コード例 #1
0
    @Override
    public SeekStatus seekCeil(BytesRef target) throws IOException {

      // already here
      if (term != null && term.equals(target)) {
        return SeekStatus.FOUND;
      }

      int startIdx = Arrays.binarySearch(indexedTermsArray, target);

      if (startIdx >= 0) {
        // we hit the term exactly... lucky us!
        TermsEnum.SeekStatus seekStatus = termsEnum.seekCeil(target);
        assert seekStatus == TermsEnum.SeekStatus.FOUND;
        ord = startIdx << indexIntervalBits;
        setTerm();
        assert term != null;
        return SeekStatus.FOUND;
      }

      // we didn't hit the term exactly
      startIdx = -startIdx - 1;

      if (startIdx == 0) {
        // our target occurs *before* the first term
        TermsEnum.SeekStatus seekStatus = termsEnum.seekCeil(target);
        assert seekStatus == TermsEnum.SeekStatus.NOT_FOUND;
        ord = 0;
        setTerm();
        assert term != null;
        return SeekStatus.NOT_FOUND;
      }

      // back up to the start of the block
      startIdx--;

      if ((ord >> indexIntervalBits) == startIdx && term != null && term.compareTo(target) <= 0) {
        // we are already in the right block and the current term is before the term we want,
        // so we don't need to seek.
      } else {
        // seek to the right block
        TermsEnum.SeekStatus seekStatus = termsEnum.seekCeil(indexedTermsArray[startIdx]);
        assert seekStatus == TermsEnum.SeekStatus.FOUND;
        ord = startIdx << indexIntervalBits;
        setTerm();
        assert term != null; // should be non-null since it's in the index
      }

      while (term != null && term.compareTo(target) < 0) {
        next();
      }

      if (term == null) {
        return SeekStatus.END;
      } else if (term.compareTo(target) == 0) {
        return SeekStatus.FOUND;
      } else {
        return SeekStatus.NOT_FOUND;
      }
    }
コード例 #2
0
    @Override
    public void seekExact(long targetOrd) throws IOException {
      int delta = (int) (targetOrd - ordBase - ord);
      // System.out.println("  seek(ord) targetOrd=" + targetOrd + " delta=" + delta + " ord=" + ord
      // + " ii=" + indexInterval);
      if (delta < 0 || delta > indexInterval) {
        final int idx = (int) (targetOrd >>> indexIntervalBits);
        final BytesRef base = indexedTermsArray[idx];
        // System.out.println("  do seek term=" + base.utf8ToString());
        ord = idx << indexIntervalBits;
        delta = (int) (targetOrd - ord);
        final TermsEnum.SeekStatus seekStatus = termsEnum.seekCeil(base);
        assert seekStatus == TermsEnum.SeekStatus.FOUND;
      } else {
        // System.out.println("seek w/in block");
      }

      while (--delta >= 0) {
        BytesRef br = termsEnum.next();
        if (br == null) {
          assert false;
          return;
        }
        ord++;
      }

      setTerm();
      assert term != null;
    }
コード例 #3
0
ファイル: LuceneUtil.java プロジェクト: INL/BlackLab
 /**
  * Find terms in the index based on a prefix. Useful for autocomplete.
  *
  * @param index the index
  * @param fieldName the field
  * @param prefix the prefix we're looking for (null or empty string for all terms)
  * @param sensitive match case-sensitively or not?
  * @param maxResults max. number of results to return (or -1 for all)
  * @return the matching terms
  */
 public static List<String> findTermsByPrefix(
     LeafReader index, String fieldName, String prefix, boolean sensitive, int maxResults) {
   boolean allTerms = prefix == null || prefix.length() == 0;
   if (allTerms) {
     prefix = "";
     sensitive = true; // don't do unnecessary work in this case
   }
   try {
     if (!sensitive) prefix = StringUtil.removeAccents(prefix).toLowerCase();
     org.apache.lucene.index.Terms terms = index.terms(fieldName);
     List<String> results = new ArrayList<>();
     TermsEnum termsEnum = terms.iterator();
     BytesRef brPrefix = new BytesRef(prefix.getBytes(LUCENE_DEFAULT_CHARSET));
     termsEnum.seekCeil(brPrefix); // find the prefix in the terms list
     while (maxResults < 0 || results.size() < maxResults) {
       BytesRef term = termsEnum.next();
       if (term == null) break;
       String termText = term.utf8ToString();
       String optDesensitized = termText;
       if (!sensitive) optDesensitized = StringUtil.removeAccents(termText).toLowerCase();
       if (!allTerms && !optDesensitized.substring(0, prefix.length()).equalsIgnoreCase(prefix)) {
         // Doesn't match prefix or different field; no more matches
         break;
       }
       // Match, add term
       results.add(termText);
     }
     return results;
   } catch (IOException e) {
     throw new RuntimeException(e);
   }
 }
コード例 #4
0
  private void getPrefixTerms(
      ObjectHashSet<Term> terms, final Term prefix, final IndexReader reader) throws IOException {
    // SlowCompositeReaderWrapper could be used... but this would merge all terms from each segment
    // into one terms
    // instance, which is very expensive. Therefore I think it is better to iterate over each leaf
    // individually.
    List<LeafReaderContext> leaves = reader.leaves();
    for (LeafReaderContext leaf : leaves) {
      Terms _terms = leaf.reader().terms(field);
      if (_terms == null) {
        continue;
      }

