Search in sources :

Example 66 with SemanticNodeProcessor

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

the class UnionProcessor method transform.

/**
 * Transform the query tree. For each union, store the fact whether both the
 * sub-queries are map-only
 *
 * @param pCtx
 *          the current parse context
 */
public ParseContext transform(ParseContext pCtx) throws SemanticException {
    // create a walker which walks the tree in a BFS manner while maintaining
    // the operator stack.
    Map<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    opRules.put(new RuleRegExp("R1", ReduceSinkOperator.getOperatorName() + "%.*" + UnionOperator.getOperatorName() + "%"), UnionProcFactory.getMapRedUnion());
    opRules.put(new RuleRegExp("R2", UnionOperator.getOperatorName() + "%.*" + UnionOperator.getOperatorName() + "%"), UnionProcFactory.getUnknownUnion());
    opRules.put(new RuleRegExp("R3", TableScanOperator.getOperatorName() + "%.*" + UnionOperator.getOperatorName() + "%"), UnionProcFactory.getMapUnion());
    // The dispatcher fires the processor for the matching rule and passes the
    // context along
    UnionProcContext uCtx = new UnionProcContext();
    uCtx.setParseContext(pCtx);
    SemanticDispatcher disp = new DefaultRuleDispatcher(UnionProcFactory.getNoUnion(), opRules, uCtx);
    LevelOrderWalker ogw = new LevelOrderWalker(disp);
    ogw.setNodeTypes(UnionOperator.class);
    // Create a list of topop nodes
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pCtx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    pCtx.setUCtx(uCtx);
    // Walk the tree again to see if the union can be removed completely
    HiveConf conf = pCtx.getConf();
    opRules.clear();
    if (conf.getBoolVar(HiveConf.ConfVars.HIVE_OPTIMIZE_UNION_REMOVE) && !conf.getVar(HiveConf.ConfVars.HIVE_EXECUTION_ENGINE).equals("spark")) {
        opRules.put(new RuleRegExp("R5", UnionOperator.getOperatorName() + "%" + ".*" + FileSinkOperator.getOperatorName() + "%"), UnionProcFactory.getUnionNoProcessFile());
        disp = new DefaultRuleDispatcher(UnionProcFactory.getNoUnion(), opRules, uCtx);
        ogw = new LevelOrderWalker(disp);
        ogw.setNodeTypes(FileSinkOperator.class);
        // Create a list of topop nodes
        topNodes.clear();
        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) LinkedHashMap(java.util.LinkedHashMap) SemanticDispatcher(org.apache.hadoop.hive.ql.lib.SemanticDispatcher) SemanticNodeProcessor(org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor) HiveConf(org.apache.hadoop.hive.conf.HiveConf) LevelOrderWalker(org.apache.hadoop.hive.ql.lib.LevelOrderWalker)

Example 67 with SemanticNodeProcessor

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

the class NullScanOptimizer method resolve.

@Override
public PhysicalContext resolve(PhysicalContext pctx) throws SemanticException {
    Map<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<>();
    opRules.put(new RuleRegExp("R1", TableScanOperator.getOperatorName() + "%.*" + FilterOperator.getOperatorName() + "%"), new WhereFalseProcessor());
    SemanticDispatcher disp = new NullScanTaskDispatcher(pctx, opRules);
    SemanticGraphWalker ogw = new DefaultGraphWalker(disp);
    List<Node> topNodes = new ArrayList<>(pctx.getRootTasks());
    ogw.startWalking(topNodes, null);
    opRules.clear();
    opRules.put(new RuleRegExp("R1", TableScanOperator.getOperatorName() + "%"), new TSMarker());
    opRules.put(new RuleRegExp("R2", LimitOperator.getOperatorName() + "%"), new Limit0Processor());
    disp = new NullScanTaskDispatcher(pctx, opRules);
    ogw = new DefaultGraphWalker(disp);
    topNodes = new ArrayList<>(pctx.getRootTasks());
    ogw.startWalking(topNodes, null);
    return pctx;
}
Also used : SemanticRule(org.apache.hadoop.hive.ql.lib.SemanticRule) 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 68 with SemanticNodeProcessor

use of org.apache.hadoop.hive.ql.lib.SemanticNodeProcessor 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<SemanticRule, SemanticNodeProcessor> opRules = new LinkedHashMap<SemanticRule, SemanticNodeProcessor>();
    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());
    opRules.put(new RuleRegExp("UDTF", UDTFOperator.getOperatorName() + "%"), StatsRulesProcFactory.getUDTFRule());
    opRules.put(new RuleRegExp("LVJ", LateralViewJoinOperator.getOperatorName() + "%"), StatsRulesProcFactory.getLateralViewJoinRule());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    SemanticDispatcher disp = new DefaultRuleDispatcher(StatsRulesProcFactory.getDefaultRule(), opRules, aspCtx);
    SemanticGraphWalker 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 : 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) LevelOrderWalker(org.apache.hadoop.hive.ql.lib.LevelOrderWalker)

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