Package org.apache.hadoop.hive.ql.exec

Source Code of org.apache.hadoop.hive.ql.exec.MapOperator

/**
* 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.hadoop.hive.ql.exec;

import java.io.Serializable;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.HashMap;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Properties;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hive.conf.HiveConf;
import org.apache.hadoop.hive.ql.io.IOContext;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.metadata.VirtualColumn;
import org.apache.hadoop.hive.ql.plan.api.OperatorType;
import org.apache.hadoop.hive.ql.plan.MapredWork;
import org.apache.hadoop.hive.ql.plan.PartitionDesc;
import org.apache.hadoop.hive.ql.plan.TableScanDesc;
import org.apache.hadoop.hive.serde2.Deserializer;
import org.apache.hadoop.hive.serde2.SerDeException;
import org.apache.hadoop.hive.serde2.SerDeUtils;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.PrimitiveObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.io.Writable;
import org.apache.hadoop.util.StringUtils;

/**
* Map operator. This triggers overall map side processing. This is a little
* different from regular operators in that it starts off by processing a
* Writable data structure from a Table (instead of a Hive Object).
**/
public class MapOperator extends Operator<MapredWork> implements Serializable {

  private static final long serialVersionUID = 1L;

  /**
   * Counter.
   *
   */
  public static enum Counter {
    DESERIALIZE_ERRORS
  }

  private final transient LongWritable deserialize_error_count = new LongWritable();
  private transient Deserializer deserializer;

  private transient Object[] rowWithPart;
  private transient Writable[] vcValues;
  private transient List<VirtualColumn> vcs;
  private transient Object[] rowWithPartAndVC;
  private transient StructObjectInspector rowObjectInspector;
  private transient boolean isPartitioned;
  private transient boolean hasVC;
  private Map<MapInputPath, MapOpCtx> opCtxMap;

  private Map<Operator<? extends Serializable>, java.util.ArrayList<String>> operatorToPaths;

  private final java.util.ArrayList<String> childrenPaths = new ArrayList<String>();

  private ArrayList<Operator<? extends Serializable>> extraChildrenToClose = null;

  private static class MapInputPath {
    String path;
    String alias;
    Operator<? extends Serializable> op;

    /**
     * @param path
     * @param alias
     * @param op
     */
    public MapInputPath(String path, String alias,
        Operator<? extends Serializable> op) {
      this.path = path;
      this.alias = alias;
      this.op = op;
    }

    @Override
    public boolean equals(Object o) {
      if (o instanceof MapInputPath) {
        MapInputPath mObj = (MapInputPath) o;
        if (mObj == null) {
          return false;
        }
        return path.equals(mObj.path) && alias.equals(mObj.alias)
            && op.equals(mObj.op);
      }

      return false;
    }

    @Override
    public int hashCode() {
      return (op == null) ? 0 : op.hashCode();
    }
  }

  private static class MapOpCtx {
    boolean isPartitioned;
    StructObjectInspector rawRowObjectInspector; //without partition
    StructObjectInspector partObjectInspector; // partition
    StructObjectInspector rowObjectInspector;
    Object[] rowWithPart;
    Deserializer deserializer;
    public String tableName;
    public String partName;

    /**
     * @param isPartitioned
     * @param rowObjectInspector
     * @param rowWithPart
     */
    public MapOpCtx(boolean isPartitioned,
        StructObjectInspector rowObjectInspector,
        StructObjectInspector rawRowObjectInspector,
        StructObjectInspector partObjectInspector,
        Object[] rowWithPart,
        Deserializer deserializer) {
      this.isPartitioned = isPartitioned;
      this.rowObjectInspector = rowObjectInspector;
      this.rawRowObjectInspector = rawRowObjectInspector;
      this.partObjectInspector = partObjectInspector;
      this.rowWithPart = rowWithPart;
      this.deserializer = deserializer;
    }

    /**
     * @return the isPartitioned
     */
    public boolean isPartitioned() {
      return isPartitioned;
    }

    /**
     * @return the rowObjectInspector
     */
    public StructObjectInspector getRowObjectInspector() {
      return rowObjectInspector;
    }

    /**
     * @return the rowWithPart
     */
    public Object[] getRowWithPart() {
      return rowWithPart;
    }

    /**
     * @return the deserializer
     */
    public Deserializer getDeserializer() {
      return deserializer;
    }
  }