      TermsEnum termsEnum = _terms.iterator();
      TermsEnum.SeekStatus seekStatus = termsEnum.seekCeil(prefix.bytes());
      if (TermsEnum.SeekStatus.END == seekStatus) {
        continue;
      }

      for (BytesRef term = termsEnum.term(); term != null; term = termsEnum.next()) {
        if (!StringHelper.startsWith(term, prefix.bytes())) {
          break;
        }

        terms.add(new Term(field, BytesRef.deepCopyOf(term)));
        if (terms.size() >= maxExpansions) {
          return;
        }
      }
    }
  }
コード例 #5
0
  public void testSeekCeilNotFound() throws Exception {
    Directory dir = newDirectory();
    RandomIndexWriter w = new RandomIndexWriter(random(), dir);
    Document doc = new Document();
    // Get empty string in there!
    doc.add(newStringField("field", "", Field.Store.NO));
    w.addDocument(doc);

    for (int i = 0; i < 36; i++) {
      doc = new Document();
      String term = "" + (char) (97 + i);
      String term2 = "a" + (char) (97 + i);
      doc.add(newTextField("field", term + " " + term2, Field.Store.NO));
      w.addDocument(doc);
    }

    w.forceMerge(1);
    IndexReader r = w.getReader();
    TermsEnum te = MultiFields.getTerms(r, "field").iterator(null);
    assertEquals(TermsEnum.SeekStatus.NOT_FOUND, te.seekCeil(new BytesRef(new byte[] {0x22})));
    assertEquals("a", te.term().utf8ToString());
    assertEquals(1L, te.ord());
    r.close();
    w.close();
    dir.close();
  }
コード例 #6
0
 @Override
 public long lookupTerm(BytesRef key) {
   try {
     switch (te.seekCeil(key)) {
       case FOUND:
         assert te.ord() >= 0;
         return te.ord();
       case NOT_FOUND:
         assert te.ord() >= 0;
         return -te.ord() - 1;
       default: /* END */
         return -numTerms() - 1;
     }
   } catch (IOException e) {
     throw new RuntimeException(e);
   }
 }
コード例 #7
0
  @Override
  public void visitMatchingTerms(IndexReader reader, String fieldName, MatchingTermVisitor mtv)
      throws IOException {
    int prefixLength = prefix.length();
    Terms terms = MultiFields.getTerms(reader, fieldName);
    if (terms != null) {
      Matcher matcher = pattern.matcher("");
      try {
        TermsEnum termsEnum = terms.iterator(null);

        TermsEnum.SeekStatus status = termsEnum.seekCeil(prefixRef);
        BytesRef text;
        if (status == TermsEnum.SeekStatus.FOUND) {
          text = prefixRef;
        } else if (status == TermsEnum.SeekStatus.NOT_FOUND) {
          text = termsEnum.term();
        } else {
          text = null;
        }

        while (text != null) {
          if (text != null && StringHelper.startsWith(text, prefixRef)) {
            String textString = text.utf8ToString();
            matcher.reset(textString.substring(prefixLength));
            if (matcher.matches()) {
              mtv.visitMatchingTerm(new Term(fieldName, textString));
            }
          } else {
            break;
          }
          text = termsEnum.next();
        }
      } finally {
        matcher.reset();
      }
    }
  }
コード例 #8
0
  /** Call this only once (if you subclass!) */
  protected void uninvert(final LeafReader reader, Bits liveDocs, final BytesRef termPrefix)
      throws IOException {
    final FieldInfo info = reader.getFieldInfos().fieldInfo(field);
    if (checkForDocValues && info != null && info.getDocValuesType() != DocValuesType.NONE) {
      throw new IllegalStateException(
          "Type mismatch: " + field + " was indexed as " + info.getDocValuesType());
    }
    // System.out.println("DTO uninvert field=" + field + " prefix=" + termPrefix);
    final long startTime = System.nanoTime();
    prefix = termPrefix == null ? null : BytesRef.deepCopyOf(termPrefix);

    final int maxDoc = reader.maxDoc();
    final int[] index =
        new int
            [maxDoc]; // immediate term numbers, or the index into the byte[] representing the last
    // number
    final int[] lastTerm = new int[maxDoc]; // last term we saw for this document
    final byte[][] bytes =
        new byte[maxDoc][]; // list of term numbers for the doc (delta encoded vInts)

    final Terms terms = reader.terms(field);
    if (terms == null) {
      // No terms
      return;
    }

    final TermsEnum te = terms.iterator();
    final BytesRef seekStart = termPrefix != null ? termPrefix : new BytesRef();
    // System.out.println("seekStart=" + seekStart.utf8ToString());
    if (te.seekCeil(seekStart) == TermsEnum.SeekStatus.END) {
      // No terms match
      return;
    }

    // For our "term index wrapper"
    final List<BytesRef> indexedTerms = new ArrayList<>();
    final PagedBytes indexedTermsBytes = new PagedBytes(15);

    // we need a minimum of 9 bytes, but round up to 12 since the space would
    // be wasted with most allocators anyway.
    byte[] tempArr = new byte[12];

    //
    // enumerate all terms, and build an intermediate form of the un-inverted field.
    //
    // During this intermediate form, every document has a (potential) byte[]
    // and the int[maxDoc()] array either contains the termNumber list directly
    // or the *end* offset of the termNumber list in its byte array (for faster
    // appending and faster creation of the final form).
    //
    // idea... if things are too large while building, we could do a range of docs
    // at a time (but it would be a fair amount slower to build)
    // could also do ranges in parallel to take advantage of multiple CPUs

