Examples with NodeProcessor org.apache.hadoop.hive.ql.lib.NodeProcessor used on opensource projects

Example 36 with NodeProcessor

use of org.apache.hadoop.hive.ql.lib.NodeProcessor in project hive by apache.

the class AnnotateWithStatistics method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    AnnotateStatsProcCtx aspCtx = new AnnotateStatsProcCtx(pctx);
    // create a walker which walks the tree in a BFS manner while maintaining the
    // operator stack. The dispatcher generates the plan from the operator tree
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("TS", TableScanOperator.getOperatorName() + "%"), StatsRulesProcFactory.getTableScanRule());
    opRules.put(new RuleRegExp("SEL", SelectOperator.getOperatorName() + "%"), StatsRulesProcFactory.getSelectRule());
    opRules.put(new RuleRegExp("FIL", FilterOperator.getOperatorName() + "%"), StatsRulesProcFactory.getFilterRule());
    opRules.put(new RuleRegExp("GBY", GroupByOperator.getOperatorName() + "%"), StatsRulesProcFactory.getGroupByRule());
    opRules.put(new RuleRegExp("JOIN", CommonJoinOperator.getOperatorName() + "%|" + MapJoinOperator.getOperatorName() + "%"), StatsRulesProcFactory.getJoinRule());
    opRules.put(new RuleRegExp("LIM", LimitOperator.getOperatorName() + "%"), StatsRulesProcFactory.getLimitRule());
    opRules.put(new RuleRegExp("RS", ReduceSinkOperator.getOperatorName() + "%"), StatsRulesProcFactory.getReduceSinkRule());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(StatsRulesProcFactory.getDefaultRule(), opRules, aspCtx);
    GraphWalker ogw = new LevelOrderWalker(disp, 0);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 37 with NodeProcessor

use of org.apache.hadoop.hive.ql.lib.NodeProcessor in project hive by apache.

the class ExprProcFactory method getExprDependency.

/**
   * Gets the expression dependencies for the expression.
   *
   * @param lctx
   *          The lineage context containing the input operators dependencies.
   * @param inpOp
   *          The input operator to the current operator.
   * @param expr
   *          The expression that is being processed.
   * @throws SemanticException
   */
public static Dependency getExprDependency(LineageCtx lctx, Operator<? extends OperatorDesc> inpOp, ExprNodeDesc expr) throws SemanticException {
    // Create the walker, the rules dispatcher and the context.
    ExprProcCtx exprCtx = new ExprProcCtx(lctx, inpOp);
    // create a walker which walks the tree in a DFS manner while maintaining
    // the operator stack. The dispatcher
    // generates the plan from the operator tree
    Map<Rule, NodeProcessor> exprRules = new LinkedHashMap<Rule, NodeProcessor>();
    exprRules.put(new RuleRegExp("R1", ExprNodeColumnDesc.class.getName() + "%"), getColumnProcessor());
    exprRules.put(new RuleRegExp("R2", ExprNodeFieldDesc.class.getName() + "%"), getFieldProcessor());
    exprRules.put(new RuleRegExp("R3", ExprNodeGenericFuncDesc.class.getName() + "%"), getGenericFuncProcessor());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(getDefaultExprProcessor(), exprRules, exprCtx);
    GraphWalker egw = new DefaultGraphWalker(disp);
    List<Node> startNodes = new ArrayList<Node>();
    startNodes.add(expr);
    HashMap<Node, Object> outputMap = new HashMap<Node, Object>();
    egw.startWalking(startNodes, outputMap);
    return (Dependency) outputMap.get(expr);
}

Example 38 with NodeProcessor

use of org.apache.hadoop.hive.ql.lib.NodeProcessor in project hive by apache.

the class AnnotateRunTimeStatsOptimizer method resolve.

@Override
public PhysicalContext resolve(PhysicalContext pctx) throws SemanticException {
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    Dispatcher disp = new AnnotateRunTimeStatsDispatcher(pctx, opRules);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getRootTasks());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 39 with NodeProcessor

use of org.apache.hadoop.hive.ql.lib.NodeProcessor in project hive by apache.

the class RewriteCanApplyCtx method populateRewriteVars.

/**
   * This method walks all the nodes starting from topOp TableScanOperator node
   * and invokes methods from {@link RewriteCanApplyProcFactory} for each of the rules
   * added to the opRules map. We use the {@link PreOrderOnceWalker} for a pre-order
   * traversal of the operator tree.
   *
   * The methods from {@link RewriteCanApplyProcFactory} set appropriate values in
   * {@link RewriteVars} enum.
   *
   * @param topOp
   * @throws SemanticException
   */
void populateRewriteVars(TableScanOperator topOp) throws SemanticException {
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    //^TS%[(SEL%)|(FIL%)]*GRY%[(FIL%)]*RS%[(FIL%)]*GRY%
    opRules.put(new RuleRegExp("R1", TableScanOperator.getOperatorName() + "%[(" + SelectOperator.getOperatorName() + "%)|(" + FilterOperator.getOperatorName() + "%)]*" + GroupByOperator.getOperatorName() + "%[" + FilterOperator.getOperatorName() + "%]*" + ReduceSinkOperator.getOperatorName() + "%[" + FilterOperator.getOperatorName() + "%]*" + GroupByOperator.getOperatorName() + "%"), RewriteCanApplyProcFactory.canApplyOnTableScanOperator(topOp));
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(getDefaultProc(), opRules, this);
    GraphWalker ogw = new PreOrderOnceWalker(disp);
    // Create a list of topop nodes
    List<Node> topNodes = new ArrayList<Node>();
    topNodes.add(topOp);
    try {
        ogw.startWalking(topNodes, null);
    } catch (SemanticException e) {
        LOG.error("Exception in walking operator tree. Rewrite variables not populated");
        LOG.error(org.apache.hadoop.util.StringUtils.stringifyException(e));
        throw new SemanticException(e.getMessage(), e);
    }
}

Example 40 with NodeProcessor

use of org.apache.hadoop.hive.ql.lib.NodeProcessor in project hive by apache.

the class SimpleFetchAggregation method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    if (pctx.getFetchTask() != null || !pctx.getQueryProperties().isQuery() || pctx.getQueryProperties().isAnalyzeRewrite() || pctx.getQueryProperties().isCTAS() || pctx.getLoadFileWork().size() > 1 || !pctx.getLoadTableWork().isEmpty()) {
        return pctx;
    }
    String GBY = GroupByOperator.getOperatorName() + "%";
    String RS = ReduceSinkOperator.getOperatorName() + "%";
    String SEL = SelectOperator.getOperatorName() + "%";
    String FS = FileSinkOperator.getOperatorName() + "%";
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", GBY + RS + GBY + SEL + FS), new SingleGBYProcessor(pctx));
    opRules.put(new RuleRegExp("R2", GBY + RS + GBY + FS), new SingleGBYProcessor(pctx));
    Dispatcher disp = new DefaultRuleDispatcher(null, opRules, null);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}