exec/java-exec/src/main/java/org/apache/drill/exec/physical/impl/join/RowKeyJoinBatch.java

/*
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.apache.drill.exec.physical.impl.join;


import java.util.List;

import org.apache.commons.lang3.tuple.Pair;
import org.apache.drill.exec.exception.OutOfMemoryException;
import org.apache.drill.exec.exception.SchemaChangeException;
import org.apache.drill.exec.ops.FragmentContext;
import org.apache.drill.exec.physical.config.RowKeyJoinPOP;
import org.apache.drill.exec.record.AbstractRecordBatch;
import org.apache.drill.exec.record.RecordBatch;
import org.apache.drill.exec.record.TransferPair;
import org.apache.drill.exec.record.VectorWrapper;
import org.apache.drill.exec.record.selection.SelectionVector2;
import org.apache.drill.exec.record.selection.SelectionVector4;
import org.apache.drill.exec.vector.SchemaChangeCallBack;
import org.apache.drill.exec.vector.ValueVector;

import org.apache.drill.shaded.guava.com.google.common.collect.Iterables;
import org.apache.drill.shaded.guava.com.google.common.collect.Lists;


public class RowKeyJoinBatch extends AbstractRecordBatch<RowKeyJoinPOP> implements RowKeyJoin {
  private static final org.slf4j.Logger logger = org.slf4j.LoggerFactory.getLogger(RowKeyJoinBatch.class);

  // primary table side record batch
  private final RecordBatch left;

  // index table side record batch
  private final RecordBatch right;

  private boolean hasRowKeyBatch;
  private IterOutcome leftUpstream = IterOutcome.NONE;
  private IterOutcome rightUpstream = IterOutcome.NONE;
  private final List<TransferPair> transfers = Lists.newArrayList();
  private int recordCount = 0;
  private final SchemaChangeCallBack callBack = new SchemaChangeCallBack();
  private RowKeyJoinState rkJoinState = RowKeyJoinState.INITIAL;

  public RowKeyJoinBatch(RowKeyJoinPOP config, FragmentContext context, RecordBatch left, RecordBatch right)
      throws OutOfMemoryException {
    super(config, context, true /* need to build schema */);
    this.left = left;
    this.right = right;
    this.hasRowKeyBatch = false;
  }

  @Override
  public int getRecordCount() {
    if (state == BatchState.DONE) {
      return 0;
    }
    return recordCount;
  }

  @Override
  public SelectionVector2 getSelectionVector2() {
    throw new UnsupportedOperationException("RowKeyJoinBatch does not support selection vector");
  }

  @Override
  public SelectionVector4 getSelectionVector4() {
    throw new UnsupportedOperationException("RowKeyJoinBatch does not support selection vector");
  }

  @Override
  protected void buildSchema() throws SchemaChangeException {
    container.clear();

    rightUpstream = next(right);

    if (leftUpstream == IterOutcome.STOP || rightUpstream == IterOutcome.STOP) {
      state = BatchState.STOP;
      return;
    }

    if (right.getRecordCount() > 0) {
      // set the hasRowKeyBatch flag such that calling next() on the left input
      // would see the correct status
      hasRowKeyBatch = true;
    }

    leftUpstream = next(left);

    for (final VectorWrapper<?> v : left) {
      final TransferPair pair = v.getValueVector().makeTransferPair(
          container.addOrGet(v.getField(), callBack));
      transfers.add(pair);
    }

    container.buildSchema(left.getSchema().getSelectionVectorMode());
  }

  @Override
  public IterOutcome innerNext() {
    if (state == BatchState.DONE) {
      return IterOutcome.NONE;
    }
    try {
      if (state == BatchState.FIRST && left.getRecordCount() > 0) {
        logger.debug("First batch, outputting the batch with {} records.", left.getRecordCount());
        // there is already a pending batch from left, output it
        outputCurrentLeftBatch();
        // Check if schema has changed (this is just to guard against potential changes to the
        // output schema by outputCurrentLeftBatch.
        if (callBack.getSchemaChangedAndReset()) {
          return IterOutcome.OK_NEW_SCHEMA;
        }
        return IterOutcome.OK;
      }

      if (rightUpstream == IterOutcome.NONE) {
        rkJoinState = RowKeyJoinState.DONE;
        state = BatchState.DONE;
        return rightUpstream;
      }
      rightUpstream = next(right);

      logger.debug("right input IterOutcome: {}", rightUpstream);