    // OPTIONAL: remap the largest df terms to the lowest 128 (single byte)
    // values.  This requires going over the field first to find the most
    // frequent terms ahead of time.

    int termNum = 0;
    postingsEnum = null;

    // Loop begins with te positioned to first term (we call
    // seek above):
    for (; ; ) {
      final BytesRef t = te.term();
      if (t == null || (termPrefix != null && !StringHelper.startsWith(t, termPrefix))) {
        break;
      }
      // System.out.println("visit term=" + t.utf8ToString() + " " + t + " termNum=" + termNum);

      visitTerm(te, termNum);

      if ((termNum & indexIntervalMask) == 0) {
        // Index this term
        sizeOfIndexedStrings += t.length;
        BytesRef indexedTerm = new BytesRef();
        indexedTermsBytes.copy(t, indexedTerm);
        // TODO: really should 1) strip off useless suffix,
        // and 2) use FST not array/PagedBytes
        indexedTerms.add(indexedTerm);
      }

      final int df = te.docFreq();
      if (df <= maxTermDocFreq) {

        postingsEnum = te.postings(postingsEnum, PostingsEnum.NONE);

        // dF, but takes deletions into account
        int actualDF = 0;

        for (; ; ) {
          int doc = postingsEnum.nextDoc();
          if (doc == DocIdSetIterator.NO_MORE_DOCS) {
            break;
          }
          // System.out.println("  chunk=" + chunk + " docs");

          actualDF++;
          termInstances++;

          // System.out.println("    docID=" + doc);
          // add TNUM_OFFSET to the term number to make room for special reserved values:
          // 0 (end term) and 1 (index into byte array follows)
          int delta = termNum - lastTerm[doc] + TNUM_OFFSET;
          lastTerm[doc] = termNum;
          int val = index[doc];

          if ((val & 0xff) == 1) {
            // index into byte array (actually the end of
            // the doc-specific byte[] when building)
            int pos = val >>> 8;
            int ilen = vIntSize(delta);
            byte[] arr = bytes[doc];
            int newend = pos + ilen;
            if (newend > arr.length) {
              // We avoid a doubling strategy to lower memory usage.
              // this faceting method isn't for docs with many terms.
              // In hotspot, objects have 2 words of overhead, then fields, rounded up to a 64-bit
              // boundary.
              // TODO: figure out what array lengths we can round up to w/o actually using more
              // memory
              // (how much space does a byte[] take up?  Is data preceded by a 32 bit length only?
              // It should be safe to round up to the nearest 32 bits in any case.
              int newLen = (newend + 3) & 0xfffffffc; // 4 byte alignment
              byte[] newarr = new byte[newLen];
              System.arraycopy(arr, 0, newarr, 0, pos);
              arr = newarr;
              bytes[doc] = newarr;
            }
            pos = writeInt(delta, arr, pos);
            index[doc] = (pos << 8) | 1; // update pointer to end index in byte[]
          } else {
            // OK, this int has data in it... find the end (a zero starting byte - not
            // part of another number, hence not following a byte with the high bit set).
            int ipos;
            if (val == 0) {
              ipos = 0;
            } else if ((val & 0x0000ff80) == 0) {
              ipos = 1;
            } else if ((val & 0x00ff8000) == 0) {
              ipos = 2;
            } else if ((val & 0xff800000) == 0) {
              ipos = 3;
            } else {
              ipos = 4;
            }

            // System.out.println("      ipos=" + ipos);

            int endPos = writeInt(delta, tempArr, ipos);
            // System.out.println("      endpos=" + endPos);
            if (endPos <= 4) {
              // System.out.println("      fits!");
              // value will fit in the integer... move bytes back
              for (int j = ipos; j < endPos; j++) {
                val |= (tempArr[j] & 0xff) << (j << 3);
              }
              index[doc] = val;
            } else {
              // value won't fit... move integer into byte[]
              for (int j = 0; j < ipos; j++) {
                tempArr[j] = (byte) val;
                val >>>= 8;
              }
              // point at the end index in the byte[]
              index[doc] = (endPos << 8) | 1;
              bytes[doc] = tempArr;
              tempArr = new byte[12];
            }
          }
        }
        setActualDocFreq(termNum, actualDF);
      }

      termNum++;
      if (te.next() == null) {
        break;
      }
    }

    numTermsInField = termNum;

    long midPoint = System.nanoTime();

    if (termInstances == 0) {
      // we didn't invert anything
      // lower memory consumption.
      tnums = null;
    } else {

      this.index = index;

      //
      // transform intermediate form into the final form, building a single byte[]
      // at a time, and releasing the intermediate byte[]s as we go to avoid
      // increasing the memory footprint.
      //

      for (int pass = 0; pass < 256; pass++) {
        byte[] target = tnums[pass];
        int pos = 0; // end in target;
        if (target != null) {
          pos = target.length;
        } else {
          target = new byte[4096];
        }

