Search in sources :

Example 66 with SemanticException

use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.

the class Registry method getQualifiedFunctionInfoUnderLock.

private FunctionInfo getQualifiedFunctionInfoUnderLock(String qualifiedName) throws SemanticException {
    FunctionInfo info = mFunctions.get(qualifiedName);
    if (info != null && info.isBlockedFunction()) {
        throw new SemanticException("UDF " + qualifiedName + " is not allowed");
    }
    if (!isNative && info != null && info.isDiscarded()) {
        // the persistent function is discarded. try reload
        mFunctions.remove(qualifiedName);
        return null;
    }
    // and if necessary load the JARs in this thread.
    if (isNative && info != null && info.isPersistent()) {
        return registerToSessionRegistry(qualifiedName, info);
    }
    if (info != null || !isNative) {
        // We have the UDF, or we are in the session registry (or both).
        return info;
    }
    // If we are in the system registry and this feature is enabled, try to get it from metastore.
    SessionState ss = SessionState.get();
    HiveConf conf = (ss == null) ? null : ss.getConf();
    if (conf == null || !HiveConf.getBoolVar(conf, ConfVars.HIVE_ALLOW_UDF_LOAD_ON_DEMAND)) {
        return null;
    }
    // This is a little bit weird. We'll do the MS call outside of the lock. Our caller calls us
    // under lock, so we'd preserve the lock state for them; their finally block will release the
    // lock correctly. See the comment on the lock field - the locking needs to be reworked.
    lock.unlock();
    try {
        return getFunctionInfoFromMetastoreNoLock(qualifiedName, conf);
    } finally {
        lock.lock();
    }
}
Also used : SessionState(org.apache.hadoop.hive.ql.session.SessionState) HiveConf(org.apache.hadoop.hive.conf.HiveConf) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException)

Example 67 with SemanticException

use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.

the class IndexWhereProcessor method process.

