Java VectorWrapper类代码示例

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
private void showSingleBatch (VectorAccessibleSerializable vcSerializable, boolean showHeader) {
  final VectorContainer vectorContainer = (VectorContainer) vcSerializable.get();

  /* show the header of the batch */
  if (showHeader) {
    System.out.println(getBatchMetaInfo(vcSerializable).toString());

    System.out.println("Schema Information");
    for (final VectorWrapper w : vectorContainer) {
      final MaterializedField field = w.getValueVector().getField();
      System.out.println (String.format("name : %s, minor_type : %s, data_mode : %s",
                                        field.toExpr(),
                                        field.getType().getMinorType().toString(),
                                        field.isNullable() ? "nullable":"non-nullable"
                        ));
    }
  }

  /* show the contents in the batch */
  VectorUtil.showVectorAccessibleContent(vectorContainer);
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
private BatchMetaInfo getBatchMetaInfo(VectorAccessibleSerializable vcSerializable) {
  final VectorAccessible vectorContainer = vcSerializable.get();

  int rows = 0;
  int selectedRows = 0;
  int totalDataSize = 0;

  rows = vectorContainer.getRecordCount();
  selectedRows = rows;

  if (vectorContainer.getSchema().getSelectionVectorMode() == SelectionVectorMode.TWO_BYTE) {
    selectedRows = vcSerializable.getSv2().getCount();
  }

  for (final VectorWrapper w : vectorContainer) {
     totalDataSize += w.getValueVector().getBufferSize();
  }

  return new BatchMetaInfo(rows, selectedRows, totalDataSize);
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
private void handleRemainder() {
  final int remainingRecordCount = incoming.getRecordCount() - remainderIndex;
  if (!doAlloc()) {
    outOfMemory = true;
    return;
  }
  final int projRecords = projector.projectRecords(remainderIndex, remainingRecordCount, 0);
  if (projRecords < remainingRecordCount) {
    setValueCount(projRecords);
    this.recordCount = projRecords;
    remainderIndex += projRecords;
  } else {
    setValueCount(remainingRecordCount);
    hasRemainder = false;
    remainderIndex = 0;
    for (final VectorWrapper<?> v : incoming) {
      v.clear();
    }
    this.recordCount = remainingRecordCount;
  }
  // In case of complex writer expression, vectors would be added to batch run-time.
  // We have to re-build the schema.
  if (complexWriters != null) {
    container.buildSchema(SelectionVectorMode.NONE);
  }
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
@Override
public IterOutcome innerNext() {
  if(!first && !noEndLimit && recordsLeft <= 0) {
    incoming.kill(true);

    IterOutcome upStream = next(incoming);
    if (upStream == IterOutcome.OUT_OF_MEMORY) {
      return upStream;
    }

    while (upStream == IterOutcome.OK || upStream == IterOutcome.OK_NEW_SCHEMA) {
      // Clear the memory for the incoming batch
      for (VectorWrapper<?> wrapper : incoming) {
        wrapper.getValueVector().clear();
      }
      upStream = next(incoming);
      if (upStream == IterOutcome.OUT_OF_MEMORY) {
        return upStream;
      }
    }

    return IterOutcome.NONE;
  }

  return super.innerNext();
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
@Override
public void buildSchema() throws SchemaChangeException {
  IterOutcome outcome = next(incoming);
  switch (outcome) {
    case NONE:
      state = BatchState.DONE;
      container.buildSchema(SelectionVectorMode.NONE);
      return;
    case OUT_OF_MEMORY:
      state = BatchState.OUT_OF_MEMORY;
      return;
    case STOP:
      state = BatchState.STOP;
      return;
  }

  if (!createAggregator()) {
    state = BatchState.DONE;
  }
  for (VectorWrapper w : container) {
    AllocationHelper.allocatePrecomputedChildCount(w.getValueVector(), 0, 0, 0);
  }
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
private void allocateOutgoing(int records) {
  // Skip the keys and only allocate for outputting the workspace values
  // (keys will be output through splitAndTransfer)
  Iterator<VectorWrapper<?>> outgoingIter = outContainer.iterator();
  for (int i = 0; i < numGroupByOutFields; i++) {
    outgoingIter.next();
  }
  while (outgoingIter.hasNext()) {
    ValueVector vv = outgoingIter.next().getValueVector();
    MajorType type = vv.getField().getType();

