Package org.apache.lucene.store

Examples of org.apache.lucene.store.ByteArrayDataInput


      this.longsSize = longsSize;
      // if (DEBUG) {
      //   System.out.println("BTTR: seg=" + segment + " field=" + fieldInfo.name + " rootBlockCode=" + rootCode + " divisor=" + indexDivisor);
      // }

      rootBlockFP = (new ByteArrayDataInput(rootCode.bytes, rootCode.offset, rootCode.length)).readVLong() >>> BlockTreeTermsWriter.OUTPUT_FLAGS_NUM_BITS;

      if (indexIn != null) {
        final IndexInput clone = indexIn.clone();
        //System.out.println("start=" + indexStartFP + " field=" + fieldInfo.name);
        clone.seek(indexStartFP);
View Full Code Here


        
          // metadata
          numBytes = in.readVInt();
          if (bytes == null) {
            bytes = new byte[ArrayUtil.oversize(numBytes, 1)];
            bytesReader = new ByteArrayDataInput();
          } else if (bytes.length < numBytes) {
            bytes = new byte[ArrayUtil.oversize(numBytes, 1)];
          }
          in.readBytes(bytes, 0, numBytes);
          bytesReader.reset(bytes, 0, numBytes);
View Full Code Here

        // metadata
        len = in.readVInt();
        if (bytes == null) {
          bytes = new byte[ArrayUtil.oversize(len, 1)];
          bytesReader = new ByteArrayDataInput();
        } else if (bytes.length < len) {
          bytes = new byte[ArrayUtil.oversize(len, 1)];
        }
        in.readBytes(bytes, 0, len);
        bytesReader.reset(bytes, 0, len);
View Full Code Here

      // Build FST:
      BytesRef previousAnalyzed = null;
      BytesRef analyzed = new BytesRef();
      BytesRef surface = new BytesRef();
      IntsRef scratchInts = new IntsRef();
      ByteArrayDataInput input = new ByteArrayDataInput();

      // Used to remove duplicate surface forms (but we
      // still index the hightest-weight one).  We clear
      // this when we see a new analyzed form, so it cannot
      // grow unbounded (at most 256 entries):
      Set<BytesRef> seenSurfaceForms = new HashSet<BytesRef>();

      int dedup = 0;
      while (reader.read(scratch)) {
        input.reset(scratch.bytes, scratch.offset, scratch.length);
        short analyzedLength = input.readShort();
        analyzed.grow(analyzedLength+2);
        input.readBytes(analyzed.bytes, 0, analyzedLength);
        analyzed.length = analyzedLength;

        long cost = input.readInt();

        surface.bytes = scratch.bytes;
        if (hasPayloads) {
          surface.length = input.readShort();
          surface.offset = input.getPosition();
        } else {
          surface.offset = input.getPosition();
          surface.length = scratch.length - surface.offset;
        }
       
        if (previousAnalyzed == null) {
          previousAnalyzed = new BytesRef();
          previousAnalyzed.copyBytes(analyzed);
          seenSurfaceForms.add(BytesRef.deepCopyOf(surface));
        } else if (analyzed.equals(previousAnalyzed)) {
          dedup++;
          if (dedup >= maxSurfaceFormsPerAnalyzedForm) {
            // More than maxSurfaceFormsPerAnalyzedForm
            // dups: skip the rest:
            continue;
          }
          if (seenSurfaceForms.contains(surface)) {
            continue;
          }
          seenSurfaceForms.add(BytesRef.deepCopyOf(surface));
        } else {
          dedup = 0;
          previousAnalyzed.copyBytes(analyzed);
          seenSurfaceForms.clear();
          seenSurfaceForms.add(BytesRef.deepCopyOf(surface));
        }

        // TODO: I think we can avoid the extra 2 bytes when
        // there is no dup (dedup==0), but we'd have to fix
        // the exactFirst logic ... which would be sort of
        // hairy because we'd need to special case the two
        // (dup/not dup)...

        // NOTE: must be byte 0 so we sort before whatever
        // is next
        analyzed.bytes[analyzed.offset+analyzed.length] = 0;
        analyzed.bytes[analyzed.offset+analyzed.length+1] = (byte) dedup;
        analyzed.length += 2;

        Util.toIntsRef(analyzed, scratchInts);
        //System.out.println("ADD: " + scratchInts + " -> " + cost + ": " + surface.utf8ToString());
        if (!hasPayloads) {
          builder.add(scratchInts, outputs.newPair(cost, BytesRef.deepCopyOf(surface)));
        } else {
          int payloadOffset = input.getPosition() + surface.length;
          int payloadLength = scratch.length - payloadOffset;
          BytesRef br = new BytesRef(surface.length + 1 + payloadLength);
          System.arraycopy(surface.bytes, surface.offset, br.bytes, 0, surface.length);
          br.bytes[surface.length] = PAYLOAD_SEP;
          System.arraycopy(scratch.bytes, payloadOffset, br.bytes, surface.length+1, payloadLength);
View Full Code Here