        // loop over documents, 0x00ppxxxx, 0x01ppxxxx, 0x02ppxxxx
        // where pp is the pass (which array we are building), and xx is all values.
        // each pass shares the same byte[] for termNumber lists.
        for (int docbase = pass << 16; docbase < maxDoc; docbase += (1 << 24)) {
          int lim = Math.min(docbase + (1 << 16), maxDoc);
          for (int doc = docbase; doc < lim; doc++) {
            // System.out.println("  pass="******" process docID=" + doc);
            int val = index[doc];
            if ((val & 0xff) == 1) {
              int len = val >>> 8;
              // System.out.println("    ptr pos=" + pos);
              index[doc] = (pos << 8) | 1; // change index to point to start of array
              if ((pos & 0xff000000) != 0) {
                // we only have 24 bits for the array index
                throw new IllegalStateException(
                    "Too many values for UnInvertedField faceting on field " + field);
              }
              byte[] arr = bytes[doc];
              /*
              for(byte b : arr) {
                //System.out.println("      b=" + Integer.toHexString((int) b));
              }
              */
              bytes[doc] = null; // IMPORTANT: allow GC to avoid OOM
              if (target.length <= pos + len) {
                int newlen = target.length;
                /**
                 * * we don't have to worry about the array getting too large since the "pos" param
                 * will overflow first (only 24 bits available) if ((newlen<<1) <= 0) { //
                 * overflow... newlen = Integer.MAX_VALUE; if (newlen <= pos + len) { throw new
                 * SolrException(400,"Too many terms to uninvert field!"); } } else { while (newlen
                 * <= pos + len) newlen<<=1; // doubling strategy } **
                 */
                while (newlen <= pos + len) newlen <<= 1; // doubling strategy
                byte[] newtarget = new byte[newlen];
                System.arraycopy(target, 0, newtarget, 0, pos);
                target = newtarget;
              }
              System.arraycopy(arr, 0, target, pos, len);
              pos += len + 1; // skip single byte at end and leave it 0 for terminator
            }
          }
        }

        // shrink array
        if (pos < target.length) {
          byte[] newtarget = new byte[pos];
          System.arraycopy(target, 0, newtarget, 0, pos);
          target = newtarget;
        }

        tnums[pass] = target;

        if ((pass << 16) > maxDoc) break;
      }
    }
    indexedTermsArray = indexedTerms.toArray(new BytesRef[indexedTerms.size()]);

    long endTime = System.nanoTime();

    total_time = (int) TimeUnit.MILLISECONDS.convert(endTime - startTime, TimeUnit.NANOSECONDS);
    phase1_time = (int) TimeUnit.MILLISECONDS.convert(midPoint - startTime, TimeUnit.NANOSECONDS);
  }
コード例 #9
0
  @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();
  }
コード例 #10
0
  @Override
  public void process(ResponseBuilder rb) throws IOException {
    SolrParams params = rb.req.getParams();
    if (!params.getBool(TermsParams.TERMS, false)) return;

    String[] fields = params.getParams(TermsParams.TERMS_FIELD);

    NamedList<Object> termsResult = new SimpleOrderedMap<>();
    rb.rsp.add("terms", termsResult);

    if (fields == null || fields.length == 0) return;

    int limit = params.getInt(TermsParams.TERMS_LIMIT, 10);
    if (limit < 0) {
      limit = Integer.MAX_VALUE;
    }

    String lowerStr = params.get(TermsParams.TERMS_LOWER);
    String upperStr = params.get(TermsParams.TERMS_UPPER);
    boolean upperIncl = params.getBool(TermsParams.TERMS_UPPER_INCLUSIVE, false);
    boolean lowerIncl = params.getBool(TermsParams.TERMS_LOWER_INCLUSIVE, true);
    boolean sort =
        !TermsParams.TERMS_SORT_INDEX.equals(
            params.get(TermsParams.TERMS_SORT, TermsParams.TERMS_SORT_COUNT));
    int freqmin = params.getInt(TermsParams.TERMS_MINCOUNT, 1);
    int freqmax = params.getInt(TermsParams.TERMS_MAXCOUNT, UNLIMITED_MAX_COUNT);
    if (freqmax < 0) {
      freqmax = Integer.MAX_VALUE;
    }
    String prefix = params.get(TermsParams.TERMS_PREFIX_STR);
    String regexp = params.get(TermsParams.TERMS_REGEXP_STR);
    Pattern pattern = regexp != null ? Pattern.compile(regexp, resolveRegexpFlags(params)) : null;

    boolean raw = params.getBool(TermsParams.TERMS_RAW, false);

    final AtomicReader indexReader = rb.req.getSearcher().getAtomicReader();
    Fields lfields = indexReader.fields();

    for (String field : fields) {
      NamedList<Integer> fieldTerms = new NamedList<>();
      termsResult.add(field, fieldTerms);

      Terms terms = lfields == null ? null : lfields.terms(field);
      if (terms == null) {
        // no terms for this field
        continue;
      }

      FieldType ft = raw ? null : rb.req.getSchema().getFieldTypeNoEx(field);
      if (ft == null) ft = new StrField();

      // prefix must currently be text
      BytesRef prefixBytes = prefix == null ? null : new BytesRef(prefix);

      BytesRef upperBytes = null;
      if (upperStr != null) {
        upperBytes = new BytesRef();
        ft.readableToIndexed(upperStr, upperBytes);
      }

      BytesRef lowerBytes;
      if (lowerStr == null) {
        // If no lower bound was specified, use the prefix
        lowerBytes = prefixBytes;
      } else {
        lowerBytes = new BytesRef();
        if (raw) {
          // TODO: how to handle binary? perhaps we don't for "raw"... or if the field exists
          // perhaps we detect if the FieldType is non-character and expect hex if so?
          lowerBytes = new BytesRef(lowerStr);
        } else {
          lowerBytes = new BytesRef();
          ft.readableToIndexed(lowerStr, lowerBytes);
        }
      }

