Search in sources :

Example 16 with DefaultRuleDispatcher

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

the class AnnotateWithOpTraits method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    AnnotateOpTraitsProcCtx annotateCtx = new AnnotateOpTraitsProcCtx(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() + "%"), OpTraitsRulesProcFactory.getTableScanRule());
    opRules.put(new RuleRegExp("RS", ReduceSinkOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getReduceSinkRule());
    opRules.put(new RuleRegExp("JOIN", JoinOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getJoinRule());
    opRules.put(new RuleRegExp("MAPJOIN", MapJoinOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getMultiParentRule());
    opRules.put(new RuleRegExp("SMB", SMBMapJoinOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getMultiParentRule());
    opRules.put(new RuleRegExp("MUX", MuxOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getMultiParentRule());
    opRules.put(new RuleRegExp("DEMUX", DemuxOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getMultiParentRule());
    opRules.put(new RuleRegExp("UNION", UnionOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getMultiParentRule());
    opRules.put(new RuleRegExp("GBY", GroupByOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getGroupByRule());
    opRules.put(new RuleRegExp("SEL", SelectOperator.getOperatorName() + "%"), OpTraitsRulesProcFactory.getSelectRule());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(OpTraitsRulesProcFactory.getDefaultRule(), opRules, annotateCtx);
    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;
}
Also used : NodeProcessor(org.apache.hadoop.hive.ql.lib.NodeProcessor) 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) Dispatcher(org.apache.hadoop.hive.ql.lib.Dispatcher) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) LinkedHashMap(java.util.LinkedHashMap) Rule(org.apache.hadoop.hive.ql.lib.Rule) LevelOrderWalker(org.apache.hadoop.hive.ql.lib.LevelOrderWalker) GraphWalker(org.apache.hadoop.hive.ql.lib.GraphWalker)

Example 17 with DefaultRuleDispatcher

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

the class PartitionConditionRemover method transform.

/*
   * (non-Javadoc)
   *
   * @see
   * org.apache.hadoop.hive.ql.optimizer.Transform#transform(org.apache.hadoop
   * .hive.ql.parse.ParseContext)
   */
@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    // create a the context for walking operators
    List<PcrOpWalkerCtx.OpToDeleteInfo> opToRemove = new ArrayList<PcrOpWalkerCtx.OpToDeleteInfo>();
    PcrOpWalkerCtx opWalkerCtx = new PcrOpWalkerCtx(pctx, opToRemove);
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", "(" + TableScanOperator.getOperatorName() + "%" + FilterOperator.getOperatorName() + "%)|(" + TableScanOperator.getOperatorName() + "%" + FilterOperator.getOperatorName() + "%" + FilterOperator.getOperatorName() + "%)"), PcrOpProcFactory.getFilterProc());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(PcrOpProcFactory.getDefaultProc(), opRules, opWalkerCtx);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    for (PcrOpWalkerCtx.OpToDeleteInfo entry : opToRemove) {
        entry.getParent().removeChildAndAdoptItsChildren(entry.getOperator());
    }
    return pctx;
}
Also used : NodeProcessor(org.apache.hadoop.hive.ql.lib.NodeProcessor) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker) Node(org.apache.hadoop.hive.ql.lib.Node) ArrayList(java.util.ArrayList) RuleRegExp(org.apache.hadoop.hive.ql.lib.RuleRegExp) Dispatcher(org.apache.hadoop.hive.ql.lib.Dispatcher) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) LinkedHashMap(java.util.LinkedHashMap) Rule(org.apache.hadoop.hive.ql.lib.Rule) GraphWalker(org.apache.hadoop.hive.ql.lib.GraphWalker) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker)

Example 18 with DefaultRuleDispatcher

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

the class MapReduceCompiler method generateTaskTree.

