Package org.apache.hadoop.hive.ql.plan

Examples of org.apache.hadoop.hive.ql.plan.MapredWork


      String hiveConfArgs = generateCmdLine(conf);

      // write out the plan to a local file
      Path planPath = new Path(ctx.getLocalTmpFileURI(), "plan.xml");
      OutputStream out = FileSystem.getLocal(conf).create(planPath);
      MapredWork plan = getWork();
      LOG.info("Generating plan file " + planPath.toString());
      Utilities.serializeMapRedWork(plan, out);

      String isSilent = "true".equalsIgnoreCase(System
          .getProperty("test.silent")) ? "-nolog" : "";
View Full Code Here


      throws SemanticException {
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = opProcCtx.getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    MapredWork plan = (MapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap =
      opProcCtx.getOpTaskMap();
    Operator<? extends Serializable> currTopOp = opProcCtx.getCurrTopOp();

    opTaskMap.put(reducer, currTask);
    plan.setReducer(reducer);
    ReduceSinkDesc desc = op.getConf();

    plan.setNumReduceTasks(desc.getNumReducers());

    List<Task<? extends Serializable>> rootTasks = opProcCtx.getRootTasks();

    if (!rootTasks.contains(currTask)) {
        rootTasks.add(currTask);
    }
    if (reducer.getClass() == JoinOperator.class) {
      plan.setNeedsTagging(true);
    }

    assert currTopOp != null;
    List<Operator<? extends Serializable>> seenOps = opProcCtx.getSeenOps();
    String currAliasId = opProcCtx.getCurrAliasId();
View Full Code Here

    assert (((pos == -1) && (readInputMapJoin)) || (pos != -1));
    int parentPos = (pos == -1) ? 0 : pos;
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(
        parentPos));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    MapredWork plan = (MapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>>  opTaskMap =
      opProcCtx.getOpTaskMap();
    Operator<? extends Serializable> currTopOp = opProcCtx.getCurrTopOp();

    // The mapjoin has already been encountered. Some context must be stored
    // about that
    if (readInputMapJoin) {
      AbstractMapJoinOperator<? extends MapJoinDesc> currMapJoinOp = opProcCtx.getCurrMapJoinOp();
      assert currMapJoinOp != null;
      boolean local = ((pos == -1) || (pos == (currMapJoinOp.getConf()).getPosBigTable())) ?
          false : true;

      if (setReducer) {
        Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
        plan.setReducer(reducer);
        opTaskMap.put(reducer, currTask);
        if (reducer.getClass() == JoinOperator.class) {
          plan.setNeedsTagging(true);
        }
        ReduceSinkDesc desc = (ReduceSinkDesc) op.getConf();
        plan.setNumReduceTasks(desc.getNumReducers());
      } else {
        opTaskMap.put(op, currTask);
      }

      if (!readInputUnion) {
View Full Code Here

      GenMRProcContext opProcCtx) throws SemanticException {
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = opProcCtx.getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    MapredWork plan = (MapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap =
      opProcCtx.getOpTaskMap();

    opTaskMap.put(reducer, currTask);
    plan.setReducer(reducer);
    ReduceSinkDesc desc = op.getConf();

    plan.setNumReduceTasks(desc.getNumReducers());

    if (reducer.getClass() == JoinOperator.class) {
      plan.setNeedsTagging(true);
    }

    initUnionPlan(opProcCtx, currTask, false);
  }
View Full Code Here

   * It is a idempotent function to add various intermediate files as the source
   * for the union. The plan has already been created.
   */
  public static void initUnionPlan(GenMRProcContext opProcCtx,
      Task<? extends Serializable> currTask, boolean local) {
    MapredWork plan = (MapredWork) currTask.getWork();
    UnionOperator currUnionOp = opProcCtx.getCurrUnionOp();
    assert currUnionOp != null;
    GenMRUnionCtx uCtx = opProcCtx.getUnionTask(currUnionOp);
    assert uCtx != null;

    List<String> taskTmpDirLst = uCtx.getTaskTmpDir();
    List<TableDesc> tt_descLst = uCtx.getTTDesc();
    assert !taskTmpDirLst.isEmpty() && !tt_descLst.isEmpty();
    assert taskTmpDirLst.size() == tt_descLst.size();
    int size = taskTmpDirLst.size();
    assert local == false;

    for (int pos = 0; pos < size; pos++) {
      String taskTmpDir = taskTmpDirLst.get(pos);
      TableDesc tt_desc = tt_descLst.get(pos);
      if (plan.getPathToAliases().get(taskTmpDir) == null) {
        plan.getPathToAliases().put(taskTmpDir, new ArrayList<String>());
        plan.getPathToAliases().get(taskTmpDir).add(taskTmpDir);
        plan.getPathToPartitionInfo().put(taskTmpDir,
            new PartitionDesc(tt_desc, null));
        plan.getAliasToWork().put(taskTmpDir, currUnionOp);
      }
    }
  }
View Full Code Here