    /*
     * In build schema we use the allocation model that specifies exact record count
     * so we need to stick with that allocation model until DRILL-2211 is resolved. Using
     * 50 as the average bytes per value as is used in HashTable.
     */
    AllocationHelper.allocatePrecomputedChildCount(vv, records, VARIABLE_WIDTH_VALUE_SIZE, 0);
  }
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
public InternalBatch(RecordBatch incoming, VectorWrapper[] ignoreWrappers){
  switch(incoming.getSchema().getSelectionVectorMode()){
  case FOUR_BYTE:
    this.sv4 = incoming.getSelectionVector4().createNewWrapperCurrent();
    this.sv2 = null;
    break;
  case TWO_BYTE:
    this.sv4 = null;
    this.sv2 = incoming.getSelectionVector2().clone();
    break;
  default:
    this.sv4 = null;
    this.sv2 = null;
  }
  this.schema = incoming.getSchema();
  this.container = VectorContainer.getTransferClone(incoming, ignoreWrappers);
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
@Override
public void buildSchema() throws SchemaChangeException {
  IterOutcome outcome = next(incoming);
  switch (outcome) {
    case NONE:
      state = BatchState.DONE;
      container.buildSchema(SelectionVectorMode.NONE);
      return;
    case OUT_OF_MEMORY:
      state = BatchState.OUT_OF_MEMORY;
      return;
    case STOP:
      state = BatchState.STOP;
      return;
  }

  if (!createAggregator()) {
    state = BatchState.DONE;
  }
  for (final VectorWrapper<?> w : container) {
    w.getValueVector().allocateNew();
  }
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
private Copier getGenerated2Copier() throws SchemaChangeException{
  Preconditions.checkArgument(incoming.getSchema().getSelectionVectorMode() == SelectionVectorMode.TWO_BYTE);

  for(VectorWrapper<?> vv : incoming){
    TransferPair tp = vv.getValueVector().makeTransferPair(container.addOrGet(vv.getField(), callBack));
  }

  try {
    final CodeGenerator<Copier> cg = CodeGenerator.get(Copier.TEMPLATE_DEFINITION2, context.getFunctionRegistry());
    CopyUtil.generateCopies(cg.getRoot(), incoming, false);
    Copier copier = context.getImplementationClass(cg);
    copier.setupRemover(context, incoming, this);

    return copier;
  } catch (ClassTransformationException | IOException e) {
    throw new SchemaChangeException("Failure while attempting to load generated class", e);
  }
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
public static Copier getGenerated4Copier(RecordBatch batch, FragmentContext context, BufferAllocator allocator, VectorContainer container, RecordBatch outgoing, SchemaChangeCallBack callBack) throws SchemaChangeException{

    for(VectorWrapper<?> vv : batch){
      ValueVector v = vv.getValueVectors()[0];
      v.makeTransferPair(container.addOrGet(v.getField(), callBack));
    }

    try {
      final CodeGenerator<Copier> cg = CodeGenerator.get(Copier.TEMPLATE_DEFINITION4, context.getFunctionRegistry());
      CopyUtil.generateCopies(cg.getRoot(), batch, true);
      Copier copier = context.getImplementationClass(cg);
      copier.setupRemover(context, batch, outgoing);