@Override
protected void generateTaskTree(List<Task<? extends Serializable>> rootTasks, ParseContext pCtx, List<Task<MoveWork>> mvTask, Set<ReadEntity> inputs, Set<WriteEntity> outputs) throws SemanticException {
    // generate map reduce plans
    ParseContext tempParseContext = getParseContext(pCtx, rootTasks);
    GenMRProcContext procCtx = new GenMRProcContext(conf, // Must be deterministic order map for consistent q-test output across Java versions
    new LinkedHashMap<Operator<? extends OperatorDesc>, Task<? extends Serializable>>(), tempParseContext, mvTask, rootTasks, new LinkedHashMap<Operator<? extends OperatorDesc>, GenMapRedCtx>(), inputs, outputs);
    // 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(new String("R1"), TableScanOperator.getOperatorName() + "%"), new GenMRTableScan1());
    opRules.put(new RuleRegExp(new String("R2"), TableScanOperator.getOperatorName() + "%.*" + ReduceSinkOperator.getOperatorName() + "%"), new GenMRRedSink1());
    opRules.put(new RuleRegExp(new String("R3"), ReduceSinkOperator.getOperatorName() + "%.*" + ReduceSinkOperator.getOperatorName() + "%"), new GenMRRedSink2());
    opRules.put(new RuleRegExp(new String("R4"), FileSinkOperator.getOperatorName() + "%"), new GenMRFileSink1());
    opRules.put(new RuleRegExp(new String("R5"), UnionOperator.getOperatorName() + "%"), new GenMRUnion1());
    opRules.put(new RuleRegExp(new String("R6"), UnionOperator.getOperatorName() + "%.*" + ReduceSinkOperator.getOperatorName() + "%"), new GenMRRedSink3());
    opRules.put(new RuleRegExp(new String("R7"), MapJoinOperator.getOperatorName() + "%"), MapJoinFactory.getTableScanMapJoin());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(new GenMROperator(), opRules, procCtx);
    GraphWalker ogw = new GenMapRedWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pCtx.getTopOps().values());
    ogw.startWalking(topNodes, null);
}
Also used : ReduceSinkOperator(org.apache.hadoop.hive.ql.exec.ReduceSinkOperator) MapJoinOperator(org.apache.hadoop.hive.ql.exec.MapJoinOperator) UnionOperator(org.apache.hadoop.hive.ql.exec.UnionOperator) FileSinkOperator(org.apache.hadoop.hive.ql.exec.FileSinkOperator) GenMROperator(org.apache.hadoop.hive.ql.optimizer.GenMROperator) TableScanOperator(org.apache.hadoop.hive.ql.exec.TableScanOperator) Operator(org.apache.hadoop.hive.ql.exec.Operator) ConditionalTask(org.apache.hadoop.hive.ql.exec.ConditionalTask) MapRedTask(org.apache.hadoop.hive.ql.exec.mr.MapRedTask) Task(org.apache.hadoop.hive.ql.exec.Task) Serializable(java.io.Serializable) Node(org.apache.hadoop.hive.ql.lib.Node) ArrayList(java.util.ArrayList) Dispatcher(org.apache.hadoop.hive.ql.lib.Dispatcher) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) LinkedHashMap(java.util.LinkedHashMap) GenMRTableScan1(org.apache.hadoop.hive.ql.optimizer.GenMRTableScan1) GenMapRedCtx(org.apache.hadoop.hive.ql.optimizer.GenMRProcContext.GenMapRedCtx) GenMRProcContext(org.apache.hadoop.hive.ql.optimizer.GenMRProcContext) GraphWalker(org.apache.hadoop.hive.ql.lib.GraphWalker) GenMROperator(org.apache.hadoop.hive.ql.optimizer.GenMROperator) NodeProcessor(org.apache.hadoop.hive.ql.lib.NodeProcessor) GenMRUnion1(org.apache.hadoop.hive.ql.optimizer.GenMRUnion1) GenMRFileSink1(org.apache.hadoop.hive.ql.optimizer.GenMRFileSink1) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) RuleRegExp(org.apache.hadoop.hive.ql.lib.RuleRegExp) GenMRRedSink2(org.apache.hadoop.hive.ql.optimizer.GenMRRedSink2) GenMRRedSink1(org.apache.hadoop.hive.ql.optimizer.GenMRRedSink1) GenMRRedSink3(org.apache.hadoop.hive.ql.optimizer.GenMRRedSink3) Rule(org.apache.hadoop.hive.ql.lib.Rule) OperatorDesc(org.apache.hadoop.hive.ql.plan.OperatorDesc)

Example 19 with DefaultRuleDispatcher

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

the class SparkCompiler method runSetReducerParallelism.