      Task<? extends Serializable> oldTask, Task<? extends Serializable> task,
      GenMRProcContext opProcCtx, int pos, boolean split,
      boolean readMapJoinData, boolean readUnionData, boolean createLocalWork)
      throws SemanticException {
    Task<? extends Serializable> currTask = task;
    MapredWork plan = (MapredWork) currTask.getWork();
    Operator<? extends Serializable> currTopOp = opProcCtx.getCurrTopOp();
    List<Task<? extends Serializable>> parTasks = null;

    // terminate the old task and make current task dependent on it
    if (split) {
View Full Code Here

   */
  public static void splitPlan(ReduceSinkOperator op, GenMRProcContext opProcCtx)
  throws SemanticException {
    // Generate a new task
    ParseContext parseCtx = opProcCtx.getParseCtx();
    MapredWork cplan = getMapRedWork(parseCtx.getConf());
    Task<? extends Serializable> redTask = TaskFactory.get(cplan, parseCtx
        .getConf());
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);

    // Add the reducer
    cplan.setReducer(reducer);
    ReduceSinkDesc desc = op.getConf();

    cplan.setNumReduceTasks(new Integer(desc.getNumReducers()));

    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap =
      opProcCtx.getOpTaskMap();
    opTaskMap.put(reducer, redTask);
    Task<? extends Serializable> currTask = opProcCtx.getCurrTask();
View Full Code Here

   * create a new plan and return.
   *
   * @return the new plan
   */
  public static MapredWork getMapRedWork(HiveConf conf) {
    MapredWork work = new MapredWork();
    work.setPathToAliases(new LinkedHashMap<String, ArrayList<String>>());
    work.setPathToPartitionInfo(new LinkedHashMap<String, PartitionDesc>());
    work.setAliasToWork(new LinkedHashMap<String, Operator<? extends Serializable>>());
    work.setTagToValueDesc(new ArrayList<TableDesc>());
    work.setReducer(null);
    work.setHadoopSupportsSplittable(
        conf.getBoolVar(HiveConf.ConfVars.HIVE_COMBINE_INPUT_FORMAT_SUPPORTS_SPLITTABLE));
    return work;
  }
View Full Code Here

    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = opProcCtx.getMapCurrCtx();
    mapCurrCtx.put(ts_op, new GenMapRedCtx(childTask, null, null));

    String streamDesc = taskTmpDir;
    MapredWork cplan = (MapredWork) childTask.getWork();

    if (setReducer) {
      Operator<? extends Serializable> reducer = op.getChildOperators().get(0);

      if (reducer.getClass() == JoinOperator.class) {
        String origStreamDesc;
        streamDesc = "$INTNAME";
        origStreamDesc = streamDesc;
        int pos = 0;
        while (cplan.getAliasToWork().get(streamDesc) != null) {
          streamDesc = origStreamDesc.concat(String.valueOf(++pos));
        }
      }

      // TODO: Allocate work to remove the temporary files and make that
      // dependent on the redTask
      if (reducer.getClass() == JoinOperator.class) {
        cplan.setNeedsTagging(true);
      }
    }

    // Add the path to alias mapping
    setTaskPlan(taskTmpDir, streamDesc, ts_op, cplan, local, tt_desc);
View Full Code Here

    GenMRUnionCtx uCtxTask = ctx.getUnionTask(union);
    Task<? extends Serializable> uTask = null;

    union.getParentOperators().get(pos);
    MapredWork uPlan = null;

    // union is encountered for the first time
    if (uCtxTask == null) {
      uCtxTask = new GenMRUnionCtx();
      uPlan = GenMapRedUtils.getMapRedWork(parseCtx.getConf());
      uTask = TaskFactory.get(uPlan, parseCtx.getConf());
      uCtxTask.setUTask(uTask);
      ctx.setUnionTask(union, uCtxTask);
    } else {
      uTask = uCtxTask.getUTask();
      uPlan = (MapredWork) uTask.getWork();
    }

    // If there is a mapjoin at position 'pos'
    if (uPrsCtx.getMapJoinSubq(pos)) {
      GenMRMapJoinCtx mjCtx = ctx.getMapJoinCtx(ctx.getCurrMapJoinOp());
      String taskTmpDir = mjCtx.getTaskTmpDir();
      if (uPlan.getPathToAliases().get(taskTmpDir) == null) {
        uPlan.getPathToAliases().put(taskTmpDir, new ArrayList<String>());
        uPlan.getPathToAliases().get(taskTmpDir).add(taskTmpDir);
        uPlan.getPathToPartitionInfo().put(taskTmpDir,
            new PartitionDesc(mjCtx.getTTDesc(), null));
        uPlan.getAliasToWork().put(taskTmpDir, mjCtx.getRootMapJoinOp());
      }

      for (Task t : currTask.getParentTasks()) {
        t.addDependentTask(uTask);
      }
View Full Code Here

TOP

Related Classes of org.apache.hadoop.hive.ql.plan.MapredWork

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.