey, null));
        return;
      }

      if (doc > facetFieldDocTermOrds.docID()) {
        facetFieldDocTermOrds.advance(doc);
      }
      boolean empty = true;
      if (doc == facetFieldDocTermOrds.docID()) {
        long ord;
        while ((ord = facetFieldDocTermOrds.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
          process(groupOrd, (int) ord);
          empty = false;
        }
      }

      if (empty) {
        process(
            groupOrd,
            facetFieldNumTerms); // this facet ord is reserved for docs not containing facet field.
      }
    }
  /**
   * Merges the sortedset docvalues from <code>toMerge</code>.
   *
   * <p>The default implementation calls {@link #addSortedSetField}, passing an Iterable that merges
   * ordinals and values and filters deleted documents .
   */
  public void mergeSortedSetField(
      FieldInfo fieldInfo, final MergeState mergeState, List<SortedSetDocValues> toMerge)
      throws IOException {

    mergeState.checkAbort.work(mergeState.segmentInfo.getDocCount());

    final AtomicReader readers[] = mergeState.readers.toArray(new AtomicReader[toMerge.size()]);
    final SortedSetDocValues dvs[] = toMerge.toArray(new SortedSetDocValues[toMerge.size()]);

    // step 1: iterate thru each sub and mark terms still in use
    TermsEnum liveTerms[] = new TermsEnum[dvs.length];
    long[] weights = new long[liveTerms.length];
    for (int sub = 0; sub < liveTerms.length; sub++) {
      AtomicReader reader = readers[sub];
      SortedSetDocValues dv = dvs[sub];
      Bits liveDocs = reader.getLiveDocs();
      if (liveDocs == null) {
        liveTerms[sub] = dv.termsEnum();
        weights[sub] = dv.getValueCount();
      } else {
        LongBitSet bitset = new LongBitSet(dv.getValueCount());
        for (int i = 0; i < reader.maxDoc(); i++) {
          if (liveDocs.get(i)) {
            dv.setDocument(i);
            long ord;
            while ((ord = dv.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
              bitset.set(ord);
            }
          }
        }
        liveTerms[sub] = new BitsFilteredTermsEnum(dv.termsEnum(), bitset);
        weights[sub] = bitset.cardinality();
      }
    }

    // step 2: create ordinal map (this conceptually does the "merging")
    final OrdinalMap map = OrdinalMap.build(this, liveTerms, weights, PackedInts.COMPACT);

    // step 3: add field
    addSortedSetField(
        fieldInfo,
        // ord -> value
        new Iterable<BytesRef>() {
          @Override
          public Iterator<BytesRef> iterator() {
            return new Iterator<BytesRef>() {
              long currentOrd;

              @Override
              public boolean hasNext() {
                return currentOrd < map.getValueCount();
              }

              @Override
              public BytesRef next() {
                if (!hasNext()) {
                  throw new NoSuchElementException();
                }
                int segmentNumber = map.getFirstSegmentNumber(currentOrd);
                long segmentOrd = map.getFirstSegmentOrd(currentOrd);
                final BytesRef term = dvs[segmentNumber].lookupOrd(segmentOrd);
                currentOrd++;
                return term;
              }

              @Override
              public void remove() {
                throw new UnsupportedOperationException();
              }
            };
          }
        },
        // doc -> ord count
        new Iterable<Number>() {
          @Override
          public Iterator<Number> iterator() {
            return new Iterator<Number>() {
              int readerUpto = -1;
              int docIDUpto;
              int nextValue;
              AtomicReader currentReader;
              Bits currentLiveDocs;
              boolean nextIsSet;

              @Override
              public boolean hasNext() {
                return nextIsSet || setNext();
              }

              @Override
              public void remove() {
                throw new UnsupportedOperationException();
              }