  /**
   * Initializes this map op as the root of the tree. It sets JobConf &
   * MapRedWork and starts initialization of the operator tree rooted at this
   * op.
   *
   * @param hconf
   * @param mrwork
   * @throws HiveException
   */
  public void initializeAsRoot(Configuration hconf, MapredWork mrwork)
      throws HiveException {
    setConf(mrwork);
    setChildren(hconf);
    initialize(hconf, null);
  }

  private static MapOpCtx initObjectInspector(MapredWork conf,
      Configuration hconf, String onefile) throws HiveException,
      ClassNotFoundException, InstantiationException, IllegalAccessException,
      SerDeException {
    PartitionDesc td = conf.getPathToPartitionInfo().get(onefile);
    LinkedHashMap<String, String> partSpec = td.getPartSpec();
    Properties tblProps = td.getProperties();

    Class sdclass = td.getDeserializerClass();
    if (sdclass == null) {
      String className = td.getSerdeClassName();
      if ((className == "") || (className == null)) {
        throw new HiveException(
            "SerDe class or the SerDe class name is not set for table: "
            + td.getProperties().getProperty("name"));
      }
      sdclass = hconf.getClassByName(className);
    }

    String tableName = String.valueOf(tblProps.getProperty("name"));
    String partName = String.valueOf(partSpec);
    // HiveConf.setVar(hconf, HiveConf.ConfVars.HIVETABLENAME, tableName);
    // HiveConf.setVar(hconf, HiveConf.ConfVars.HIVEPARTITIONNAME, partName);
    Deserializer deserializer = (Deserializer) sdclass.newInstance();
    deserializer.initialize(hconf, tblProps);
    StructObjectInspector rawRowObjectInspector = (StructObjectInspector) deserializer
        .getObjectInspector();

    MapOpCtx opCtx = null;
    // Next check if this table has partitions and if so
    // get the list of partition names as well as allocate
    // the serdes for the partition columns
    String pcols = tblProps
        .getProperty(org.apache.hadoop.hive.metastore.api.Constants.META_TABLE_PARTITION_COLUMNS);
    // Log LOG = LogFactory.getLog(MapOperator.class.getName());
    if (pcols != null && pcols.length() > 0) {
      String[] partKeys = pcols.trim().split("/");
      List<String> partNames = new ArrayList<String>(partKeys.length);
      Object[] partValues = new Object[partKeys.length];
      List<ObjectInspector> partObjectInspectors = new ArrayList<ObjectInspector>(
          partKeys.length);
      for (int i = 0; i < partKeys.length; i++) {
        String key = partKeys[i];
        partNames.add(key);
        // Partitions do not exist for this table
        if (partSpec == null) {
          partValues[i] = new Text();
        } else {
          partValues[i] = new Text(partSpec.get(key));
        }
        partObjectInspectors
            .add(PrimitiveObjectInspectorFactory.writableStringObjectInspector);
      }
      StructObjectInspector partObjectInspector = ObjectInspectorFactory
          .getStandardStructObjectInspector(partNames, partObjectInspectors);

      Object[] rowWithPart = new Object[2];
      rowWithPart[1] = partValues;
      StructObjectInspector rowObjectInspector = ObjectInspectorFactory
          .getUnionStructObjectInspector(Arrays
          .asList(new StructObjectInspector[] {rawRowObjectInspector, partObjectInspector}));
      // LOG.info("dump " + tableName + " " + partName + " " +
      // rowObjectInspector.getTypeName());
      opCtx = new MapOpCtx(true, rowObjectInspector, rawRowObjectInspector ,partObjectInspector,rowWithPart, deserializer);
    } else {
      // LOG.info("dump2 " + tableName + " " + partName + " " +
      // rowObjectInspector.getTypeName());
      opCtx = new MapOpCtx(false, rawRowObjectInspector, rawRowObjectInspector, null, null, deserializer);
    }
    opCtx.tableName = tableName;
    opCtx.partName = partName;
    return opCtx;
  }

  public void setChildren(Configuration hconf) throws HiveException {

    Path fpath = new Path((new Path(HiveConf.getVar(hconf,
        HiveConf.ConfVars.HADOOPMAPFILENAME))).toUri().getPath());
    ArrayList<Operator<? extends Serializable>> children = new ArrayList<Operator<? extends Serializable>>();
    opCtxMap = new HashMap<MapInputPath, MapOpCtx>();
    operatorToPaths = new HashMap<Operator<? extends Serializable>, java.util.ArrayList<String>>();

    statsMap.put(Counter.DESERIALIZE_ERRORS, deserialize_error_count);

    try {
      boolean done = false;
      for (String onefile : conf.getPathToAliases().keySet()) {
        MapOpCtx opCtx = initObjectInspector(conf, hconf, onefile);
        Path onepath = new Path(new Path(onefile).toUri().getPath());
        List<String> aliases = conf.getPathToAliases().get(onefile);