private void runSetReducerParallelism(OptimizeSparkProcContext procCtx) throws SemanticException {
    ParseContext pCtx = procCtx.getParseContext();
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("Set parallelism - ReduceSink", ReduceSinkOperator.getOperatorName() + "%"), new SetSparkReducerParallelism(pCtx.getConf()));
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(null, opRules, procCtx);
    GraphWalker ogw = new PreOrderWalker(disp);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pCtx.getTopOps().values());
    ogw.startWalking(topNodes, null);
}
Also used : NodeProcessor(org.apache.hadoop.hive.ql.lib.NodeProcessor) 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) Dispatcher(org.apache.hadoop.hive.ql.lib.Dispatcher) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) SetSparkReducerParallelism(org.apache.hadoop.hive.ql.optimizer.spark.SetSparkReducerParallelism) LinkedHashMap(java.util.LinkedHashMap) ParseContext(org.apache.hadoop.hive.ql.parse.ParseContext) Rule(org.apache.hadoop.hive.ql.lib.Rule) TypeRule(org.apache.hadoop.hive.ql.lib.TypeRule) PreOrderWalker(org.apache.hadoop.hive.ql.lib.PreOrderWalker) GraphWalker(org.apache.hadoop.hive.ql.lib.GraphWalker) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker)

Example 20 with DefaultRuleDispatcher

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

the class SparkCompiler method runDynamicPartitionPruning.

private void runDynamicPartitionPruning(OptimizeSparkProcContext procCtx) throws SemanticException {
    if (!conf.getBoolVar(HiveConf.ConfVars.SPARK_DYNAMIC_PARTITION_PRUNING)) {
        return;
    }
    ParseContext parseContext = procCtx.getParseContext();
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp(new String("Dynamic Partition Pruning"), FilterOperator.getOperatorName() + "%"), new DynamicPartitionPruningOptimization());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(null, opRules, procCtx);
    GraphWalker ogw = new ForwardWalker(disp);
    List<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(parseContext.getTopOps().values());
    ogw.startWalking(topNodes, null);
    // of "and true and true" conditions.
    if (procCtx.getConf().getBoolVar(HiveConf.ConfVars.HIVEOPTCONSTANTPROPAGATION)) {
        new ConstantPropagate().transform(parseContext);
    }
}
Also used : NodeProcessor(org.apache.hadoop.hive.ql.lib.NodeProcessor) ForwardWalker(org.apache.hadoop.hive.ql.lib.ForwardWalker) 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) Dispatcher(org.apache.hadoop.hive.ql.lib.Dispatcher) DefaultRuleDispatcher(org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher) LinkedHashMap(java.util.LinkedHashMap) DynamicPartitionPruningOptimization(org.apache.hadoop.hive.ql.optimizer.DynamicPartitionPruningOptimization) ParseContext(org.apache.hadoop.hive.ql.parse.ParseContext) Rule(org.apache.hadoop.hive.ql.lib.Rule) TypeRule(org.apache.hadoop.hive.ql.lib.TypeRule) GraphWalker(org.apache.hadoop.hive.ql.lib.GraphWalker) DefaultGraphWalker(org.apache.hadoop.hive.ql.lib.DefaultGraphWalker) ConstantPropagate(org.apache.hadoop.hive.ql.optimizer.ConstantPropagate)

Aggregations

DefaultRuleDispatcher (org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher)64 Node (org.apache.hadoop.hive.ql.lib.Node)64 GraphWalker (org.apache.hadoop.hive.ql.lib.GraphWalker)63 ArrayList (java.util.ArrayList)62 Dispatcher (org.apache.hadoop.hive.ql.lib.Dispatcher)62 NodeProcessor (org.apache.hadoop.hive.ql.lib.NodeProcessor)55 Rule (org.apache.hadoop.hive.ql.lib.Rule)55 LinkedHashMap (java.util.LinkedHashMap)54 DefaultGraphWalker (org.apache.hadoop.hive.ql.lib.DefaultGraphWalker)49 RuleRegExp (org.apache.hadoop.hive.ql.lib.RuleRegExp)48 HashMap (java.util.HashMap)16 SemanticException (org.apache.hadoop.hive.ql.parse.SemanticException)12 ExprNodeColumnDesc (org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc)11 ExprNodeDesc (org.apache.hadoop.hive.ql.plan.ExprNodeDesc)11 ExprNodeGenericFuncDesc (org.apache.hadoop.hive.ql.plan.ExprNodeGenericFuncDesc)11 TypeRule (org.apache.hadoop.hive.ql.lib.TypeRule)10 ExprNodeConstantDesc (org.apache.hadoop.hive.ql.plan.ExprNodeConstantDesc)7 GenericUDFOPEqualOrGreaterThan (org.apache.hadoop.hive.ql.udf.generic.GenericUDFOPEqualOrGreaterThan)7 Test (org.junit.Test)7 List (java.util.List)6