コード例 #1
0
  @Test
  public void testSvValues() throws IOException {
    int numDocs = 1000000;
    int numOrdinals = numDocs / 4;
    Map<Integer, Long> controlDocToOrdinal = new HashMap<>();
    OrdinalsBuilder builder = new OrdinalsBuilder(numDocs);
    long ordinal = builder.currentOrdinal();
    for (int doc = 0; doc < numDocs; doc++) {
      if (doc % numOrdinals == 0) {
        ordinal = builder.nextOrdinal();
      }
      controlDocToOrdinal.put(doc, ordinal);
      builder.addDoc(doc);
    }

    Ordinals ords = builder.build(ImmutableSettings.EMPTY);
    assertThat(ords, instanceOf(SinglePackedOrdinals.class));
    RandomAccessOrds docs = ords.ordinals();
    final SortedDocValues singleOrds = DocValues.unwrapSingleton(docs);
    assertNotNull(singleOrds);

    for (Map.Entry<Integer, Long> entry : controlDocToOrdinal.entrySet()) {
      assertThat(entry.getValue(), equalTo((long) singleOrds.getOrd(entry.getKey())));
    }
  }
コード例 #2
0
    @Override
    public MultiGeoPointValues getGeoPointValues() {
      final RandomAccessOrds ords = ordinals.ordinals();
      final SortedDocValues singleOrds = DocValues.unwrapSingleton(ords);
      if (singleOrds != null) {
        final GeoPoint point = new GeoPoint();
        final GeoPointValues values =
            new GeoPointValues() {
              @Override
              public GeoPoint get(int docID) {
                final int ord = singleOrds.getOrd(docID);
                if (ord >= 0) {
                  return point.reset(lat.get(ord), lon.get(ord));
                }
                return point.reset(Double.NaN, Double.NaN);
              }
            };
        return FieldData.singleton(values, DocValues.docsWithValue(singleOrds, maxDoc));
      } else {
        final GeoPoint point = new GeoPoint();
        return new MultiGeoPointValues() {

          @Override
          public GeoPoint valueAt(int index) {
            final long ord = ords.ordAt(index);
            if (ord >= 0) {
              return point.reset(lat.get(ord), lon.get(ord));
            }
            return point.reset(Double.NaN, Double.NaN);
          }

          @Override
          public void setDocument(int docId) {
            ords.setDocument(docId);
          }

          @Override
          public int count() {
            return ords.cardinality();
          }
        };
      }
    }
コード例 #3
0
  private static SortedNumericDocValues withOrdinals(
      Ordinals ordinals, final LongValues values, int maxDoc) {
    final RandomAccessOrds ords = ordinals.ordinals();
    final SortedDocValues singleOrds = DocValues.unwrapSingleton(ords);
    if (singleOrds != null) {
      final NumericDocValues singleValues =
          new NumericDocValues() {
            @Override
            public long get(int docID) {
              final int ord = singleOrds.getOrd(docID);
              if (ord >= 0) {
                return values.get(singleOrds.getOrd(docID));
              } else {
                return 0;
              }
            }
          };
      return DocValues.singleton(singleValues, DocValues.docsWithValue(ords, maxDoc));
    } else {
      return new SortedNumericDocValues() {
        @Override
        public long valueAt(int index) {
          return values.get(ords.ordAt(index));
        }

        @Override
        public void setDocument(int doc) {
          ords.setDocument(doc);
        }

        @Override
        public int count() {
          return ords.cardinality();
        }
      };
    }
  }
コード例 #4
0
  public static NamedList<Integer> getCounts(
      SolrIndexSearcher searcher,
      DocSet docs,
      String fieldName,
      int offset,
      int limit,
      int mincount,
      boolean missing,
      String sort,
      String prefix,
      String contains,
      boolean ignoreCase)
      throws IOException {
    SchemaField schemaField = searcher.getSchema().getField(fieldName);
    FieldType ft = schemaField.getType();
    NamedList<Integer> res = new NamedList<>();

    // TODO: remove multiValuedFieldCache(), check dv type / uninversion type?
    final boolean multiValued = schemaField.multiValued() || ft.multiValuedFieldCache();

    final SortedSetDocValues si; // for term lookups only
    OrdinalMap ordinalMap = null; // for mapping per-segment ords to global ones
    if (multiValued) {
      si = searcher.getLeafReader().getSortedSetDocValues(fieldName);
      if (si instanceof MultiSortedSetDocValues) {
        ordinalMap = ((MultiSortedSetDocValues) si).mapping;
      }
    } else {
      SortedDocValues single = searcher.getLeafReader().getSortedDocValues(fieldName);
      si = single == null ? null : DocValues.singleton(single);
      if (single instanceof MultiSortedDocValues) {
        ordinalMap = ((MultiSortedDocValues) single).mapping;
      }
    }
    if (si == null) {
      return finalize(res, searcher, schemaField, docs, -1, missing);
    }
    if (si.getValueCount() >= Integer.MAX_VALUE) {
      throw new UnsupportedOperationException(
          "Currently this faceting method is limited to " + Integer.MAX_VALUE + " unique terms");
    }

    final BytesRefBuilder prefixRef;
    if (prefix == null) {
      prefixRef = null;
    } else if (prefix.length() == 0) {
      prefix = null;
      prefixRef = null;
    } else {
      prefixRef = new BytesRefBuilder();
      prefixRef.copyChars(prefix);
    }

    int startTermIndex, endTermIndex;
    if (prefix != null) {
      startTermIndex = (int) si.lookupTerm(prefixRef.get());
      if (startTermIndex < 0) startTermIndex = -startTermIndex - 1;
      prefixRef.append(UnicodeUtil.BIG_TERM);
      endTermIndex = (int) si.lookupTerm(prefixRef.get());
      assert endTermIndex < 0;
      endTermIndex = -endTermIndex - 1;
    } else {
      startTermIndex = -1;
      endTermIndex = (int) si.getValueCount();
    }

    final int nTerms = endTermIndex - startTermIndex;
    int missingCount = -1;
    final CharsRefBuilder charsRef = new CharsRefBuilder();
    if (nTerms > 0 && docs.size() >= mincount) {