        for (String onealias : aliases) {
          Operator<? extends Serializable> op = conf.getAliasToWork().get(
              onealias);
          LOG.info("Adding alias " + onealias + " to work list for file "
              + onefile);
          MapInputPath inp = new MapInputPath(onefile, onealias, op);
          opCtxMap.put(inp, opCtx);
          if (operatorToPaths.get(op) == null) {
            operatorToPaths.put(op, new java.util.ArrayList<String>());
          }
          operatorToPaths.get(op).add(onefile);

          op
              .setParentOperators(new ArrayList<Operator<? extends Serializable>>());
          op.getParentOperators().add(this);
          // check for the operators who will process rows coming to this Map
          // Operator
          if (!onepath.toUri().relativize(fpath.toUri()).equals(fpath.toUri())) {
            children.add(op);
            childrenPaths.add(onefile);
            LOG.info("dump " + op.getName() + " "
                + opCtxMap.get(inp).getRowObjectInspector().getTypeName());
            if (!done) {
              deserializer = opCtxMap.get(inp).getDeserializer();
              isPartitioned = opCtxMap.get(inp).isPartitioned();
              rowWithPart = opCtxMap.get(inp).getRowWithPart();
              rowObjectInspector = opCtxMap.get(inp).getRowObjectInspector();
              StructObjectInspector rawRowObjectInspector = opCtxMap.get(inp).rawRowObjectInspector;
              StructObjectInspector partObjectInspector = opCtxMap.get(inp).partObjectInspector;
              if (op instanceof TableScanOperator) {
                TableScanOperator tsOp = (TableScanOperator) op;
                TableScanDesc tsDesc = tsOp.getConf();
                if(tsDesc != null) {
                  this.vcs = tsDesc.getVirtualCols();
                  if (vcs != null && vcs.size() > 0) {
                    this.hasVC = true;
                    List<String> vcNames = new ArrayList<String>(vcs.size());
                    this.vcValues = new Writable[vcs.size()];
                    List<ObjectInspector> vcsObjectInspectors = new ArrayList<ObjectInspector>(vcs.size());
                    for (int i = 0; i < vcs.size(); i++) {
                      VirtualColumn vc = vcs.get(i);
                      vcsObjectInspectors.add(
                          PrimitiveObjectInspectorFactory.getPrimitiveWritableObjectInspector(
                              ((PrimitiveTypeInfo) vc.getTypeInfo()).getPrimitiveCategory()));
                      vcNames.add(vc.getName());
                    }
                    StructObjectInspector vcStructObjectInspector = ObjectInspectorFactory
                        .getStandardStructObjectInspector(vcNames,
                            vcsObjectInspectors);
                    if (isPartitioned) {
                      this.rowWithPartAndVC = new Object[3];
                      this.rowWithPartAndVC[1] = this.rowWithPart[1];
                    } else {
                      this.rowWithPartAndVC = new Object[2];
                    }
                    if(partObjectInspector == null) {
                      this.rowObjectInspector = ObjectInspectorFactory.getUnionStructObjectInspector(Arrays
                          .asList(new StructObjectInspector[] {
                              rowObjectInspector, vcStructObjectInspector }));
                    } else {
                      this.rowObjectInspector = ObjectInspectorFactory.getUnionStructObjectInspector(Arrays
                          .asList(new StructObjectInspector[] {
                              rawRowObjectInspector, partObjectInspector, vcStructObjectInspector }));
                    }
                    opCtxMap.get(inp).rowObjectInspector = this.rowObjectInspector;
                  }
                }
              }
              done = true;
            }
          }
        }
      }
      if (children.size() == 0) {
        // didn't find match for input file path in configuration!
        // serious problem ..
        LOG.error("Configuration does not have any alias for path: "
            + fpath.toUri().getPath());
        throw new HiveException("Configuration and input path are inconsistent");
      }

      // we found all the operators that we are supposed to process.
      setChildOperators(children);
    } catch (Exception e) {
      throw new HiveException(e);
    }
  }

  @Override
  public void initializeOp(Configuration hconf) throws HiveException {
    // set that parent initialization is done and call initialize on children
    state = State.INIT;
    List<Operator<? extends Serializable>> children = getChildOperators();

    for (Entry<MapInputPath, MapOpCtx> entry : opCtxMap.entrySet()) {
      // Add alias, table name, and partitions to hadoop conf so that their
      // children will
      // inherit these
      HiveConf.setVar(hconf, HiveConf.ConfVars.HIVETABLENAME,
          entry.getValue().tableName);
      HiveConf.setVar(hconf, HiveConf.ConfVars.HIVEPARTITIONNAME, entry
          .getValue().partName);
      MapInputPath input = entry.getKey();
      Operator<? extends Serializable> op = input.op;
      // op is not in the children list, so need to remember it and close it
      // afterwards
      if (children.indexOf(op) == -1) {
        if (extraChildrenToClose == null) {
          extraChildrenToClose = new ArrayList<Operator<? extends Serializable>>();
        }
        extraChildrenToClose.add(op);
      }

