Search in sources :

Example 11 with SemanticNodeProcessor

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

the class PredicatePushDown method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    pGraphContext = pctx;
    // create a the context for walking operators
    OpWalkerInfo opWalkerInfo = new OpWalkerInfo(pGraphContext);
    Map<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    opRules.put(new RuleRegExp("R1", FilterOperator.getOperatorName() + "%"), OpProcFactory.getFilterProc());
    opRules.put(new RuleRegExp("R2", PTFOperator.getOperatorName() + "%"), OpProcFactory.getPTFProc());
    opRules.put(new RuleRegExp("R3", CommonJoinOperator.getOperatorName() + "%"), OpProcFactory.getJoinProc());
    opRules.put(new RuleRegExp("R4", TableScanOperator.getOperatorName() + "%"), OpProcFactory.getTSProc());
    opRules.put(new RuleRegExp("R5", ScriptOperator.getOperatorName() + "%"), OpProcFactory.getSCRProc());
    opRules.put(new RuleRegExp("R6", LimitOperator.getOperatorName() + "%"), OpProcFactory.getLIMProc());
    opRules.put(new RuleRegExp("R7", UDTFOperator.getOperatorName() + "%"), OpProcFactory.getUDTFProc());
    opRules.put(new RuleRegExp("R8", LateralViewForwardOperator.getOperatorName() + "%"), OpProcFactory.getLVFProc());
    opRules.put(new RuleRegExp("R9", LateralViewJoinOperator.getOperatorName() + "%"), OpProcFactory.getLVJProc());
    opRules.put(new RuleRegExp("R10", ReduceSinkOperator.getOperatorName() + "%"), OpProcFactory.getRSProc());
    opRules.put(new RuleRegExp("R11", GroupByOperator.getOperatorName() + "%"), OpProcFactory.getGBYProc());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    SemanticDispatcher disp = new DefaultRuleDispatcher(OpProcFactory.getDefaultProc(), opRules, opWalkerInfo);
    SemanticGraphWalker ogw = new DefaultGraphWalker(disp);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pGraphContext.getTopOps().values());
    ogw.startWalking(topNodes, null);
    if (LOG.isDebugEnabled()) {
        LOG.debug("After PPD:\n" + Operator.toString(pctx.getTopOps().values()));
    }
    return pGraphContext;
}
Also used : SemanticRule(org.apache.hadoop.hive.ql.lib.SemanticRule) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker) Node(org.apache.hadoop.hive.ql.lib.Node) RuleRegExp(org.apache.hadoop.hive.ql.lib.RuleRegExp) ArrayList(java.util.ArrayList) SemanticGraphWalker(org.apache.hadoop.hive.ql.lib.SemanticGraphWalker) LinkedHashMap(java.util.LinkedHashMap) SemanticDispatcher(org.apache.hadoop.hive.ql.lib.SemanticDispatcher) SemanticNodeProcessor(org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor)

Example 12 with SemanticNodeProcessor

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

the class SimplePredicatePushDown method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    pGraphContext = pctx;
    // create a the context for walking operators
    OpWalkerInfo opWalkerInfo = new OpWalkerInfo(pGraphContext);
    Map<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    opRules.put(new RuleRegExp("R1", FilterOperator.getOperatorName() + "%"), OpProcFactory.getFilterSyntheticJoinPredicateProc());
    opRules.put(new RuleRegExp("R2", PTFOperator.getOperatorName() + "%"), OpProcFactory.getPTFProc());
    opRules.put(new RuleRegExp("R3", CommonJoinOperator.getOperatorName() + "%"), OpProcFactory.getJoinProc());
    opRules.put(new RuleRegExp("R4", TableScanOperator.getOperatorName() + "%"), OpProcFactory.getTSProc());
    opRules.put(new RuleRegExp("R5", ScriptOperator.getOperatorName() + "%"), OpProcFactory.getSCRProc());
    opRules.put(new RuleRegExp("R6", LimitOperator.getOperatorName() + "%"), OpProcFactory.getLIMProc());
    opRules.put(new RuleRegExp("R7", UDTFOperator.getOperatorName() + "%"), OpProcFactory.getUDTFProc());
    opRules.put(new RuleRegExp("R8", LateralViewForwardOperator.getOperatorName() + "%"), OpProcFactory.getLVFProc());
    opRules.put(new RuleRegExp("R9", LateralViewJoinOperator.getOperatorName() + "%"), OpProcFactory.getLVJProc());
    opRules.put(new RuleRegExp("R10", ReduceSinkOperator.getOperatorName() + "%"), OpProcFactory.getRSProc());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    SemanticDispatcher disp = new DefaultRuleDispatcher(OpProcFactory.getDefaultProc(), opRules, opWalkerInfo);
    SemanticGraphWalker ogw = new DefaultGraphWalker(disp);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pGraphContext.getTopOps().values());
    ogw.startWalking(topNodes, null);
    if (LOG.isDebugEnabled()) {
        LOG.debug("After PPD:\n" + Operator.toString(pctx.getTopOps().values()));
    }
    return pGraphContext;
}
Also used : SemanticRule(org.apache.hadoop.hive.ql.lib.SemanticRule) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker) Node(org.apache.hadoop.hive.ql.lib.Node) RuleRegExp(org.apache.hadoop.hive.ql.lib.RuleRegExp) ArrayList(java.util.ArrayList) SemanticGraphWalker(org.apache.hadoop.hive.ql.lib.SemanticGraphWalker) LinkedHashMap(java.util.LinkedHashMap) SemanticDispatcher(org.apache.hadoop.hive.ql.lib.SemanticDispatcher) SemanticNodeProcessor(org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor)