@Override
public /**
   * Process a node of the operator tree. This matches on the rule in IndexWhereTaskDispatcher
   */
Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procCtx, Object... nodeOutputs) throws SemanticException {
    TableScanOperator operator = (TableScanOperator) nd;
    List<Node> opChildren = operator.getChildren();
    TableScanDesc operatorDesc = operator.getConf();
    if (operatorDesc == null || !tsToIndices.containsKey(operator)) {
        return null;
    }
    List<Index> indexes = tsToIndices.get(operator);
    ExprNodeDesc predicate = operatorDesc.getFilterExpr();
    IndexWhereProcCtx context = (IndexWhereProcCtx) procCtx;
    ParseContext pctx = context.getParseContext();
    LOG.info("Processing predicate for index optimization");
    if (predicate == null) {
        LOG.info("null predicate pushed down");
        return null;
    }
    LOG.info(predicate.getExprString());
    // check if we have tsToIndices on all partitions in this table scan
    Set<Partition> queryPartitions;
    try {
        queryPartitions = IndexUtils.checkPartitionsCoveredByIndex(operator, pctx, indexes);
        if (queryPartitions == null) {
            // partitions not covered
            return null;
        }
    } catch (HiveException e) {
        LOG.error("Fatal Error: problem accessing metastore", e);
        throw new SemanticException(e);
    }
    // we can only process MapReduce tasks to check input size
    if (!context.getCurrentTask().isMapRedTask()) {
        return null;
    }
    MapRedTask currentTask = (MapRedTask) context.getCurrentTask();
    // get potential reentrant index queries from each index
    Map<Index, HiveIndexQueryContext> queryContexts = new HashMap<Index, HiveIndexQueryContext>();
    // make sure we have an index on the table being scanned
    TableDesc tblDesc = operator.getTableDesc();
    Map<String, List<Index>> indexesByType = new HashMap<String, List<Index>>();
    for (Index indexOnTable : indexes) {
        if (indexesByType.get(indexOnTable.getIndexHandlerClass()) == null) {
            List<Index> newType = new ArrayList<Index>();
            newType.add(indexOnTable);
            indexesByType.put(indexOnTable.getIndexHandlerClass(), newType);
        } else {
            indexesByType.get(indexOnTable.getIndexHandlerClass()).add(indexOnTable);
        }
    }
    // choose index type with most tsToIndices of the same type on the table
    // TODO HIVE-2130 This would be a good place for some sort of cost based choice?
    List<Index> bestIndexes = indexesByType.values().iterator().next();
    for (List<Index> indexTypes : indexesByType.values()) {
        if (bestIndexes.size() < indexTypes.size()) {
            bestIndexes = indexTypes;
        }
    }
    // rewrite index queries for the chosen index type
    HiveIndexQueryContext tmpQueryContext = new HiveIndexQueryContext();
    tmpQueryContext.setQueryPartitions(queryPartitions);
    rewriteForIndexes(predicate, bestIndexes, pctx, currentTask, tmpQueryContext);
    List<Task<?>> indexTasks = tmpQueryContext.getQueryTasks();
    if (indexTasks != null && indexTasks.size() > 0) {
        queryContexts.put(bestIndexes.get(0), tmpQueryContext);
    }
    // choose an index rewrite to use
    if (queryContexts.size() > 0) {
        // TODO HIVE-2130 This would be a good place for some sort of cost based choice?
        Index chosenIndex = queryContexts.keySet().iterator().next();
        // modify the parse context to use indexing
        // we need to delay this until we choose one index so that we don't attempt to modify pctx multiple times
        HiveIndexQueryContext queryContext = queryContexts.get(chosenIndex);
        // prepare the map reduce job to use indexing
        MapWork work = currentTask.getWork().getMapWork();
        work.setInputformat(queryContext.getIndexInputFormat());
        work.addIndexIntermediateFile(queryContext.getIndexIntermediateFile());
        // modify inputs based on index query
        Set<ReadEntity> inputs = pctx.getSemanticInputs();
        inputs.addAll(queryContext.getAdditionalSemanticInputs());
        List<Task<?>> chosenRewrite = queryContext.getQueryTasks();
        // add dependencies so index query runs first
        insertIndexQuery(pctx, context, chosenRewrite);
    }
    return null;
}
Also used : HiveIndexQueryContext(org.apache.hadoop.hive.ql.index.HiveIndexQueryContext) TableScanOperator(org.apache.hadoop.hive.ql.exec.TableScanOperator) Task(org.apache.hadoop.hive.ql.exec.Task) MapRedTask(org.apache.hadoop.hive.ql.exec.mr.MapRedTask) HiveException(org.apache.hadoop.hive.ql.metadata.HiveException) HashMap(java.util.HashMap) Node(org.apache.hadoop.hive.ql.lib.Node) ArrayList(java.util.ArrayList) Index(org.apache.hadoop.hive.metastore.api.Index) MapRedTask(org.apache.hadoop.hive.ql.exec.mr.MapRedTask) ArrayList(java.util.ArrayList) List(java.util.List) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException) Partition(org.apache.hadoop.hive.ql.metadata.Partition) TableScanDesc(org.apache.hadoop.hive.ql.plan.TableScanDesc) ReadEntity(org.apache.hadoop.hive.ql.hooks.ReadEntity) MapWork(org.apache.hadoop.hive.ql.plan.MapWork) ParseContext(org.apache.hadoop.hive.ql.parse.ParseContext) TableDesc(org.apache.hadoop.hive.ql.plan.TableDesc)

Example 68 with SemanticException

use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.

the class TableBasedIndexHandler method generateIndexBuildTaskList.

@Override
public List<Task<?>> generateIndexBuildTaskList(org.apache.hadoop.hive.ql.metadata.Table baseTbl, org.apache.hadoop.hive.metastore.api.Index index, List<Partition> indexTblPartitions, List<Partition> baseTblPartitions, org.apache.hadoop.hive.ql.metadata.Table indexTbl, Set<ReadEntity> inputs, Set<WriteEntity> outputs) throws HiveException {
    try {
        TableDesc desc = Utilities.getTableDesc(indexTbl);
        List<Partition> newBaseTblPartitions = new ArrayList<Partition>();
        List<Task<?>> indexBuilderTasks = new ArrayList<Task<?>>();
        if (!baseTbl.isPartitioned()) {
            // the table does not have any partition, then create index for the
            // whole table
            Task<?> indexBuilder = getIndexBuilderMapRedTask(inputs, outputs, index, false, new PartitionDesc(desc, null), indexTbl.getTableName(), new PartitionDesc(Utilities.getTableDesc(baseTbl), null), baseTbl.getTableName(), indexTbl.getDbName());
            indexBuilderTasks.add(indexBuilder);
        } else {
            // table
            for (int i = 0; i < indexTblPartitions.size(); i++) {
                Partition indexPart = indexTblPartitions.get(i);
                Partition basePart = null;
                for (int j = 0; j < baseTblPartitions.size(); j++) {
                    if (baseTblPartitions.get(j).getName().equals(indexPart.getName())) {
                        basePart = baseTblPartitions.get(j);
                        newBaseTblPartitions.add(baseTblPartitions.get(j));
                        break;
                    }
                }
                if (basePart == null) {
                    throw new RuntimeException("Partitions of base table and index table are inconsistent.");
                }
                // for each partition, spawn a map reduce task.
                Task<?> indexBuilder = getIndexBuilderMapRedTask(inputs, outputs, index, true, new PartitionDesc(indexPart), indexTbl.getTableName(), new PartitionDesc(basePart), baseTbl.getTableName(), indexTbl.getDbName());
                indexBuilderTasks.add(indexBuilder);
            }
        }
        return indexBuilderTasks;
    } catch (Exception e) {
        throw new SemanticException(e);
    }
}
Also used : Partition(org.apache.hadoop.hive.ql.metadata.Partition) Task(org.apache.hadoop.hive.ql.exec.Task) ArrayList(java.util.ArrayList) PartitionDesc(org.apache.hadoop.hive.ql.plan.PartitionDesc) TableDesc(org.apache.hadoop.hive.ql.plan.TableDesc) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException) HiveException(org.apache.hadoop.hive.ql.metadata.HiveException) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException)