      // multiple input paths may corresponding the same operator (tree). The
      // below logic is to avoid initialize one operator multiple times if there
      // is one input path in this mapper's input paths.
      boolean shouldInit = true;
      List<String> paths = operatorToPaths.get(op);
      for (String path : paths) {
        if (childrenPaths.contains(path) && !path.equals(input.path)) {
          shouldInit = false;
          break;
        }
      }
      if (shouldInit) {
        op.initialize(hconf, new ObjectInspector[] {entry.getValue().getRowObjectInspector()});
      }
    }
  }

  /**
   * close extra child operators that are initialized but are not executed.
   */
  @Override
  public void closeOp(boolean abort) throws HiveException {
    if (extraChildrenToClose != null) {
      for (Operator<? extends Serializable> op : extraChildrenToClose) {
        op.close(abort);
      }
    }
  }

  public void process(Writable value) throws HiveException {
    Object row = null;
    try {
      if (this.hasVC) {
        this.rowWithPartAndVC[0] = deserializer.deserialize(value);
        int vcPos = isPartitioned ? 2 : 1;
        populateVirtualColumnValues();
        this.rowWithPartAndVC[vcPos] = this.vcValues;
      } else if (!isPartitioned) {
        row = deserializer.deserialize((Writable)value);
      } else {
        rowWithPart[0] = deserializer.deserialize((Writable)value);
      }
    } catch (Exception e) {
      // Serialize the row and output.
      String rawRowString;
      try {
        rawRowString = value.toString();
      } catch (Exception e2) {
        rawRowString = "[Error getting row data with exception " +
            StringUtils.stringifyException(e2) + " ]";
      }

      // TODO: policy on deserialization errors
      deserialize_error_count.set(deserialize_error_count.get() + 1);
      throw new HiveException("Hive Runtime Error while processing writable " + rawRowString, e);
    }

    try {
      if (this.hasVC) {
        forward(this.rowWithPartAndVC, this.rowObjectInspector);
      } else if (!isPartitioned) {
        forward(row, rowObjectInspector);
      } else {
        forward(rowWithPart, rowObjectInspector);
      }
    } catch (Exception e) {
      // Serialize the row and output the error message.
      String rowString;
      try {
        if (this.hasVC) {
          rowString = SerDeUtils.getJSONString(rowWithPartAndVC, rowObjectInspector);
        } else if (!isPartitioned) {
          rowString = SerDeUtils.getJSONString(row, rowObjectInspector);
        } else {
          rowString = SerDeUtils.getJSONString(rowWithPart, rowObjectInspector);
        }
      } catch (Exception e2) {
        rowString = "[Error getting row data with exception " +
            StringUtils.stringifyException(e2) + " ]";
      }
      throw new HiveException("Hive Runtime Error while processing row " + rowString, e);
    }
  }

  private void populateVirtualColumnValues() {
    if (this.vcs != null) {
      ExecMapperContext mapExecCxt = this.getExecContext();
      IOContext ioCxt = mapExecCxt.getIoCxt();
      for (int i = 0; i < vcs.size(); i++) {
        VirtualColumn vc = vcs.get(i);
        if (vc.equals(VirtualColumn.FILENAME) && mapExecCxt.inputFileChanged()) {
          this.vcValues[i] = new Text(mapExecCxt.getCurrentInputFile());
        } else if (vc.equals(VirtualColumn.BLOCKOFFSET)) {
          long current = ioCxt.getCurrentBlockStart();
          LongWritable old = (LongWritable) this.vcValues[i];
          if (old == null) {
            old = new LongWritable(current);
            this.vcValues[i] = old;
            continue;
          }
          if (current != old.get()) {
            old.set(current);
          }
        }
      }
    }
  }

  @Override
  public void processOp(Object row, int tag) throws HiveException {
    throw new HiveException("Hive 2 Internal error: should not be called!");
  }

  @Override
  public String getName() {
    return "MAP";
  }

  @Override
  public OperatorType getType() {
    return null;
  }

}
TOP

Related Classes of org.apache.hadoop.hive.ql.exec.MapOperator

TOP
Copyright © 2018 www.massapi.com. 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.