      TermsEnum termsEnum = terms.iterator(null);
      BytesRef term = null;

      if (lowerBytes != null) {
        if (termsEnum.seekCeil(lowerBytes) == TermsEnum.SeekStatus.END) {
          termsEnum = null;
        } else {
          term = termsEnum.term();
          // Only advance the enum if we are excluding the lower bound and the lower Term actually
          // matches
          if (lowerIncl == false && term.equals(lowerBytes)) {
            term = termsEnum.next();
          }
        }
      } else {
        // position termsEnum on first term
        term = termsEnum.next();
      }

      int i = 0;
      BoundedTreeSet<CountPair<BytesRef, Integer>> queue =
          (sort ? new BoundedTreeSet<CountPair<BytesRef, Integer>>(limit) : null);
      CharsRef external = new CharsRef();
      while (term != null && (i < limit || sort)) {
        boolean externalized = false; // did we fill in "external" yet for this term?

        // stop if the prefix doesn't match
        if (prefixBytes != null && !StringHelper.startsWith(term, prefixBytes)) break;

        if (pattern != null) {
          // indexed text or external text?
          // TODO: support "raw" mode?
          ft.indexedToReadable(term, external);
          externalized = true;
          if (!pattern.matcher(external).matches()) {
            term = termsEnum.next();
            continue;
          }
        }

        if (upperBytes != null) {
          int upperCmp = term.compareTo(upperBytes);
          // if we are past the upper term, or equal to it (when don't include upper) then stop.
          if (upperCmp > 0 || (upperCmp == 0 && !upperIncl)) break;
        }

        // This is a good term in the range.  Check if mincount/maxcount conditions are satisfied.
        int docFreq = termsEnum.docFreq();
        if (docFreq >= freqmin && docFreq <= freqmax) {
          // add the term to the list
          if (sort) {
            queue.add(new CountPair<>(BytesRef.deepCopyOf(term), docFreq));
          } else {

            // TODO: handle raw somehow
            if (!externalized) {
              ft.indexedToReadable(term, external);
            }
            fieldTerms.add(external.toString(), docFreq);
            i++;
          }
        }

        term = termsEnum.next();
      }

      if (sort) {
        for (CountPair<BytesRef, Integer> item : queue) {
          if (i >= limit) break;
          ft.indexedToReadable(item.key, external);
          fieldTerms.add(external.toString(), item.val);
          i++;
        }
      }
    }
  }
コード例 #11
0
  @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();
  }
コード例 #12
0
ファイル: IndexDatabase.java プロジェクト: battbeach/OpenGrok
  /**
   * Update the content of this index database
   *
   * @throws IOException if an error occurs
   * @throws HistoryException if an error occurs when accessing the history
   */
  public void update() throws IOException, HistoryException {
    synchronized (lock) {
      if (running) {
        throw new IOException("Indexer already running!");
      }
      running = true;
      interrupted = false;
    }

    String ctgs = RuntimeEnvironment.getInstance().getCtags();
    if (ctgs != null) {
      ctags = new Ctags();
      ctags.setBinary(ctgs);
    }
    if (ctags == null) {
      log.severe("Unable to run ctags! searching definitions will not work!");
    }

    if (ctags != null) {
      String filename = RuntimeEnvironment.getInstance().getCTagsExtraOptionsFile();
      if (filename != null) {
        ctags.setCTagsExtraOptionsFile(filename);
      }
    }

    try {
      Analyzer analyzer = AnalyzerGuru.getAnalyzer();
      IndexWriterConfig iwc = new IndexWriterConfig(SearchEngine.LUCENE_VERSION, analyzer);
      iwc.setOpenMode(OpenMode.CREATE_OR_APPEND);
      // iwc.setRAMBufferSizeMB(256.0);  //TODO check what is the sweet spot
      writer = new IndexWriter(indexDirectory, iwc);
      writer.commit(); // to make sure index exists on the disk
      // writer.setMaxFieldLength(RuntimeEnvironment.getInstance().getIndexWordLimit());

      if (directories.isEmpty()) {
        if (project == null) {
          directories.add("");
        } else {
          directories.add(project.getPath());
        }
      }

      for (String dir : directories) {
        File sourceRoot;
        if ("".equals(dir)) {
          sourceRoot = RuntimeEnvironment.getInstance().getSourceRootFile();
        } else {
          sourceRoot = new File(RuntimeEnvironment.getInstance().getSourceRootFile(), dir);
        }

        HistoryGuru.getInstance().ensureHistoryCacheExists(sourceRoot);

