Examples of UnionParseContext


Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      UnionOperator union = (UnionOperator) nd;
      UnionProcContext ctx = (UnionProcContext) procCtx;

      // find the branch on which this processor was invoked
      int pos = getPositionParent(union, stack);
      UnionParseContext uCtx = ctx.getUnionParseContext(union);
      if (uCtx == null) {
        uCtx = new UnionParseContext(union.getConf().getNumInputs());
      }

      uCtx.setMapOnlySubq(pos, true);
      uCtx.setRootTask(pos, true);
      ctx.setUnionParseContext(union, uCtx);
      return null;
    }
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      UnionOperator union = (UnionOperator) nd;
      UnionProcContext ctx = (UnionProcContext) procCtx;

      // find the branch on which this processor was invoked
      int pos = getPositionParent(union, stack);
      UnionParseContext uCtx = ctx.getUnionParseContext(union);
      if (uCtx == null) {
        uCtx = new UnionParseContext(union.getConf().getNumInputs());
      }

      uCtx.setMapJoinSubq(pos, true);
      ctx.setUnionParseContext(union, uCtx);
      return null;
    }
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      UnionOperator union = (UnionOperator) nd;
      UnionProcContext ctx = (UnionProcContext) procCtx;

      // find the branch on which this processor was invoked
      int pos = getPositionParent(union, stack);
      UnionParseContext uCtx = ctx.getUnionParseContext(union);
      if (uCtx == null) {
        uCtx = new UnionParseContext(union.getConf().getNumInputs());
      }

      uCtx.setMapOnlySubq(pos, true);
      uCtx.setRootTask(pos, false);
      ctx.setUnionParseContext(union, uCtx);
      return null;
    }
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

        }
        ctx.setCurrTask(uTask);
      }
    }

    UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);
    ctx.getMapCurrCtx().put(
        (Operator<? extends OperatorDesc>) union,
        new GenMapRedCtx(ctx.getCurrTask(),
            ctx.getCurrAliasId()));
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      mapCurrCtx.put((Operator<? extends OperatorDesc>) nd,
        new GenMapRedCtx(ctx.getCurrTask(),ctx.getCurrAliasId()));
      return null;
    }

    UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);

    ctx.setCurrUnionOp(union);
    // The plan needs to be broken only if one of the sub-queries involve a
    // map-reduce job
    if (uPrsCtx.allMapOnlySubQ()) {
      return processMapOnlyUnion(union, stack, ctx, uCtx);
    }

    assert uPrsCtx != null;

    Task<? extends Serializable> currTask = ctx.getCurrTask();
    int pos = UnionProcFactory.getPositionParent(union, stack);

    Task<? extends Serializable> uTask = null;
    MapredWork uPlan = null;

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

    // Copy into the current union task plan if
    if (uPrsCtx.getMapOnlySubq(pos) && uPrsCtx.getRootTask(pos)) {
      processSubQueryUnionMerge(ctx, uCtxTask, union, stack);
      if (ctx.getRootTasks().contains(currTask)) {
        ctx.getRootTasks().remove(currTask);
      }
    }
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      UnionOperator union = (UnionOperator) nd;
      UnionProcContext ctx = (UnionProcContext) procCtx;

      // find the branch on which this processor was invoked
      int pos = getPositionParent(union, stack);
      UnionParseContext uCtx = ctx.getUnionParseContext(union);
      if (uCtx == null) {
        uCtx = new UnionParseContext(union.getConf().getNumInputs());
      }
      ctx.setMapOnlySubq(false);
      uCtx.setMapOnlySubq(pos, false);
      uCtx.setRootTask(pos, false);
      ctx.setUnionParseContext(union, uCtx);
      return null;
    }
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      UnionOperator union = (UnionOperator) nd;
      UnionProcContext ctx = (UnionProcContext) procCtx;

      // find the branch on which this processor was invoked
      int pos = getPositionParent(union, stack);
      UnionParseContext uCtx = ctx.getUnionParseContext(union);
      if (uCtx == null) {
        uCtx = new UnionParseContext(union.getConf().getNumInputs());
      }

      uCtx.setMapOnlySubq(pos, true);
      uCtx.setRootTask(pos, true);
      ctx.setUnionParseContext(union, uCtx);
      return null;
    }
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

      UnionOperator union = (UnionOperator) nd;
      UnionProcContext ctx = (UnionProcContext) procCtx;

