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

Example 36 with GraphWalker

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

the class SimpleFetchAggregation method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    if (pctx.getFetchTask() != null || !pctx.getQueryProperties().isQuery() || pctx.getQueryProperties().isAnalyzeRewrite() || pctx.getQueryProperties().isCTAS() || pctx.getLoadFileWork().size() > 1 || !pctx.getLoadTableWork().isEmpty()) {
        return pctx;
    }
    String GBY = GroupByOperator.getOperatorName() + "%";
    String RS = ReduceSinkOperator.getOperatorName() + "%";
    String SEL = SelectOperator.getOperatorName() + "%";
    String FS = FileSinkOperator.getOperatorName() + "%";
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", GBY + RS + GBY + SEL + FS), new SingleGBYProcessor(pctx));
    opRules.put(new RuleRegExp("R2", GBY + RS + GBY + FS), new SingleGBYProcessor(pctx));
    Dispatcher disp = new DefaultRuleDispatcher(null, opRules, null);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 37 with GraphWalker

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

the class GroupByOptimizer method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    HiveConf conf = pctx.getConf();
    if (!HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVEGROUPBYSKEW)) {
        // process group-by pattern
        opRules.put(new RuleRegExp("R1", GroupByOperator.getOperatorName() + "%" + ReduceSinkOperator.getOperatorName() + "%" + GroupByOperator.getOperatorName() + "%"), getMapSortedGroupbyProc(pctx));
    } else {
        // If hive.groupby.skewindata is set to true, the operator tree is as below
        opRules.put(new RuleRegExp("R2", GroupByOperator.getOperatorName() + "%" + ReduceSinkOperator.getOperatorName() + "%" + GroupByOperator.getOperatorName() + "%" + ReduceSinkOperator.getOperatorName() + "%" + GroupByOperator.getOperatorName() + "%"), getMapSortedGroupbySkewProc(pctx));
    }
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(getDefaultProc(), opRules, new GroupByOptimizerContext(conf));
    GraphWalker ogw = new DefaultGraphWalker(disp);
    // Create a list of topop nodes
    List<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 38 with GraphWalker

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

the class IdentityProjectRemover method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    // 0. We check the conditions to apply this transformation,
    //    if we do not meet them we bail out
    final boolean cboEnabled = HiveConf.getBoolVar(pctx.getConf(), HiveConf.ConfVars.HIVE_CBO_ENABLED);
    final boolean returnPathEnabled = HiveConf.getBoolVar(pctx.getConf(), HiveConf.ConfVars.HIVE_CBO_RETPATH_HIVEOP);
    final boolean cboSucceeded = pctx.getContext().isCboSucceeded();
    if (cboEnabled && returnPathEnabled && cboSucceeded) {
        return pctx;
    }
    // 1. We apply the transformation
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", "(" + SelectOperator.getOperatorName() + "%)"), new ProjectRemover());
    GraphWalker ogw = new DefaultGraphWalker(new DefaultRuleDispatcher(null, opRules, null));
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 39 with GraphWalker

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

the class LimitPushdownOptimizer method transform.

@Override
public ParseContext transform(ParseContext pctx) throws SemanticException {
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", ReduceSinkOperator.getOperatorName() + "%" + ".*" + LimitOperator.getOperatorName() + "%"), new TopNReducer());
    opRules.put(new RuleRegExp("R2", ReduceSinkOperator.getOperatorName() + "%" + ".*" + ReduceSinkOperator.getOperatorName() + "%"), new TopNPropagator());
    LimitPushdownContext context = new LimitPushdownContext(pctx.getConf());
    Dispatcher disp = new DefaultRuleDispatcher(null, opRules, context);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    List<Node> topNodes = new ArrayList<Node>(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}

Example 40 with GraphWalker

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

the class DynamicPartitionPruningOptimization method collectDynamicPruningConditions.

private Map<Node, Object> collectDynamicPruningConditions(ExprNodeDesc pred, NodeProcessorCtx ctx) 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 RuleRegExp("R1", ExprNodeDynamicListDesc.class.getName() + "%"), new DynamicPartitionPrunerProc());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(null, 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;
}