      // count collection array only needs to be as big as the number of terms we are
      // going to collect counts for.
      final int[] counts = new int[nTerms];

      Filter filter = docs.getTopFilter();
      List<LeafReaderContext> leaves = searcher.getTopReaderContext().leaves();
      for (int subIndex = 0; subIndex < leaves.size(); subIndex++) {
        LeafReaderContext leaf = leaves.get(subIndex);
        DocIdSet dis =
            filter.getDocIdSet(leaf, null); // solr docsets already exclude any deleted docs
        DocIdSetIterator disi = null;
        if (dis != null) {
          disi = dis.iterator();
        }
        if (disi != null) {
          if (multiValued) {
            SortedSetDocValues sub = leaf.reader().getSortedSetDocValues(fieldName);
            if (sub == null) {
              sub = DocValues.emptySortedSet();
            }
            final SortedDocValues singleton = DocValues.unwrapSingleton(sub);
            if (singleton != null) {
              // some codecs may optimize SORTED_SET storage for single-valued fields
              accumSingle(counts, startTermIndex, singleton, disi, subIndex, ordinalMap);
            } else {
              accumMulti(counts, startTermIndex, sub, disi, subIndex, ordinalMap);
            }
          } else {
            SortedDocValues sub = leaf.reader().getSortedDocValues(fieldName);
            if (sub == null) {
              sub = DocValues.emptySorted();
            }
            accumSingle(counts, startTermIndex, sub, disi, subIndex, ordinalMap);
          }
        }
      }

      if (startTermIndex == -1) {
        missingCount = counts[0];
      }

      // IDEA: we could also maintain a count of "other"... everything that fell outside
      // of the top 'N'

      int off = offset;
      int lim = limit >= 0 ? limit : Integer.MAX_VALUE;

      if (sort.equals(FacetParams.FACET_SORT_COUNT)
          || sort.equals(FacetParams.FACET_SORT_COUNT_LEGACY)) {
        int maxsize = limit > 0 ? offset + limit : Integer.MAX_VALUE - 1;
        maxsize = Math.min(maxsize, nTerms);
        LongPriorityQueue queue =
            new LongPriorityQueue(Math.min(maxsize, 1000), maxsize, Long.MIN_VALUE);

        int min = mincount - 1; // the smallest value in the top 'N' values
        for (int i = (startTermIndex == -1) ? 1 : 0; i < nTerms; i++) {
          int c = counts[i];
          if (contains != null) {
            final BytesRef term = si.lookupOrd(startTermIndex + i);
            if (!SimpleFacets.contains(term.utf8ToString(), contains, ignoreCase)) {
              continue;
            }
          }
          if (c > min) {
            // NOTE: we use c>min rather than c>=min as an optimization because we are going in
            // index order, so we already know that the keys are ordered.  This can be very
            // important if a lot of the counts are repeated (like zero counts would be).

            // smaller term numbers sort higher, so subtract the term number instead
            long pair = (((long) c) << 32) + (Integer.MAX_VALUE - i);
            boolean displaced = queue.insert(pair);
            if (displaced) min = (int) (queue.top() >>> 32);
          }
        }

        // if we are deep paging, we don't have to order the highest "offset" counts.
        int collectCount = Math.max(0, queue.size() - off);
        assert collectCount <= lim;

        // the start and end indexes of our list "sorted" (starting with the highest value)
        int sortedIdxStart = queue.size() - (collectCount - 1);
        int sortedIdxEnd = queue.size() + 1;
        final long[] sorted = queue.sort(collectCount);

        for (int i = sortedIdxStart; i < sortedIdxEnd; i++) {
          long pair = sorted[i];
          int c = (int) (pair >>> 32);
          int tnum = Integer.MAX_VALUE - (int) pair;
          final BytesRef term = si.lookupOrd(startTermIndex + tnum);
          ft.indexedToReadable(term, charsRef);
          res.add(charsRef.toString(), c);
        }

      } else {
        // add results in index order
        int i = (startTermIndex == -1) ? 1 : 0;
        if (mincount <= 0 && contains == null) {
          // if mincount<=0 and we're not examining the values for contains, then
          // we won't discard any terms and we know exactly where to start.
          i += off;
          off = 0;
        }

        for (; i < nTerms; i++) {
          int c = counts[i];
          if (c < mincount) continue;
          BytesRef term = null;
          if (contains != null) {
            term = si.lookupOrd(startTermIndex + i);
            if (!SimpleFacets.contains(term.utf8ToString(), contains, ignoreCase)) {
              continue;
            }
          }
          if (--off >= 0) continue;
          if (--lim < 0) break;
          if (term == null) {
            term = si.lookupOrd(startTermIndex + i);
          }
          ft.indexedToReadable(term, charsRef);
          res.add(charsRef.toString(), c);
        }
      }
    }

    return finalize(res, searcher, schemaField, docs, missingCount, missing);
  }