      // find the branch on which this processor was invoked
      int pos = getPositionParent(union, stack);
      UnionParseContext uCtx = ctx.getUnionParseContext(union);
      if (uCtx == null) {
        uCtx = new UnionParseContext(union.getConf().getNumInputs());
      }
      int start = stack.size() - 2;
      UnionOperator parentUnionOperator = null;
      while (start >= 0) {
        Operator<? extends OperatorDesc> parent =
          (Operator<? extends OperatorDesc>) stack.get(start);
        if (parent instanceof UnionOperator) {
          parentUnionOperator = (UnionOperator) parent;
          break;
        }
        start--;
      }
      assert parentUnionOperator != null;

      // default to false
      boolean mapOnly = false;
      boolean rootTask = false;
      UnionParseContext parentUCtx =
          ctx.getUnionParseContext(parentUnionOperator);
      if (parentUCtx != null && parentUCtx.allMapOnlySubQSet()) {
        mapOnly = parentUCtx.allMapOnlySubQ();
        rootTask = parentUCtx.allRootTasks();
      }

      uCtx.setMapOnlySubq(pos, mapOnly);

      uCtx.setRootTask(pos, rootTask);
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

        }
        pos--;
      }

      UnionProcContext ctx = (UnionProcContext) procCtx;
      UnionParseContext uCtx = ctx.getUnionParseContext(union);

      // No need for this if all sub-queries are map-only queries
      // If all the queries are map-only, anyway the query is most optimized
      if ((uCtx != null) && (uCtx.allMapOnlySubQ())) {
        return null;
      }

      pos++;
      pushOperatorsAboveUnion(union, stack, pos);
View Full Code Here

Examples of org.apache.hadoop.hive.ql.optimizer.unionproc.UnionProcContext.UnionParseContext

        if (tsk != null) {
          ctx.setCurrTask(tsk);
        }
      }

      UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);
      if ((uPrsCtx != null) && (uPrsCtx.getMapJoinQuery())) {
        GenMapRedUtils.mergeMapJoinUnion(union, ctx, UnionProcFactory
            .getPositionParent(union, stack));
      } else {
        mapCurrCtx.put((Operator<? extends Serializable>) nd, new GenMapRedCtx(
            ctx.getCurrTask(), ctx.getCurrTopOp(), ctx.getCurrAliasId()));
      }
      return null;
    }

    ctx.setCurrUnionOp(union);

    UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);
    assert uPrsCtx != null;

    Task<? extends Serializable> currTask = ctx.getCurrTask();
    int pos = UnionProcFactory.getPositionParent(union, stack);

    // is the current task a root task
    if (uPrsCtx.getRootTask(pos) && (!ctx.getRootTasks().contains(currTask))) {
      ctx.getRootTasks().add(currTask);
    }

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

    Operator<? extends Serializable> parent = 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)) {
      AbstractMapJoinOperator<? extends MapJoinDesc> mjOp = ctx.getCurrMapJoinOp();
      assert mjOp != null;
      GenMRMapJoinCtx mjCtx = ctx.getMapJoinCtx(mjOp);
      assert mjCtx != null;
      MapredWork plan = (MapredWork) currTask.getWork();

      String taskTmpDir = mjCtx.getTaskTmpDir();
      TableDesc tt_desc = mjCtx.getTTDesc();
      assert 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, mjCtx.getRootMapJoinOp());
    }

    TableDesc tt_desc = PlanUtils.getIntermediateFileTableDesc(PlanUtils
        .getFieldSchemasFromRowSchema(parent.getSchema(), "temporarycol"));

    // generate the temporary file
    Context baseCtx = parseCtx.getContext();
    String taskTmpDir = baseCtx.getMRTmpFileURI();

    // Add the path to alias mapping
    uCtxTask.addTaskTmpDir(taskTmpDir);
    uCtxTask.addTTDesc(tt_desc);

    // The union task is empty. The files created for all the inputs are
    // assembled in the
    // union context and later used to initialize the union plan

    // Create a file sink operator for this file name
    Operator<? extends Serializable> fs_op = OperatorFactory.get(
        new FileSinkDesc(taskTmpDir, tt_desc, parseCtx.getConf().getBoolVar(
        HiveConf.ConfVars.COMPRESSINTERMEDIATE)), parent.getSchema());

    assert parent.getChildOperators().size() == 1;
    parent.getChildOperators().set(0, fs_op);

    List<Operator<? extends Serializable>> parentOpList =
      new ArrayList<Operator<? extends Serializable>>();
    parentOpList.add(parent);
    fs_op.setParentOperators(parentOpList);

    currTask.addDependentTask(uTask);

    // If it is map-only task, add the files to be processed
    if (uPrsCtx.getMapOnlySubq(pos) && uPrsCtx.getRootTask(pos)) {
      GenMapRedUtils.setTaskPlan(ctx.getCurrAliasId(), ctx.getCurrTopOp(),
          (MapredWork) currTask.getWork(), false, ctx);
    }

    ctx.setCurrTask(uTask);
View Full Code Here
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.