      switch(rightUpstream) {
      case NONE:
      case STOP:
        rkJoinState = RowKeyJoinState.DONE;
        state = BatchState.DONE;
        return rightUpstream;
      case OK_NEW_SCHEMA:
      case OK:
        // we got a new batch from the right input, set this flag
        // such that subsequent check by a scan would indicate availability
        // of the row keys.
        while ((rightUpstream == IterOutcome.OK || rightUpstream == IterOutcome.OK_NEW_SCHEMA) &&
            right.getRecordCount() == 0) {
          rightUpstream = next(right);
          logger.trace("rowkeyjoin loop when recordCount == 0. rightUpstream {}", rightUpstream);
        }

        if (!hasRowKeyBatch && right.getRecordCount() > 0) {
          hasRowKeyBatch = true;
        }

        logger.debug("right input num records = {}", right.getRecordCount());

        if (hasRowKeyBatch) {
          // get the next batch from left input
          leftUpstream = next(left);

          logger.debug("left input IterOutcome: {}", leftUpstream);

          if (leftUpstream == IterOutcome.OK || leftUpstream == IterOutcome.OK_NEW_SCHEMA) {
            logger.debug("left input num records = {}", left.getRecordCount());
            if (left.getRecordCount() > 0) {
              logger.debug("Outputting the left batch with {} records.", left.getRecordCount());
              outputCurrentLeftBatch();
              // Check if schema has changed (this is just to guard against potential changes to the
              // output schema by outputCurrentLeftBatch, but in general the leftUpstream status should
              // be sufficient)
              if (callBack.getSchemaChangedAndReset()) {
                return IterOutcome.OK_NEW_SCHEMA;
              }
            }
          }
        }

        if (leftUpstream == IterOutcome.NONE) {
          container.setRecordCount(0);
          this.recordCount = 0;
          return rightUpstream;
        } else {
          return leftUpstream;
        }

      default:
        throw new IllegalStateException(String.format("Unknown state %s.", rightUpstream));
      }
    } finally {
      if (state == BatchState.FIRST) {
        state = BatchState.NOT_FIRST;
      }
      if (leftUpstream == IterOutcome.NONE && rkJoinState == RowKeyJoinState.PROCESSING) {
        rkJoinState = RowKeyJoinState.INITIAL;
      }
    }
  }

  private void outputCurrentLeftBatch() {
    // Schema change when state is FIRST shouldn't happen as buildSchema should
    // take care of building the schema for the first batch. This check is introduced
    // to guard against any schema change after buildSchema phase and reading
    // the first batch of rows.
    if (leftUpstream == IterOutcome.OK_NEW_SCHEMA && state == BatchState.FIRST ||
        state == BatchState.NOT_FIRST) {
      container.zeroVectors();
      transfers.clear();

      for (final VectorWrapper<?> v : left) {
        final TransferPair pair = v.getValueVector().makeTransferPair(
            container.addOrGet(v.getField(), callBack));
        transfers.add(pair);
      }

      if (container.isSchemaChanged()) {
        container.buildSchema(left.getSchema().getSelectionVectorMode());
      }
    }

    for (TransferPair t : transfers) {
      t.transfer();
    }

    container.setRecordCount(left.getRecordCount());
    this.recordCount = left.getRecordCount();
  }

  @Override  // implement RowKeyJoin interface
  public boolean hasRowKeyBatch() {
    return hasRowKeyBatch;
  }

  @Override  // implement RowKeyJoin interface
  public Pair<ValueVector, Integer> nextRowKeyBatch() {
    if (hasRowKeyBatch && right.getRecordCount() > 0 ) {
      // since entire right row key batch will be returned to the caller, reset
      // the hasRowKeyBatch to false
      hasRowKeyBatch = false;
      VectorWrapper<?> vw = Iterables.get(right, 0);
      ValueVector vv = vw.getValueVector();
      return Pair.of(vv, right.getRecordCount()-1);
    }
    return null;
  }

  @Override   // implement RowKeyJoin interface
  public BatchState getBatchState() {
    return state;
  }

  @Override  // implement RowKeyJoin interface
  public void setBatchState(BatchState newState) {
    state = newState;
  }

  @Override
  public void setRowKeyJoinState(RowKeyJoinState newState) {
    this.rkJoinState = newState;
  }

  @Override
  public RowKeyJoinState getRowKeyJoinState() {
    return rkJoinState;
  }

  @Override
  public void killIncoming(boolean sendUpstream) {
    left.kill(sendUpstream);
    right.kill(sendUpstream);
  }

  @Override
  public void close() {
    rkJoinState = RowKeyJoinState.DONE;
    super.close();
  }

  @Override
  public void dump() {
    logger.error("RowKeyJoinBatch[container={}, left={}, right={}, hasRowKeyBatch={}, rkJoinState={}]",
        container, left, right, hasRowKeyBatch, rkJoinState);
  }

}