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

Example 11 with RuleRegExp

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

the class ReduceSinkDeDuplication method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    pGraphContext = pctx;
    // generate pruned column list for all relevant operators
    ReduceSinkDeduplicateProcCtx cppCtx = new ReduceSinkDeduplicateProcCtx(pGraphContext);
    // for auto convert map-joins, it not safe to dedup in here (todo)
    boolean mergeJoins = !pctx.getConf().getBoolVar(HIVECONVERTJOIN) && !pctx.getConf().getBoolVar(HIVECONVERTJOINNOCONDITIONALTASK) && !pctx.getConf().getBoolVar(ConfVars.HIVE_CONVERT_JOIN_BUCKET_MAPJOIN_TEZ) && !pctx.getConf().getBoolVar(ConfVars.HIVEDYNAMICPARTITIONHASHJOIN);
    // If multiple rules can be matched with same cost, last rule will be choosen as a processor
    // see DefaultRuleDispatcher#dispatch()
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", RS + "%.*%" + RS + "%"), ReduceSinkDeduplicateProcFactory.getReducerReducerProc());
    opRules.put(new RuleRegExp("R2", RS + "%" + GBY + "%.*%" + RS + "%"), ReduceSinkDeduplicateProcFactory.getGroupbyReducerProc());
    if (mergeJoins) {
        opRules.put(new RuleRegExp("R3", JOIN + "%.*%" + RS + "%"), ReduceSinkDeduplicateProcFactory.getJoinReducerProc());
    }
    // TODO RS+JOIN
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(ReduceSinkDeduplicateProcFactory.getDefaultProc(), opRules, cppCtx);
    GraphWalker 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);
    return pGraphContext;
}

Example 12 with RuleRegExp

use of org.apache.hadoop.hive.ql.lib.RuleRegExp 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;
}

Example 13 with RuleRegExp

use of org.apache.hadoop.hive.ql.lib.RuleRegExp 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;
}

Example 14 with RuleRegExp

use of org.apache.hadoop.hive.ql.lib.RuleRegExp 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);
}

Example 15 with RuleRegExp

use of org.apache.hadoop.hive.ql.lib.RuleRegExp 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);
}