Example 69 with SemanticException

use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.

the class HiveOpConverter method visit.

OpAttr visit(HiveSortLimit sortRel) throws SemanticException {
    OpAttr inputOpAf = dispatch(sortRel.getInput());
    if (LOG.isDebugEnabled()) {
        LOG.debug("Translating operator rel#" + sortRel.getId() + ":" + sortRel.getRelTypeName() + " with row type: [" + sortRel.getRowType() + "]");
        if (sortRel.getCollation() == RelCollations.EMPTY) {
            LOG.debug("Operator rel#" + sortRel.getId() + ":" + sortRel.getRelTypeName() + " consists of limit");
        } else if (sortRel.fetch == null) {
            LOG.debug("Operator rel#" + sortRel.getId() + ":" + sortRel.getRelTypeName() + " consists of sort");
        } else {
            LOG.debug("Operator rel#" + sortRel.getId() + ":" + sortRel.getRelTypeName() + " consists of sort+limit");
        }
    }
    Operator<?> inputOp = inputOpAf.inputs.get(0);
    Operator<?> resultOp = inputOpAf.inputs.get(0);
    // of their columns
    if (sortRel.getCollation() != RelCollations.EMPTY) {
        // In strict mode, in the presence of order by, limit must be specified.
        if (sortRel.fetch == null) {
            String error = StrictChecks.checkNoLimit(hiveConf);
            if (error != null)
                throw new SemanticException(error);
        }
        // 1.a. Extract order for each column from collation
        // Generate sortCols and order
        ImmutableBitSet.Builder sortColsPosBuilder = ImmutableBitSet.builder();
        ImmutableBitSet.Builder sortOutputColsPosBuilder = ImmutableBitSet.builder();
        Map<Integer, RexNode> obRefToCallMap = sortRel.getInputRefToCallMap();
        List<ExprNodeDesc> sortCols = new ArrayList<ExprNodeDesc>();
        StringBuilder order = new StringBuilder();
        StringBuilder nullOrder = new StringBuilder();
        for (RelFieldCollation sortInfo : sortRel.getCollation().getFieldCollations()) {
            int sortColumnPos = sortInfo.getFieldIndex();
            ColumnInfo columnInfo = new ColumnInfo(inputOp.getSchema().getSignature().get(sortColumnPos));
            ExprNodeColumnDesc sortColumn = new ExprNodeColumnDesc(columnInfo.getType(), columnInfo.getInternalName(), columnInfo.getTabAlias(), columnInfo.getIsVirtualCol());
            sortCols.add(sortColumn);
            if (sortInfo.getDirection() == RelFieldCollation.Direction.DESCENDING) {
                order.append("-");
            } else {
                order.append("+");
            }
            if (sortInfo.nullDirection == RelFieldCollation.NullDirection.FIRST) {
                nullOrder.append("a");
            } else if (sortInfo.nullDirection == RelFieldCollation.NullDirection.LAST) {
                nullOrder.append("z");
            } else {
                // Default
                nullOrder.append(sortInfo.getDirection() == RelFieldCollation.Direction.DESCENDING ? "z" : "a");
            }
            if (obRefToCallMap != null) {
                RexNode obExpr = obRefToCallMap.get(sortColumnPos);
                sortColsPosBuilder.set(sortColumnPos);
                if (obExpr == null) {
                    sortOutputColsPosBuilder.set(sortColumnPos);
                }
            }
        }
        // Use only 1 reducer for order by
        int numReducers = 1;
        // We keep the columns only the columns that are part of the final output
        List<String> keepColumns = new ArrayList<String>();
        final ImmutableBitSet sortColsPos = sortColsPosBuilder.build();
        final ImmutableBitSet sortOutputColsPos = sortOutputColsPosBuilder.build();
        final ArrayList<ColumnInfo> inputSchema = inputOp.getSchema().getSignature();
        for (int pos = 0; pos < inputSchema.size(); pos++) {
            if ((sortColsPos.get(pos) && sortOutputColsPos.get(pos)) || (!sortColsPos.get(pos) && !sortOutputColsPos.get(pos))) {
                keepColumns.add(inputSchema.get(pos).getInternalName());
            }
        }
        // 1.b. Generate reduce sink and project operator
        resultOp = genReduceSinkAndBacktrackSelect(resultOp, sortCols.toArray(new ExprNodeDesc[sortCols.size()]), 0, new ArrayList<ExprNodeDesc>(), order.toString(), nullOrder.toString(), numReducers, Operation.NOT_ACID, hiveConf, keepColumns);
    }
    // 2. If we need to generate limit
    if (sortRel.fetch != null) {
        int limit = RexLiteral.intValue(sortRel.fetch);
        int offset = sortRel.offset == null ? 0 : RexLiteral.intValue(sortRel.offset);
        LimitDesc limitDesc = new LimitDesc(offset, limit);
        ArrayList<ColumnInfo> cinfoLst = createColInfos(resultOp);
        resultOp = OperatorFactory.getAndMakeChild(limitDesc, new RowSchema(cinfoLst), resultOp);
        if (LOG.isDebugEnabled()) {
            LOG.debug("Generated " + resultOp + " with row schema: [" + resultOp.getSchema() + "]");
        }
    }
    // 3. Return result
    return inputOpAf.clone(resultOp);
}
Also used : RowSchema(org.apache.hadoop.hive.ql.exec.RowSchema) ImmutableBitSet(org.apache.calcite.util.ImmutableBitSet) ArrayList(java.util.ArrayList) ColumnInfo(org.apache.hadoop.hive.ql.exec.ColumnInfo) LimitDesc(org.apache.hadoop.hive.ql.plan.LimitDesc) RelFieldCollation(org.apache.calcite.rel.RelFieldCollation) ExprNodeColumnDesc(org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException) RexNode(org.apache.calcite.rex.RexNode)

