Package org.apache.hadoop.hive.ql.parse

Examples of org.apache.hadoop.hive.ql.parse.ParseContext


   */
  public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx opProcCtx, Object... nodeOutputs) throws SemanticException {
    ReduceSinkOperator op = (ReduceSinkOperator)nd;
    GenMRProcContext ctx = (GenMRProcContext)opProcCtx;

    ParseContext parseCtx = ctx.getParseCtx();

    // map-join consisted on a bunch of map-only jobs, and it has been split after the mapjoin
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx.getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
View Full Code Here


  public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx opProcCtx,
      Object... nodeOutputs) throws SemanticException {
    ReduceSinkOperator op = (ReduceSinkOperator) nd;
    GenMRProcContext ctx = (GenMRProcContext) opProcCtx;

    ParseContext parseCtx = ctx.getParseCtx();
    UnionProcContext uCtx = parseCtx.getUCtx();

    // union was map only - no special processing needed
    if (uCtx.isMapOnlySubq()) {
      return (new GenMRRedSink1()).process(nd, stack, opProcCtx, nodeOutputs);
    }
View Full Code Here

    public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procCtx,
        Object... nodeOutputs) throws SemanticException {

      // LineageCtx
      LineageCtx lCtx = (LineageCtx) procCtx;
      ParseContext pctx = lCtx.getParseCtx();

      // Table scan operator.
      TableScanOperator top = (TableScanOperator)nd;
      org.apache.hadoop.hive.ql.metadata.Table t = pctx.getTopToTable().get(top);
      Table tab = t.getTTable();

      // Generate the mappings
      RowSchema rs = top.getSchema();
      List<FieldSchema> cols = t.getAllCols();
      Map<String, FieldSchema> fieldSchemaMap = new HashMap<String, FieldSchema>();
      for(FieldSchema col : cols) {
        fieldSchemaMap.put(col.getName(), col);
      }

      Iterator<VirtualColumn> vcs = VirtualColumn.getRegistry(pctx.getConf()).iterator();
      while (vcs.hasNext()) {
        VirtualColumn vc = vcs.next();
        fieldSchemaMap.put(vc.getName(), new FieldSchema(vc.getName(),
            vc.getTypeInfo().getTypeName(), ""));
      }
View Full Code Here

   */
  public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx opProcCtx,
      Object... nodeOutputs) throws SemanticException {
    UnionOperator union = (UnionOperator) nd;
    GenMRProcContext ctx = (GenMRProcContext) opProcCtx;
    ParseContext parseCtx = ctx.getParseCtx();
    UnionProcContext uCtx = parseCtx.getUCtx();

    // Map-only subqueries can be optimized in future to not write to a file in
    // future
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
        .getMapCurrCtx();

    // The plan needs to be broken only if one of the sub-queries involve a
    // map-reduce job
    if (uCtx.isMapOnlySubq()) {
      // merge currTask from multiple topOps
      HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx
          .getOpTaskMap();
      if (opTaskMap != null && opTaskMap.size() > 0) {
        Task<? extends Serializable> tsk = opTaskMap.get(null);
        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);
      uTask = TaskFactory.get(uPlan, parseCtx.getConf());
      uCtxTask.setUTask(uTask);
      ctx.setUnionTask(union, uCtxTask);
    } else {
      uTask = uCtxTask.getUTask();
    }

    // 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);

View Full Code Here

   * @throws SemanticException
   */
  public static ParseContext generateOperatorTree(HiveConf conf,
      String command) throws SemanticException{
    Context ctx;
    ParseContext subPCtx = null;
    try {
      ctx = new Context(conf);
      ParseDriver pd = new ParseDriver();
      ASTNode tree = pd.parse(command, ctx);
      tree = ParseUtils.findRootNonNullToken(tree);
View Full Code Here

   */
  private static void doSemanticAnalysis(SemanticAnalyzer sem,
      ASTNode ast, Context ctx) throws SemanticException {
    QB qb = new QB(null, null, false);
    ASTNode child = ast;
    ParseContext subPCtx = ((SemanticAnalyzer) sem).getParseContext();
    subPCtx.setContext(ctx);
    ((SemanticAnalyzer) sem).init(subPCtx);