        String startuid = Util.path2uid(dir, "");
        IndexReader reader = DirectoryReader.open(indexDirectory); // open existing index
        Terms terms = null;
        int numDocs = reader.numDocs();
        if (numDocs > 0) {
          Fields uFields = MultiFields.getFields(reader); // reader.getTermVectors(0);
          terms = uFields.terms(QueryBuilder.U);
        }

        try {
          if (numDocs > 0) {
            uidIter = terms.iterator(null);
            TermsEnum.SeekStatus stat = uidIter.seekCeil(new BytesRef(startuid), true); // init uid
            if (stat == TermsEnum.SeekStatus.END || stat == TermsEnum.SeekStatus.NOT_FOUND) {
              uidIter = null;
            }
          }
          // TODO below should be optional, since it traverses the tree once more to get total
          // count! :(
          int file_cnt = 0;
          if (RuntimeEnvironment.getInstance().isPrintProgress()) {
            log.log(Level.INFO, "Counting files in {0} ...", dir);
            file_cnt = indexDown(sourceRoot, dir, true, 0, 0);
            if (log.isLoggable(Level.INFO)) {
              log.log(
                  Level.INFO, "Need to process: {0} files for {1}", new Object[] {file_cnt, dir});
            }
          }

          indexDown(sourceRoot, dir, false, 0, file_cnt);

          while (uidIter != null
              && uidIter.term() != null
              && uidIter.term().utf8ToString().startsWith(startuid)) {
            removeFile();
            uidIter.next();
          }
        } finally {
          reader.close();
        }
      }
    } finally {
      if (writer != null) {
        try {
          writer.prepareCommit();
          writer.commit();
          writer.close();
        } catch (IOException e) {
          log.log(Level.WARNING, "An error occured while closing writer", e);
        }
      }

      if (ctags != null) {
        try {
          ctags.close();
        } catch (IOException e) {
          log.log(Level.WARNING, "An error occured while closing ctags process", e);
        }
      }

      synchronized (lock) {
        running = false;
      }
    }

    if (!isInterrupted() && isDirty()) {
      if (RuntimeEnvironment.getInstance().isOptimizeDatabase()) {
        optimize();
      }
      createSpellingSuggestions();
      RuntimeEnvironment env = RuntimeEnvironment.getInstance();
      File timestamp = new File(env.getDataRootFile(), "timestamp");
      if (timestamp.exists()) {
        if (!timestamp.setLastModified(System.currentTimeMillis())) {
          log.log(
              Level.WARNING,
              "Failed to set last modified time on ''{0}'', used for timestamping the index database.",
              timestamp.getAbsolutePath());
        }
      } else {
        if (!timestamp.createNewFile()) {
          log.log(
              Level.WARNING,
              "Failed to create file ''{0}'', used for timestamping the index database.",
              timestamp.getAbsolutePath());
        }
      }
    }
  }
コード例 #13
0
    @Override
    protected void doSetNextReader(LeafReaderContext context) throws IOException {
      if (segmentFacetCounts != null) {
        segmentResults.add(createSegmentResult());
      }

      groupFieldTermsIndex = DocValues.getSorted(context.reader(), groupField);
      facetFieldDocTermOrds = DocValues.getSortedSet(context.reader(), facetField);
      facetFieldNumTerms = (int) facetFieldDocTermOrds.getValueCount();
      if (facetFieldNumTerms == 0) {
        facetOrdTermsEnum = null;
      } else {
        facetOrdTermsEnum = facetFieldDocTermOrds.termsEnum();
      }
      // [facetFieldNumTerms() + 1] for all possible facet values and docs not containing facet
      // field
      segmentFacetCounts = new int[facetFieldNumTerms + 1];
      segmentTotalCount = 0;

      segmentGroupedFacetHits.clear();
      for (GroupedFacetHit groupedFacetHit : groupedFacetHits) {
        int groupOrd =
            groupedFacetHit.groupValue == null
                ? -1
                : groupFieldTermsIndex.lookupTerm(groupedFacetHit.groupValue);
        if (groupedFacetHit.groupValue != null && groupOrd < 0) {
          continue;
        }

        int facetOrd;
        if (groupedFacetHit.facetValue != null) {
          if (facetOrdTermsEnum == null
              || !facetOrdTermsEnum.seekExact(groupedFacetHit.facetValue)) {
            continue;
          }
          facetOrd = (int) facetOrdTermsEnum.ord();
        } else {
          facetOrd = facetFieldNumTerms;
        }

        // (facetFieldDocTermOrds.numTerms() + 1) for all possible facet values and docs not
        // containing facet field
        int segmentGroupedFacetsIndex = groupOrd * (facetFieldNumTerms + 1) + facetOrd;
        segmentGroupedFacetHits.put(segmentGroupedFacetsIndex);
      }

      if (facetPrefix != null) {
        TermsEnum.SeekStatus seekStatus;
        if (facetOrdTermsEnum != null) {
          seekStatus = facetOrdTermsEnum.seekCeil(facetPrefix);
        } else {
          seekStatus = TermsEnum.SeekStatus.END;
        }

        if (seekStatus != TermsEnum.SeekStatus.END) {
          startFacetOrd = (int) facetOrdTermsEnum.ord();
        } else {
          startFacetOrd = 0;
          endFacetOrd = 0;
          return;
        }

        BytesRefBuilder facetEndPrefix = new BytesRefBuilder();
        facetEndPrefix.append(facetPrefix);
        facetEndPrefix.append(UnicodeUtil.BIG_TERM);
        seekStatus = facetOrdTermsEnum.seekCeil(facetEndPrefix.get());
        if (seekStatus != TermsEnum.SeekStatus.END) {
          endFacetOrd = (int) facetOrdTermsEnum.ord();
        } else {
          endFacetOrd = facetFieldNumTerms; // Don't include null...
        }
      } else {
        startFacetOrd = 0;
        endFacetOrd = facetFieldNumTerms + 1;
      }
    }
コード例 #14
0
  private void assertTermsSeeking(Terms leftTerms, Terms rightTerms) throws Exception {
    TermsEnum leftEnum = null;
    TermsEnum rightEnum = null;

