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

Example 31 with GraphWalker

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

the class MetadataOnlyOptimizer method resolve.

@Override
public PhysicalContext resolve(PhysicalContext pctx) throws SemanticException {
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", TableScanOperator.getOperatorName() + "%"), new TableScanProcessor());
    opRules.put(new RuleRegExp("R2", GroupByOperator.getOperatorName() + "%.*" + FileSinkOperator.getOperatorName() + "%"), new FileSinkProcessor());
    Dispatcher disp = new NullScanTaskDispatcher(pctx, opRules);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getRootTasks());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 32 with GraphWalker

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

the class NullScanTaskDispatcher method dispatch.

@Override
public Object dispatch(Node nd, Stack<Node> stack, Object... nodeOutputs) throws SemanticException {
    Task<? extends Serializable> task = (Task<? extends Serializable>) nd;
    // create a the context for walking operators
    ParseContext parseContext = physicalContext.getParseContext();
    WalkerCtx walkerCtx = new WalkerCtx();
    List<MapWork> mapWorks = new ArrayList<MapWork>(task.getMapWork());
    Collections.sort(mapWorks, new Comparator<MapWork>() {

        @Override
        public int compare(MapWork o1, MapWork o2) {
            return o1.getName().compareTo(o2.getName());
        }
    });
    for (MapWork mapWork : mapWorks) {
        LOG.debug("Looking at: " + mapWork.getName());
        Collection<Operator<? extends OperatorDesc>> topOperators = mapWork.getAliasToWork().values();
        if (topOperators.size() == 0) {
            LOG.debug("No top operators");
            return null;
        }
        LOG.debug("Looking for table scans where optimization is applicable");
        // The dispatcher fires the processor corresponding to the closest
        // matching rule and passes the context along
        Dispatcher disp = new DefaultRuleDispatcher(null, rules, walkerCtx);
        GraphWalker ogw = new PreOrderOnceWalker(disp);
        // Create a list of topOp nodes
        ArrayList<Node> topNodes = new ArrayList<Node>();
        // Get the top Nodes for this task
        for (Operator<? extends OperatorDesc> workOperator : topOperators) {
            if (parseContext.getTopOps().values().contains(workOperator)) {
                topNodes.add(workOperator);
            }
        }
        Operator<? extends OperatorDesc> reducer = task.getReducer(mapWork);
        if (reducer != null) {
            topNodes.add(reducer);
        }
        ogw.startWalking(topNodes, null);
        LOG.debug(String.format("Found %d null table scans", walkerCtx.getMetadataOnlyTableScans().size()));
        if (walkerCtx.getMetadataOnlyTableScans().size() > 0)
            processAlias(mapWork, walkerCtx.getMetadataOnlyTableScans());
    }
    return null;
}

Example 33 with GraphWalker

use of org.apache.hadoop.hive.ql.lib.GraphWalker in project SQLWindowing by hbutani.

the class WindowingTypeCheckProcFactory method genExprNode.

public static HashMap<Node, Object> genExprNode(ASTNode expr, TypeCheckCtx tcCtx) throws SemanticException {
    // Create the walker, the rules dispatcher and the context.
    // 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> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", Windowing2Parser.NULL + "%"), getNullExprProcessor());
    opRules.put(new RuleRegExp("R2", Windowing2Parser.Number + "%|" + Windowing2Parser.TinyintLiteral + "%|" + Windowing2Parser.SmallintLiteral + "%|" + Windowing2Parser.BigintLiteral + "%"), getNumExprProcessor());
    opRules.put(new RuleRegExp("R3", Windowing2Parser.Identifier + "%|" + Windowing2Parser.StringLiteral + "%|" + Windowing2Parser.CHARSETLITERAL + "%|" + Windowing2Parser.STRINGLITERALSEQUENCE + "%|" + "%|" + Windowing2Parser.IF + "%|" + Windowing2Parser.CASE + "%|" + Windowing2Parser.WHEN + "%|" + Windowing2Parser.IN + "%|" + Windowing2Parser.ARRAY + "%|" + Windowing2Parser.MAP + "%|" + Windowing2Parser.BETWEEN + "%|" + Windowing2Parser.STRUCT + "%"), getStrExprProcessor());
    opRules.put(new RuleRegExp("R4", Windowing2Parser.TRUE + "%|" + Windowing2Parser.FALSE + "%"), getBoolExprProcessor());
    opRules.put(new RuleRegExp("R5", Windowing2Parser.TABLEORCOL + "%"), getColumnExprProcessor());
    // The dispatcher fires the processor corresponding to the closest
    // matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(getDefaultExprProcessor(), opRules, tcCtx);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.add(expr);
    HashMap<Node, Object> nodeOutputs = new HashMap<Node, Object>();
    ogw.startWalking(topNodes, nodeOutputs);
    return nodeOutputs;
}

Example 34 with GraphWalker

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

the class PrunerUtils method walkExprTree.

/**
   * Walk expression tree for pruner generation.
   *
   * @param pred
   * @param ctx
   * @param colProc
   * @param fieldProc
   * @param genFuncProc
   * @param defProc
   * @return
   * @throws SemanticException
   */
public static Map<Node, Object> walkExprTree(ExprNodeDesc pred, NodeProcessorCtx ctx, NodeProcessor colProc, NodeProcessor fieldProc, NodeProcessor genFuncProc, NodeProcessor defProc) throws SemanticException {
    // 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 TypeRule(ExprNodeColumnDesc.class), colProc);
    exprRules.put(new TypeRule(ExprNodeFieldDesc.class), fieldProc);
    exprRules.put(new TypeRule(ExprNodeGenericFuncDesc.class), genFuncProc);
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(defProc, exprRules, ctx);
    GraphWalker egw = new DefaultGraphWalker(disp);
    List<Node> startNodes = new ArrayList<Node>();
    startNodes.add(pred);
    HashMap<Node, Object> outputMap = new HashMap<Node, Object>();
    egw.startWalking(startNodes, outputMap);
    return outputMap;
}

Example 35 with GraphWalker

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

the class RedundantDynamicPruningConditionsRemoval method transform.

/**
   * Transform the query tree.
   *
   * @param pctx the current parse context
   */
@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    // Make sure semijoin is not enabled. If it is, then do not remove the dynamic partition pruning predicates.
    if (!pctx.getConf().getBoolVar(HiveConf.ConfVars.TEZ_DYNAMIC_SEMIJOIN_REDUCTION)) {
        Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
        opRules.put(new RuleRegExp("R1", TableScanOperator.getOperatorName() + "%" + FilterOperator.getOperatorName() + "%"), new FilterTransformer());
        Dispatcher disp = new DefaultRuleDispatcher(null, opRules, null);
        GraphWalker ogw = new DefaultGraphWalker(disp);
        List<Node> topNodes = new ArrayList<Node>();
        topNodes.addAll(pctx.getTopOps().values());
        ogw.startWalking(topNodes, null);
    }
    return pctx;
}