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

Example 51 with GraphWalker

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

the class TestAccumuloRangeGenerator method testRangeDisjunction.

@Test
public void testRangeDisjunction() throws Exception {
    // rowId >= 'f'
    ExprNodeDesc column = new ExprNodeColumnDesc(TypeInfoFactory.stringTypeInfo, "rid", null, false);
    ExprNodeDesc constant = new ExprNodeConstantDesc(TypeInfoFactory.stringTypeInfo, "f");
    List<ExprNodeDesc> children = Lists.newArrayList();
    children.add(column);
    children.add(constant);
    ExprNodeDesc node = new ExprNodeGenericFuncDesc(TypeInfoFactory.stringTypeInfo, new GenericUDFOPEqualOrGreaterThan(), children);
    assertNotNull(node);
    // rowId <= 'm'
    ExprNodeDesc column2 = new ExprNodeColumnDesc(TypeInfoFactory.stringTypeInfo, "rid", null, false);
    ExprNodeDesc constant2 = new ExprNodeConstantDesc(TypeInfoFactory.stringTypeInfo, "m");
    List<ExprNodeDesc> children2 = Lists.newArrayList();
    children2.add(column2);
    children2.add(constant2);
    ExprNodeDesc node2 = new ExprNodeGenericFuncDesc(TypeInfoFactory.stringTypeInfo, new GenericUDFOPEqualOrLessThan(), children2);
    assertNotNull(node2);
    // Or UDF
    List<ExprNodeDesc> bothFilters = Lists.newArrayList();
    bothFilters.add(node);
    bothFilters.add(node2);
    ExprNodeGenericFuncDesc both = new ExprNodeGenericFuncDesc(TypeInfoFactory.stringTypeInfo, new GenericUDFOPOr(), bothFilters);
    // Should generate (-inf,+inf)
    List<Range> expectedRanges = Arrays.asList(new Range());
    AccumuloRangeGenerator rangeGenerator = new AccumuloRangeGenerator(handler, rowIdMapping, "rid");
    Dispatcher disp = new DefaultRuleDispatcher(rangeGenerator, Collections.<Rule, NodeProcessor>emptyMap(), null);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.add(both);
    HashMap<Node, Object> nodeOutput = new HashMap<Node, Object>();
    try {
        ogw.startWalking(topNodes, nodeOutput);
    } catch (SemanticException ex) {
        throw new RuntimeException(ex);
    }
    Object result = nodeOutput.get(both);
    Assert.assertNotNull(result);
    Assert.assertTrue("Result from graph walk was not a List", result instanceof List);
    @SuppressWarnings("unchecked") List<Range> actualRanges = (List<Range>) result;
    Assert.assertEquals(expectedRanges, actualRanges);
}

Example 52 with GraphWalker

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

the class IndexWhereTaskDispatcher method dispatch.

@Override
public Object dispatch(Node nd, Stack<Node> stack, Object... nodeOutputs) throws SemanticException {
    Task<? extends Serializable> task = (Task<? extends Serializable>) nd;
    ParseContext pctx = physicalContext.getParseContext();
    // create the regex's so the walker can recognize our WHERE queries
    Map<Rule, NodeProcessor> operatorRules = createOperatorRules(pctx);
    // check for no indexes on any table
    if (operatorRules == null) {
        return null;
    }
    // create context so the walker can carry the current task with it.
    IndexWhereProcCtx indexWhereOptimizeCtx = new IndexWhereProcCtx(task, pctx);
    // create the dispatcher, which fires the processor according to the rule that
    // best matches
    Dispatcher dispatcher = new DefaultRuleDispatcher(getDefaultProcessor(), operatorRules, indexWhereOptimizeCtx);
    // walk the mapper operator(not task) tree for each specific task
    GraphWalker ogw = new DefaultGraphWalker(dispatcher);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    if (task.getWork() instanceof MapredWork) {
        topNodes.addAll(((MapredWork) task.getWork()).getMapWork().getAliasToWork().values());
    } else {
        return null;
    }
    ogw.startWalking(topNodes, null);
    return null;
}

Example 53 with GraphWalker

use of org.apache.hadoop.hive.ql.lib.GraphWalker 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<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    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());
    // The dispatcher fires the processor corresponding to the closest matching
    // rule and passes the context along
    Dispatcher disp = new DefaultRuleDispatcher(StatsRulesProcFactory.getDefaultRule(), opRules, aspCtx);
    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 54 with GraphWalker

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

the class IndexPredicateAnalyzer method analyzePredicate.

/**
   * Analyzes a predicate.
   *
   * @param predicate predicate to be analyzed
   *
   * @param searchConditions receives conditions produced by analysis
   *
   * @return residual predicate which could not be translated to
   * searchConditions
   */
public ExprNodeDesc analyzePredicate(ExprNodeDesc predicate, final List<IndexSearchCondition> searchConditions) {
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    NodeProcessor nodeProcessor = new NodeProcessor() {

        @Override
        public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procCtx, Object... nodeOutputs) throws SemanticException {
            // a pure conjunction:  reject OR, CASE, etc.
            for (Node ancestor : stack) {
                if (nd == ancestor) {
                    break;
                }
                if (!FunctionRegistry.isOpAnd((ExprNodeDesc) ancestor)) {
                    return nd;
                }
            }
            return analyzeExpr((ExprNodeGenericFuncDesc) nd, searchConditions, nodeOutputs);
        }
    };
    Dispatcher disp = new DefaultRuleDispatcher(nodeProcessor, opRules, null);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> topNodes = new ArrayList<Node>();
    topNodes.add(predicate);
    HashMap<Node, Object> nodeOutput = new HashMap<Node, Object>();
    try {
        ogw.startWalking(topNodes, nodeOutput);
    } catch (SemanticException ex) {
        throw new RuntimeException(ex);
    }
    ExprNodeDesc residualPredicate = (ExprNodeDesc) nodeOutput.get(predicate);
    return residualPredicate;
}

Example 55 with GraphWalker

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

the class HiveOpConverterPostProc 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. Initialize aux data structures
    this.pctx = pctx;
    this.aliasToOpInfo = new HashMap<String, Operator<? extends OperatorDesc>>();
    // 2. Trigger transformation
    Map<Rule, NodeProcessor> opRules = new LinkedHashMap<Rule, NodeProcessor>();
    opRules.put(new RuleRegExp("R1", JoinOperator.getOperatorName() + "%"), new JoinAnnotate());
    opRules.put(new RuleRegExp("R2", TableScanOperator.getOperatorName() + "%"), new TableScanAnnotate());
    Dispatcher disp = new DefaultRuleDispatcher(null, opRules, null);
    GraphWalker ogw = new ForwardWalker(disp);
    List<Node> topNodes = new ArrayList<Node>();
    topNodes.addAll(pctx.getTopOps().values());
    ogw.startWalking(topNodes, null);
    return pctx;
}