    LOG.info("Starting Sub-query Semantic Analysis");
    sem.doPhase1(child, qb, sem.initPhase1Ctx());
    LOG.info("Completed phase 1 of Sub-query Semantic Analysis");
View Full Code Here

       
        if(ctx.contains(childReduceSink)) {
          return null;
        }
       
        ParseContext pGraphContext = ctx.getPctx();
        HashMap<String, String> childColumnMapping = getPartitionAndKeyColumnMapping(childReduceSink);
        ReduceSinkOperator parentRS = null;
        parentRS = findSingleParentReduceSink(childReduceSink, pGraphContext);
        if (parentRS == null) {
          ctx.addRejectedReduceSinkOperator(childReduceSink);
View Full Code Here

    public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procCtx,
        Object... nodeOutputs) throws SemanticException {
      AbstractMapJoinOperator<MapJoinDesc> mapJoin = (AbstractMapJoinOperator<MapJoinDesc>) nd;
      GenMRProcContext opProcCtx = (GenMRProcContext) procCtx;

      ParseContext parseCtx = opProcCtx.getParseCtx();
      MapredWork cplan = GenMapRedUtils.getMapRedWork(parseCtx);
      Task<? extends Serializable> redTask = TaskFactory.get(cplan, parseCtx
          .getConf());
      Task<? extends Serializable> currTask = opProcCtx.getCurrTask();

      // find the branch on which this processor was invoked
      int pos = getPositionParent(mapJoin, stack);
View Full Code Here

      AbstractMapJoinOperator<MapJoinDesc> mapJoin = (AbstractMapJoinOperator<MapJoinDesc>) sel.getParentOperators().get(
          0);
      assert sel.getParentOperators().size() == 1;

      GenMRProcContext ctx = (GenMRProcContext) procCtx;
      ParseContext parseCtx = ctx.getParseCtx();

      // is the mapjoin followed by a reducer
      List<AbstractMapJoinOperator<? extends MapJoinDesc>> listMapJoinOps = parseCtx
          .getListMapJoinOpsNoReducer();

      if (listMapJoinOps.contains(mapJoin)) {
        ctx.setCurrAliasId(null);
        ctx.setCurrTopOp(null);
        Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
            .getMapCurrCtx();
        mapCurrCtx.put((Operator<? extends Serializable>) nd, new GenMapRedCtx(
            ctx.getCurrTask(), null, null));
        return null;
      }

      ctx.setCurrMapJoinOp(mapJoin);

      Task<? extends Serializable> currTask = ctx.getCurrTask();
      GenMRMapJoinCtx mjCtx = ctx.getMapJoinCtx(mapJoin);
      if (mjCtx == null) {
        mjCtx = new GenMRMapJoinCtx();
        ctx.setMapJoinCtx(mapJoin, mjCtx);
      }

      MapredWork mjPlan = GenMapRedUtils.getMapRedWork(parseCtx);
      Task<? extends Serializable> mjTask = TaskFactory.get(mjPlan, parseCtx
          .getConf());

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

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

      // Add the path to alias mapping
      mjCtx.setTaskTmpDir(taskTmpDir);
      mjCtx.setTTDesc(tt_desc);
      mjCtx.setRootMapJoinOp(sel);

      sel.setParentOperators(null);

      // 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)), mapJoin.getSchema());

      assert mapJoin.getChildOperators().size() == 1;
      mapJoin.getChildOperators().set(0, fs_op);
View Full Code Here

    @Override
    public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procCtx,
        Object... nodeOutputs) throws SemanticException {
      GenMRProcContext ctx = (GenMRProcContext) procCtx;

      ParseContext parseCtx = ctx.getParseCtx();
      UnionProcContext uCtx = parseCtx.getUCtx();

      // union was map only - no special processing needed
      if (uCtx.isMapOnlySubq()) {
        return (new TableScanMapJoin())
            .process(nd, stack, procCtx, nodeOutputs);
View Full Code Here

TOP

Related Classes of org.apache.hadoop.hive.ql.parse.ParseContext

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.