Example 13 with SemanticNodeProcessor

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

the class PredicateTransitivePropagate method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    pGraphContext = pctx;
    Map<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    opRules.put(new RuleRegExp("R1", "(" + FilterOperator.getOperatorName() + "%" + ReduceSinkOperator.getOperatorName() + "%" + JoinOperator.getOperatorName() + "%)"), new JoinTransitive());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    TransitiveContext context = new TransitiveContext();
    SemanticDispatcher disp = new DefaultRuleDispatcher(null, opRules, context);
    SemanticGraphWalker ogw = new LevelOrderWalker(disp, 2);
    // Create a list of topop nodes
    List<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pGraphContext.getTopOps().values());
    ogw.startWalking(topNodes, null);
    Map<ReduceSinkOperator, List<ExprNodeDesc>> newFilters = context.getNewfilters();
    // insert new filter between RS and parent of RS
    for (Map.Entry<ReduceSinkOperator, List<ExprNodeDesc>> entry : newFilters.entrySet()) {
        ReduceSinkOperator reducer = entry.getKey();
        Operator<?> parent = reducer.getParentOperators().get(0);
        List<ExprNodeDesc> exprs = entry.getValue();
        if (parent instanceof FilterOperator) {
            exprs = ExprNodeDescUtils.split(((FilterOperator) parent).getConf().getPredicate(), exprs);
            ExprNodeDesc merged = ExprNodeDescUtils.mergePredicates(exprs);
            ((FilterOperator) parent).getConf().setPredicate(merged);
        } else {
            ExprNodeDesc merged = ExprNodeDescUtils.mergePredicates(exprs);
            RowSchema parentRS = parent.getSchema();
            Operator<FilterDesc> newFilter = createFilter(reducer, parent, parentRS, merged);
        }
    }
    return pGraphContext;
}
Also used : Node(org.apache.hadoop.hive.ql.lib.Node) ArrayList(java.util.ArrayList) LinkedHashMap(java.util.LinkedHashMap) ArrayList(java.util.ArrayList) List(java.util.List) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) RowSchema(org.apache.hadoop.hive.ql.exec.RowSchema) SemanticRule(org.apache.hadoop.hive.ql.lib.SemanticRule) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) RuleRegExp(org.apache.hadoop.hive.ql.lib.RuleRegExp) SemanticGraphWalker(org.apache.hadoop.hive.ql.lib.SemanticGraphWalker) FilterOperator(org.apache.hadoop.hive.ql.exec.FilterOperator) FilterDesc(org.apache.hadoop.hive.ql.plan.FilterDesc) ReduceSinkOperator(org.apache.hadoop.hive.ql.exec.ReduceSinkOperator) SemanticDispatcher(org.apache.hadoop.hive.ql.lib.SemanticDispatcher) SemanticNodeProcessor(org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor) LevelOrderWalker(org.apache.hadoop.hive.ql.lib.LevelOrderWalker) HashMap(java.util.HashMap) LinkedHashMap(java.util.LinkedHashMap) Map(java.util.Map)