              @Override
              public Number next() {
                if (!hasNext()) {
                  throw new NoSuchElementException();
                }
                assert nextIsSet;
                nextIsSet = false;
                // TODO make a mutable number
                return nextValue;
              }

              private boolean setNext() {
                while (true) {
                  if (readerUpto == readers.length) {
                    return false;
                  }

                  if (currentReader == null || docIDUpto == currentReader.maxDoc()) {
                    readerUpto++;
                    if (readerUpto < readers.length) {
                      currentReader = readers[readerUpto];
                      currentLiveDocs = currentReader.getLiveDocs();
                    }
                    docIDUpto = 0;
                    continue;
                  }

                  if (currentLiveDocs == null || currentLiveDocs.get(docIDUpto)) {
                    nextIsSet = true;
                    SortedSetDocValues dv = dvs[readerUpto];
                    dv.setDocument(docIDUpto);
                    nextValue = 0;
                    while (dv.nextOrd() != SortedSetDocValues.NO_MORE_ORDS) {
                      nextValue++;
                    }
                    docIDUpto++;
                    return true;
                  }

                  docIDUpto++;
                }
              }
            };
          }
        },
        // ords
        new Iterable<Number>() {
          @Override
          public Iterator<Number> iterator() {
            return new Iterator<Number>() {
              int readerUpto = -1;
              int docIDUpto;
              long nextValue;
              AtomicReader currentReader;
              Bits currentLiveDocs;
              LongValues currentMap;
              boolean nextIsSet;
              long ords[] = new long[8];
              int ordUpto;
              int ordLength;

              @Override
              public boolean hasNext() {
                return nextIsSet || setNext();
              }

              @Override
              public void remove() {
                throw new UnsupportedOperationException();
              }

              @Override
              public Number next() {
                if (!hasNext()) {
                  throw new NoSuchElementException();
                }
                assert nextIsSet;
                nextIsSet = false;
                // TODO make a mutable number
                return nextValue;
              }

              private boolean setNext() {
                while (true) {
                  if (readerUpto == readers.length) {
                    return false;
                  }

                  if (ordUpto < ordLength) {
                    nextValue = ords[ordUpto];
                    ordUpto++;
                    nextIsSet = true;
                    return true;
                  }

                  if (currentReader == null || docIDUpto == currentReader.maxDoc()) {
                    readerUpto++;
                    if (readerUpto < readers.length) {
                      currentReader = readers[readerUpto];
                      currentLiveDocs = currentReader.getLiveDocs();
                      currentMap = map.getGlobalOrds(readerUpto);
                    }
                    docIDUpto = 0;
                    continue;
                  }

                  if (currentLiveDocs == null || currentLiveDocs.get(docIDUpto)) {
                    assert docIDUpto < currentReader.maxDoc();
                    SortedSetDocValues dv = dvs[readerUpto];
                    dv.setDocument(docIDUpto);
                    ordUpto = ordLength = 0;
                    long ord;
                    while ((ord = dv.nextOrd()) != SortedSetDocValues.NO_MORE_ORDS) {
                      if (ordLength == ords.length) {
                        ords = ArrayUtil.grow(ords, ordLength + 1);
                      }
                      ords[ordLength] = currentMap.get(ord);
                      ordLength++;
                    }
                    docIDUpto++;
                    continue;
                  }

                  docIDUpto++;
                }
              }
            };
          }
        });
  }
  public void testDocValuesMemoryIndexVsNormalIndex() throws Exception {
    Document doc = new Document();
    long randomLong = random().nextLong();
    doc.add(new NumericDocValuesField("numeric", randomLong));
    if (random().nextBoolean()) {
      doc.add(new LegacyLongField("numeric", randomLong, Field.Store.NO));
    }
    int numValues = atLeast(5);
    for (int i = 0; i < numValues; i++) {
      randomLong = random().nextLong();
      doc.add(new SortedNumericDocValuesField("sorted_numeric", randomLong));
      if (random().nextBoolean()) {
        // randomly duplicate field/value
        doc.add(new SortedNumericDocValuesField("sorted_numeric", randomLong));
      }
      if (random().nextBoolean()) {
        doc.add(new LegacyLongField("numeric", randomLong, Field.Store.NO));
      }
    }
    BytesRef randomTerm = new BytesRef(randomTerm());
    doc.add(new BinaryDocValuesField("binary", randomTerm));
    if (random().nextBoolean()) {
      doc.add(new StringField("binary", randomTerm, Field.Store.NO));
    }
    randomTerm = new BytesRef(randomTerm());
    doc.add(new SortedDocValuesField("sorted", randomTerm));
    if (random().nextBoolean()) {
      doc.add(new StringField("sorted", randomTerm, Field.Store.NO));
    }
    numValues = atLeast(5);
    for (int i = 0; i < numValues; i++) {
      randomTerm = new BytesRef(randomTerm());
      doc.add(new SortedSetDocValuesField("sorted_set", randomTerm));
      if (random().nextBoolean()) {
        // randomly duplicate field/value
        doc.add(new SortedSetDocValuesField("sorted_set", randomTerm));
      }
      if (random().nextBoolean()) {
        // randomily just add a normal string field
        doc.add(new StringField("sorted_set", randomTerm, Field.Store.NO));
      }
    }