      return copier;
    } catch (ClassTransformationException | IOException e) {
      throw new SchemaChangeException("Failure while attempting to load generated class", e);
    }
  }
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
@Override
public int copyRecords(int index, int recordCount){
  for(VectorWrapper<?> out : outgoing){
    MajorType type = out.getField().getType();
    if (!Types.isFixedWidthType(type) || Types.isRepeated(type)) {
      out.getValueVector().allocateNew();
    } else {
      AllocationHelper.allocate(out.getValueVector(), recordCount, 1);
    }
  }

  int outgoingPosition = 0;
  for(int svIndex = index; svIndex < index + recordCount; svIndex++, outgoingPosition++){
    int deRefIndex = sv4.get(svIndex);
    doEval(deRefIndex, outgoingPosition);
  }
  return outgoingPosition;
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
@Override
public int copyRecords(int index, int recordCount){
  for(VectorWrapper<?> out : outgoing){
    MajorType type = out.getField().getType();
    if (!Types.isFixedWidthType(type) || Types.isRepeated(type)) {
      out.getValueVector().allocateNew();
    } else {
      AllocationHelper.allocate(out.getValueVector(), recordCount, 1);
    }
  }

  int outgoingPosition = 0;

  for(int svIndex = index; svIndex < index + recordCount; svIndex++, outgoingPosition++){
    doEval(sv2.getIndex(svIndex), outgoingPosition);
  }
  return outgoingPosition;
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
public WindowDataBatch(final VectorAccessible batch, final OperatorContext oContext) {
  this.oContext = oContext;
  recordCount = batch.getRecordCount();

  List<ValueVector> vectors = Lists.newArrayList();

  for (VectorWrapper<?> v : batch) {
    if (v.isHyper()) {
      throw new UnsupportedOperationException("Record batch data can't be created based on a hyper batch.");
    }
    TransferPair tp = v.getValueVector().getTransferPair();
    tp.transfer();
    vectors.add(tp.getTo());
  }

  container = new VectorContainer(oContext);
  container.addCollection(vectors);
  container.setRecordCount(recordCount);
  container.buildSchema(batch.getSchema().getSelectionVectorMode());
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
public RecordBatchData(VectorAccessible batch) {
  List<ValueVector> vectors = Lists.newArrayList();
  recordCount = batch.getRecordCount();

  if (batch instanceof RecordBatch && batch.getSchema().getSelectionVectorMode() == SelectionVectorMode.TWO_BYTE) {
    this.sv2 = ((RecordBatch)batch).getSelectionVector2().clone();
  } else {
    this.sv2 = null;
  }

  for (VectorWrapper<?> v : batch) {
    if (v.isHyper()) {
      throw new UnsupportedOperationException("Record batch data can't be created based on a hyper batch.");
    }
    TransferPair tp = v.getValueVector().getTransferPair();
    tp.transfer();
    vectors.add(tp.getTo());
  }

  container.addCollection(vectors);
  container.setRecordCount(recordCount);
  container.buildSchema(batch.getSchema().getSelectionVectorMode());
}
 

import org.apache.drill.exec.record.VectorWrapper; //导入依赖的package包/类
private List<List<String>> getOutput(List<QueryDataBatch> batches) throws SchemaChangeException {
  List<List<String>> output = new ArrayList<>();
  RecordBatchLoader loader = new RecordBatchLoader(getAllocator());
  int last = 0;
  for(QueryDataBatch batch : batches) {
    int rows = batch.getHeader().getRowCount();
    if(batch.getData() != null) {
      loader.load(batch.getHeader().getDef(), batch.getData());
      // TODO:  Clean:  DRILL-2933:  That load(...) no longer throws
      // SchemaChangeException, so check/clean throws clause above.
      for (int i = 0; i < rows; ++i) {
        output.add(new ArrayList<String>());
        for (VectorWrapper<?> vw: loader) {
          ValueVector.Accessor accessor = vw.getValueVector().getAccessor();
          Object o = accessor.getObject(i);
          output.get(last).add(o == null ? null: o.toString());
        }
        ++last;
      }
    }
    loader.clear();
    batch.release();
  }
  return output;
}