    // just an upper bound
    int numTests = atLeast(20);
    Random random = random();

    // collect this number of terms from the left side
    HashSet<BytesRef> tests = new HashSet<BytesRef>();
    int numPasses = 0;
    while (numPasses < 10 && tests.size() < numTests) {
      leftEnum = leftTerms.iterator(leftEnum);
      BytesRef term = null;
      while ((term = leftEnum.next()) != null) {
        int code = random.nextInt(10);
        if (code == 0) {
          // the term
          tests.add(BytesRef.deepCopyOf(term));
        } else if (code == 1) {
          // truncated subsequence of term
          term = BytesRef.deepCopyOf(term);
          if (term.length > 0) {
            // truncate it
            term.length = random.nextInt(term.length);
          }
        } else if (code == 2) {
          // term, but ensure a non-zero offset
          byte newbytes[] = new byte[term.length + 5];
          System.arraycopy(term.bytes, term.offset, newbytes, 5, term.length);
          tests.add(new BytesRef(newbytes, 5, term.length));
        }
      }
      numPasses++;
    }

    ArrayList<BytesRef> shuffledTests = new ArrayList<BytesRef>(tests);
    Collections.shuffle(shuffledTests, random);

    for (BytesRef b : shuffledTests) {
      leftEnum = leftTerms.iterator(leftEnum);
      rightEnum = rightTerms.iterator(rightEnum);

      assertEquals(leftEnum.seekExact(b), rightEnum.seekExact(b));
      assertEquals(leftEnum.seekExact(b), rightEnum.seekExact(b));

      SeekStatus leftStatus;
      SeekStatus rightStatus;

      leftStatus = leftEnum.seekCeil(b);
      rightStatus = rightEnum.seekCeil(b);
      assertEquals(leftStatus, rightStatus);
      if (leftStatus != SeekStatus.END) {
        assertEquals(leftEnum.term(), rightEnum.term());
      }

      leftStatus = leftEnum.seekCeil(b);
      rightStatus = rightEnum.seekCeil(b);
      assertEquals(leftStatus, rightStatus);
      if (leftStatus != SeekStatus.END) {
        assertEquals(leftEnum.term(), rightEnum.term());
      }
    }
  }
コード例 #15
0
  @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, NavigableSet<String>> 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);

      final 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) {
          NavigableSet<String> parentIds;
          if (childValueToParentIds.containsKey(childValue)) {
            parentIds = childValueToParentIds.lget();
          } else {
            childValueToParentIds.put(childValue, parentIds = new TreeSet<>());
          }
          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.Searcher(ChildrenConstantScoreQueryTests.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 = scaledRandomIntBetween(0, numberOfUpdatableParents);
        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);
      QueryBuilder queryBuilder;
      if (random().nextBoolean()) {
        queryBuilder =
            hasChildQuery("child", termQuery("field1", childValue))
                .setShortCircuitCutoff(shortCircuitParentDocSet);
      } else {
        queryBuilder =
            constantScoreQuery(
                hasChildFilter("child", termQuery("field1", childValue))
                    .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());
      searcher.search(query, collector);
      FixedBitSet actualResult = collector.getResult();

      FixedBitSet expectedResult = new FixedBitSet(indexReader.maxDoc());
      if (childValueToParentIds.containsKey(childValue)) {
        LeafReader slowLeafReader = SlowCompositeReaderWrapper.wrap(indexReader);
        Terms terms = slowLeafReader.terms(UidFieldMapper.NAME);
        if (terms != null) {
          NavigableSet<String> parentIds = childValueToParentIds.lget();
          TermsEnum termsEnum = terms.iterator(null);
          PostingsEnum docsEnum = null;
          for (String id : parentIds) {
            TermsEnum.SeekStatus seekStatus =
                termsEnum.seekCeil(Uid.createUidAsBytes("parent", id));
            if (seekStatus == TermsEnum.SeekStatus.FOUND) {
              docsEnum =
                  termsEnum.postings(slowLeafReader.getLiveDocs(), docsEnum, PostingsEnum.NONE);
              expectedResult.set(docsEnum.nextDoc());
            } else if (seekStatus == TermsEnum.SeekStatus.END) {
              break;
            }
          }
        }
      }

      assertBitSet(actualResult, expectedResult, searcher);
    }

    indexWriter.close();
    indexReader.close();
    directory.close();
  }
コード例 #16
0
  /**
   * Returns a list of terms in the specified field along with the corresponding count of documents
   * in the set that match that constraint. This method uses the FilterCache to get the intersection
   * count between <code>docs</code> and the DocSet for each term in the filter.
   *
   * @see FacetParams#FACET_LIMIT
   * @see FacetParams#FACET_ZEROS
   * @see FacetParams#FACET_MISSING
   */
  public NamedList<Integer> getFacetTermEnumCounts(
      SolrIndexSearcher searcher,
      DocSet docs,
      String field,
      int offset,
      int limit,
      int mincount,
      boolean missing,
      String sort,
      String prefix,
      String contains,
      boolean ignoreCase,
      SolrParams params)
      throws IOException {

    /* :TODO: potential optimization...
     * cache the Terms with the highest docFreq and try them first
     * don't enum if we get our max from them
     */