    MockAnalyzer mockAnalyzer = new MockAnalyzer(random());
    MemoryIndex memoryIndex = MemoryIndex.fromDocument(doc, mockAnalyzer);
    IndexReader indexReader = memoryIndex.createSearcher().getIndexReader();
    LeafReader leafReader = indexReader.leaves().get(0).reader();

    Directory dir = newDirectory();
    IndexWriter writer = new IndexWriter(dir, newIndexWriterConfig(random(), mockAnalyzer));
    writer.addDocument(doc);
    writer.close();
    IndexReader controlIndexReader = DirectoryReader.open(dir);
    LeafReader controlLeafReader = controlIndexReader.leaves().get(0).reader();

    NumericDocValues numericDocValues = leafReader.getNumericDocValues("numeric");
    NumericDocValues controlNumericDocValues = controlLeafReader.getNumericDocValues("numeric");
    assertEquals(controlNumericDocValues.get(0), numericDocValues.get(0));

    SortedNumericDocValues sortedNumericDocValues =
        leafReader.getSortedNumericDocValues("sorted_numeric");
    sortedNumericDocValues.setDocument(0);
    SortedNumericDocValues controlSortedNumericDocValues =
        controlLeafReader.getSortedNumericDocValues("sorted_numeric");
    controlSortedNumericDocValues.setDocument(0);
    assertEquals(controlSortedNumericDocValues.count(), sortedNumericDocValues.count());
    for (int i = 0; i < controlSortedNumericDocValues.count(); i++) {
      assertEquals(controlSortedNumericDocValues.valueAt(i), sortedNumericDocValues.valueAt(i));
    }

    BinaryDocValues binaryDocValues = leafReader.getBinaryDocValues("binary");
    BinaryDocValues controlBinaryDocValues = controlLeafReader.getBinaryDocValues("binary");
    assertEquals(controlBinaryDocValues.get(0), binaryDocValues.get(0));

    SortedDocValues sortedDocValues = leafReader.getSortedDocValues("sorted");
    SortedDocValues controlSortedDocValues = controlLeafReader.getSortedDocValues("sorted");
    assertEquals(controlSortedDocValues.getValueCount(), sortedDocValues.getValueCount());
    assertEquals(controlSortedDocValues.get(0), sortedDocValues.get(0));
    assertEquals(controlSortedDocValues.getOrd(0), sortedDocValues.getOrd(0));
    assertEquals(controlSortedDocValues.lookupOrd(0), sortedDocValues.lookupOrd(0));

    SortedSetDocValues sortedSetDocValues = leafReader.getSortedSetDocValues("sorted_set");
    sortedSetDocValues.setDocument(0);
    SortedSetDocValues controlSortedSetDocValues =
        controlLeafReader.getSortedSetDocValues("sorted_set");
    controlSortedSetDocValues.setDocument(0);
    assertEquals(controlSortedSetDocValues.getValueCount(), sortedSetDocValues.getValueCount());
    for (long controlOrd = controlSortedSetDocValues.nextOrd();
        controlOrd != SortedSetDocValues.NO_MORE_ORDS;
        controlOrd = controlSortedSetDocValues.nextOrd()) {
      assertEquals(controlOrd, sortedSetDocValues.nextOrd());
      assertEquals(
          controlSortedSetDocValues.lookupOrd(controlOrd),
          sortedSetDocValues.lookupOrd(controlOrd));
    }
    assertEquals(SortedSetDocValues.NO_MORE_ORDS, sortedSetDocValues.nextOrd());

    indexReader.close();
    controlIndexReader.close();
    dir.close();
  }