Example 14 with SemanticNodeProcessor

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

the class SyntheticJoinPredicate method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    boolean enabled = false;
    String queryEngine = pctx.getConf().getVar(ConfVars.HIVE_EXECUTION_ENGINE);
    if (queryEngine.equals("tez") && pctx.getConf().getBoolVar(ConfVars.TEZ_DYNAMIC_PARTITION_PRUNING)) {
        enabled = true;
    } else if ((queryEngine.equals("spark") && pctx.getConf().isSparkDPPAny())) {
        enabled = true;
    }
    if (!enabled) {
        return pctx;
    }
    Map<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    opRules.put(new RuleRegExp("R1", "(" + TableScanOperator.getOperatorName() + "%" + ".*" + ReduceSinkOperator.getOperatorName() + "%" + JoinOperator.getOperatorName() + "%)"), new JoinSynthetic());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    SyntheticContext context = new SyntheticContext(pctx);
    SemanticDispatcher disp = new DefaultRuleDispatcher(null, opRules, context);
    SemanticGraphWalker ogw = new PreOrderOnceWalker(disp);
    // Create a list of top op nodes
    List<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}
Also used : SemanticRule(org.apache.hadoop.hive.ql.lib.SemanticRule) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) Node(org.apache.hadoop.hive.ql.lib.Node) RuleRegExp(org.apache.hadoop.hive.ql.lib.RuleRegExp) ArrayList(java.util.ArrayList) SemanticGraphWalker(org.apache.hadoop.hive.ql.lib.SemanticGraphWalker) LinkedHashMap(java.util.LinkedHashMap) SemanticDispatcher(org.apache.hadoop.hive.ql.lib.SemanticDispatcher) SemanticNodeProcessor(org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor) PreOrderOnceWalker(org.apache.hadoop.hive.ql.lib.PreOrderOnceWalker)

Example 15 with SemanticNodeProcessor

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

the class LineageInfo method getLineageInfo.

/**
 * parses given query and gets the lineage info.
 *
 * @param query
 */
public void getLineageInfo(String query, Context ctx) throws Exception {
    /*
     * Get the AST tree
     */
    ASTNode tree = ParseUtils.parse(query, ctx);
    while ((tree.getToken() == null) && (tree.getChildCount() > 0)) {
        tree = (ASTNode) tree.getChild(0);
    }
    /*
     * initialize Event Processor and dispatcher.
     */
    inputTableList.clear();
    OutputTableList.clear();
    // 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<SemanticRule, SemanticNodeProcessor> rules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    SemanticDispatcher disp = new DefaultRuleDispatcher(this, rules, null);
    SemanticGraphWalker ogw = new DefaultGraphWalker(disp);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.add(tree);
    ogw.startWalking(topNodes, null);
}
Also used : SemanticRule(org.apache.hadoop.hive.ql.lib.SemanticRule) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker) ASTNode(org.apache.hadoop.hive.ql.parse.ASTNode) Node(org.apache.hadoop.hive.ql.lib.Node) ASTNode(org.apache.hadoop.hive.ql.parse.ASTNode) SemanticDispatcher(org.apache.hadoop.hive.ql.lib.SemanticDispatcher) ArrayList(java.util.ArrayList) SemanticGraphWalker(org.apache.hadoop.hive.ql.lib.SemanticGraphWalker) SemanticNodeProcessor(org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor) LinkedHashMap(java.util.LinkedHashMap)

Aggregations

SemanticNodeProcessor (org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor)68 LinkedHashMap (java.util.LinkedHashMap)66 Node (org.apache.hadoop.hive.ql.lib.Node)66 SemanticGraphWalker (org.apache.hadoop.hive.ql.lib.SemanticGraphWalker)66 SemanticRule (org.apache.hadoop.hive.ql.lib.SemanticRule)66 SemanticDispatcher (org.apache.hadoop.hive.ql.lib.SemanticDispatcher)65 DefaultRuleDispatcher (org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher)63 ArrayList (java.util.ArrayList)62 RuleRegExp (org.apache.hadoop.hive.ql.lib.RuleRegExp)57 DefaultGraphWalker (org.apache.hadoop.hive.ql.lib.DefaultGraphWalker)36 MapJoinOperator (org.apache.hadoop.hive.ql.exec.MapJoinOperator)11 ReduceSinkOperator (org.apache.hadoop.hive.ql.exec.ReduceSinkOperator)11 JoinOperator (org.apache.hadoop.hive.ql.exec.JoinOperator)9 Operator (org.apache.hadoop.hive.ql.exec.Operator)9 FileSinkOperator (org.apache.hadoop.hive.ql.exec.FileSinkOperator)8 ForwardWalker (org.apache.hadoop.hive.ql.lib.ForwardWalker)8 TableScanOperator (org.apache.hadoop.hive.ql.exec.TableScanOperator)7 UnionOperator (org.apache.hadoop.hive.ql.exec.UnionOperator)7 AppMasterEventOperator (org.apache.hadoop.hive.ql.exec.AppMasterEventOperator)6 DummyStoreOperator (org.apache.hadoop.hive.ql.exec.DummyStoreOperator)6