              BytesRef output = Util.get(index, prefix);
              if (output == null) {
                out.println("      broken seek state: prefix is not final in index");
                throw new RuntimeException("seek state is broken");
              } else if (isSeekFrame && !f.isFloor) {
                final ByteArrayDataInput reader = new ByteArrayDataInput(output.bytes, output.offset, output.length);
                final long codeOrig = reader.readVLong();
                final long code = (f.fp << BlockTreeTermsWriter.OUTPUT_FLAGS_NUM_BITS) | (f.hasTerms ? BlockTreeTermsWriter.OUTPUT_FLAG_HAS_TERMS:0) | (f.isFloor ? BlockTreeTermsWriter.OUTPUT_FLAG_IS_FLOOR:0);
                if (codeOrig != code) {
                  out.println("      broken seek state: output code=" + codeOrig + " doesn't match frame code=" + code);
                  throw new RuntimeException("seek state is broken");
                }
View Full Code Here

          // that's rare so won't help much
          // metadata
          numBytes = in.readVInt();
          if (bytes == null) {
            bytes = new byte[ArrayUtil.oversize(numBytes, 1)];
            bytesReader = new ByteArrayDataInput();
          } else if (bytes.length < numBytes) {
            bytes = new byte[ArrayUtil.oversize(numBytes, 1)];
          }
          in.readBytes(bytes, 0, numBytes);
          bytesReader.reset(bytes, 0, numBytes);
View Full Code Here

      final int inputLines = info.lines;
      reader = new Sort.ByteSequencesReader(tempSorted);
      long line = 0;
      int previousBucket = 0;
      int previousScore = 0;
      ByteArrayDataInput input = new ByteArrayDataInput();
      BytesRef tmp1 = new BytesRef();
      BytesRef tmp2 = new BytesRef();
      while (reader.read(tmp1)) {
        input.reset(tmp1.bytes);
        int currentScore = input.readInt();

        int bucket;
        if (line > 0 && currentScore == previousScore) {
          bucket = previousBucket;
        } else {
          bucket = (int) (line * buckets / inputLines);
        }
        previousScore = currentScore;
        previousBucket = bucket;

        // Only append the input, discard the weight.
        tmp2.bytes = tmp1.bytes;
        tmp2.offset = input.getPosition();
        tmp2.length = tmp1.length - input.getPosition();
        builder.add(tmp2, bucket);

        line++;
        count++;
      }
View Full Code Here

      /** Decodes metadata into customized term state */
      abstract void decodeMetaData() throws IOException;

      BaseTermsEnum() throws IOException {
        this.state = postingsReader.newTermState();
        this.bytesReader = new ByteArrayDataInput();
        this.term = null;
        // NOTE: metadata will only be initialized in child class
      }
View Full Code Here

  public void readTermsBlock(IndexInput termsIn, FieldInfo fieldInfo, BlockTermState _termState) throws IOException {
    //System.out.println("PR.readTermsBlock state=" + _termState);
    final PulsingTermState termState = (PulsingTermState) _termState;
    if (termState.inlinedBytes == null) {
      termState.inlinedBytes = new byte[128];
      termState.inlinedBytesReader = new ByteArrayDataInput();
    }
    int len = termsIn.readVInt();
    //System.out.println("  len=" + len + " fp=" + termsIn.getFilePointer());
    if (termState.inlinedBytes.length < len) {
      termState.inlinedBytes = new byte[ArrayUtil.oversize(len, 1)];
View Full Code Here

      final int inputLines = info.lines;
      reader = new Sort.ByteSequencesReader(tempSorted);
      long line = 0;
      int previousBucket = 0;
      int previousScore = 0;
      ByteArrayDataInput input = new ByteArrayDataInput();
      BytesRef tmp1 = new BytesRef();
      BytesRef tmp2 = new BytesRef();
      while (reader.read(tmp1)) {
        input.reset(tmp1.bytes);
        int currentScore = input.readInt();

        int bucket;
        if (line > 0 && currentScore == previousScore) {
          bucket = previousBucket;
        } else {
          bucket = (int) (line * buckets / inputLines);
        }
        previousScore = currentScore;
        previousBucket = bucket;

        // Only append the input, discard the weight.
        tmp2.bytes = tmp1.bytes;
        tmp2.offset = input.getPosition();
        tmp2.length = tmp1.length - input.getPosition();
        builder.add(tmp2, bucket);

        line++;
      }

View Full Code Here

TOP

Related Classes of org.apache.lucene.store.ByteArrayDataInput

Copyright © 2018 www.massapicom. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.