    // Minimum term docFreq in order to use the filterCache for that term.
    int minDfFilterCache = global.getFieldInt(field, FacetParams.FACET_ENUM_CACHE_MINDF, 0);

    // make sure we have a set that is fast for random access, if we will use it for that
    DocSet fastForRandomSet = docs;
    if (minDfFilterCache > 0 && docs instanceof SortedIntDocSet) {
      SortedIntDocSet sset = (SortedIntDocSet) docs;
      fastForRandomSet = new HashDocSet(sset.getDocs(), 0, sset.size());
    }

    IndexSchema schema = searcher.getSchema();
    LeafReader r = searcher.getLeafReader();
    FieldType ft = schema.getFieldType(field);

    boolean sortByCount = sort.equals("count") || sort.equals("true");
    final int maxsize = limit >= 0 ? offset + limit : Integer.MAX_VALUE - 1;
    final BoundedTreeSet<CountPair<BytesRef, Integer>> queue =
        sortByCount ? new BoundedTreeSet<CountPair<BytesRef, Integer>>(maxsize) : null;
    final NamedList<Integer> res = new NamedList<>();

    int min = mincount - 1; // the smallest value in the top 'N' values
    int off = offset;
    int lim = limit >= 0 ? limit : Integer.MAX_VALUE;

    BytesRef prefixTermBytes = null;
    if (prefix != null) {
      String indexedPrefix = ft.toInternal(prefix);
      prefixTermBytes = new BytesRef(indexedPrefix);
    }

    Fields fields = r.fields();
    Terms terms = fields == null ? null : fields.terms(field);
    TermsEnum termsEnum = null;
    SolrIndexSearcher.DocsEnumState deState = null;
    BytesRef term = null;
    if (terms != null) {
      termsEnum = terms.iterator();

      // TODO: OPT: if seek(ord) is supported for this termsEnum, then we could use it for
      // facet.offset when sorting by index order.

      if (prefixTermBytes != null) {
        if (termsEnum.seekCeil(prefixTermBytes) == TermsEnum.SeekStatus.END) {
          termsEnum = null;
        } else {
          term = termsEnum.term();
        }
      } else {
        // position termsEnum on first term
        term = termsEnum.next();
      }
    }

    PostingsEnum postingsEnum = null;
    CharsRefBuilder charsRef = new CharsRefBuilder();

    if (docs.size() >= mincount) {
      while (term != null) {

        if (prefixTermBytes != null && !StringHelper.startsWith(term, prefixTermBytes)) break;

        if (contains == null || contains(term.utf8ToString(), contains, ignoreCase)) {
          int df = termsEnum.docFreq();

          // If we are sorting, we can use df>min (rather than >=) since we
          // are going in index order.  For certain term distributions this can
          // make a large difference (for example, many terms with df=1).
          if (df > 0 && df > min) {
            int c;

            if (df >= minDfFilterCache) {
              // use the filter cache

              if (deState == null) {
                deState = new SolrIndexSearcher.DocsEnumState();
                deState.fieldName = field;
                deState.liveDocs = r.getLiveDocs();
                deState.termsEnum = termsEnum;
                deState.postingsEnum = postingsEnum;
              }

              c = searcher.numDocs(docs, deState);

              postingsEnum = deState.postingsEnum;
            } else {
              // iterate over TermDocs to calculate the intersection

              // TODO: specialize when base docset is a bitset or hash set (skipDocs)?  or does it
              // matter for this?
              // TODO: do this per-segment for better efficiency (MultiDocsEnum just uses base class
              // impl)
              // TODO: would passing deleted docs lead to better efficiency over checking the
              // fastForRandomSet?
              postingsEnum = termsEnum.postings(postingsEnum, PostingsEnum.NONE);
              c = 0;

              if (postingsEnum instanceof MultiPostingsEnum) {
                MultiPostingsEnum.EnumWithSlice[] subs =
                    ((MultiPostingsEnum) postingsEnum).getSubs();
                int numSubs = ((MultiPostingsEnum) postingsEnum).getNumSubs();
                for (int subindex = 0; subindex < numSubs; subindex++) {
                  MultiPostingsEnum.EnumWithSlice sub = subs[subindex];
                  if (sub.postingsEnum == null) continue;
                  int base = sub.slice.start;
                  int docid;
                  while ((docid = sub.postingsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) {
                    if (fastForRandomSet.exists(docid + base)) c++;
                  }
                }
              } else {
                int docid;
                while ((docid = postingsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) {
                  if (fastForRandomSet.exists(docid)) c++;
                }
              }
            }

            if (sortByCount) {
              if (c > min) {
                BytesRef termCopy = BytesRef.deepCopyOf(term);
                queue.add(new CountPair<>(termCopy, c));
                if (queue.size() >= maxsize) min = queue.last().val;
              }
            } else {
              if (c >= mincount && --off < 0) {
                if (--lim < 0) break;
                ft.indexedToReadable(term, charsRef);
                res.add(charsRef.toString(), c);
              }
            }
          }
        }
        term = termsEnum.next();
      }
    }

    if (sortByCount) {
      for (CountPair<BytesRef, Integer> p : queue) {
        if (--off >= 0) continue;
        if (--lim < 0) break;
        ft.indexedToReadable(p.key, charsRef);
        res.add(charsRef.toString(), p.val);
      }
    }

    if (missing) {
      res.add(null, getFieldMissingCount(searcher, docs, field));
    }

    return res;
  }
コード例 #17
0
  @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();
  }