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

Example 36 with DefaultGraphWalker

use of org.apache.hadoop.hive.ql.lib.DefaultGraphWalker 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 37 with DefaultGraphWalker

use of org.apache.hadoop.hive.ql.lib.DefaultGraphWalker 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 38 with DefaultGraphWalker

use of org.apache.hadoop.hive.ql.lib.DefaultGraphWalker 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 39 with DefaultGraphWalker

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

the class SemanticAnalyzer method replaceViewReferenceWithDefinition.

private void replaceViewReferenceWithDefinition(QB qb, Table tab, String tab_name, String alias) throws SemanticException {
    ASTNode viewTree;
    final ASTNodeOrigin viewOrigin = new ASTNodeOrigin("VIEW", tab.getTableName(), tab.getViewExpandedText(), alias, qb.getParseInfo().getSrcForAlias(alias));
    try {
        String viewText = tab.getViewExpandedText();
        // Reparse text, passing null for context to avoid clobbering
        // the top-level token stream.
        ASTNode tree = ParseUtils.parse(viewText, ctx, false);
        viewTree = tree;
        Dispatcher nodeOriginDispatcher = new Dispatcher() {

            @Override
            public Object dispatch(Node nd, java.util.Stack<Node> stack, Object... nodeOutputs) {
                ((ASTNode) nd).setOrigin(viewOrigin);
                return null;
            }
        };
        GraphWalker nodeOriginTagger = new DefaultGraphWalker(nodeOriginDispatcher);
        nodeOriginTagger.startWalking(java.util.Collections.<Node>singleton(viewTree), null);
    } catch (ParseException e) {
        // A user could encounter this if a stored view definition contains
        // an old SQL construct which has been eliminated in a later Hive
        // version, so we need to provide full debugging info to help
        // with fixing the view definition.
        LOG.error(org.apache.hadoop.util.StringUtils.stringifyException(e));
        StringBuilder sb = new StringBuilder();
        sb.append(e.getMessage());
        ErrorMsg.renderOrigin(sb, viewOrigin);
        throw new SemanticException(sb.toString(), e);
    }
    QBExpr qbexpr = new QBExpr(alias);
    doPhase1QBExpr(viewTree, qbexpr, qb.getId(), alias, true);
    // if HIVE_STATS_COLLECT_SCANCOLS is enabled, check.
    if ((!this.skipAuthorization() && !qb.isInsideView() && HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVE_AUTHORIZATION_ENABLED)) || HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVE_STATS_COLLECT_SCANCOLS)) {
        qb.rewriteViewToSubq(alias, tab_name, qbexpr, tab);
    } else {
        qb.rewriteViewToSubq(alias, tab_name, qbexpr, null);
    }
}

Example 40 with DefaultGraphWalker

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

the class AccumuloPredicateHandler method generateRanges.

/**
   * Encapsulates the traversal over some {@link ExprNodeDesc} tree for the generation of Accumuluo
   * Ranges using expressions involving the Accumulo rowid-mapped Hive column
   *
   * @param columnMapper
   *          Mapping of Hive to Accumulo columns for the query
   * @param hiveRowIdColumnName
   *          Name of the hive column mapped to the Accumulo rowid
   * @param root
   *          Root of some ExprNodeDesc tree to traverse, the WHERE clause
   * @return An object representing the result from the ExprNodeDesc tree traversal using the
   *         AccumuloRangeGenerator
   */
protected Object generateRanges(ColumnMapper columnMapper, String hiveRowIdColumnName, ExprNodeDesc root) {
    AccumuloRangeGenerator rangeGenerator = new AccumuloRangeGenerator(handler, columnMapper.getRowIdMapping(), hiveRowIdColumnName);
    Dispatcher disp = new DefaultRuleDispatcher(rangeGenerator, Collections.<Rule, NodeProcessor>emptyMap(), null);
    GraphWalker ogw = new DefaultGraphWalker(disp);
    ArrayList<Node> roots = new ArrayList<Node>();
    roots.add(root);
    HashMap<Node, Object> nodeOutput = new HashMap<Node, Object>();
    try {
        ogw.startWalking(roots, nodeOutput);
    } catch (SemanticException ex) {
        throw new RuntimeException(ex);
    }
    return nodeOutput.get(root);
}