Example 70 with SemanticException

use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.

the class HiveOpConverter method updateExprNode.

/*
   * This method updates the input expr, changing all the
   * ExprNodeColumnDesc in it to refer to columns given by the
   * colExprMap.
   *
   * For instance, "col_0 = 1" would become "VALUE.col_0 = 1";
   * the execution engine expects filters in the Join operators
   * to be expressed that way.
   */
private static int updateExprNode(ExprNodeDesc expr, final Map<String, Byte> reversedExprs, final Map<String, ExprNodeDesc> colExprMap) throws SemanticException {
    int inputPos = -1;
    if (expr instanceof ExprNodeGenericFuncDesc) {
        ExprNodeGenericFuncDesc func = (ExprNodeGenericFuncDesc) expr;
        List<ExprNodeDesc> newChildren = new ArrayList<ExprNodeDesc>();
        for (ExprNodeDesc functionChild : func.getChildren()) {
            if (functionChild instanceof ExprNodeColumnDesc) {
                String colRef = functionChild.getExprString();
                int pos = reversedExprs.get(colRef);
                if (pos != -1) {
                    if (inputPos == -1) {
                        inputPos = pos;
                    } else if (inputPos != pos) {
                        throw new SemanticException("UpdateExprNode is expecting only one position for join operator convert. But there are more than one.");
                    }
                }
                newChildren.add(colExprMap.get(colRef));
            } else {
                int pos = updateExprNode(functionChild, reversedExprs, colExprMap);
                if (pos != -1) {
                    if (inputPos == -1) {
                        inputPos = pos;
                    } else if (inputPos != pos) {
                        throw new SemanticException("UpdateExprNode is expecting only one position for join operator convert. But there are more than one.");
                    }
                }
                newChildren.add(functionChild);
            }
        }
        func.setChildren(newChildren);
    }
    return inputPos;
}
Also used : ArrayList(java.util.ArrayList) ExprNodeColumnDesc(org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc) ExprNodeGenericFuncDesc(org.apache.hadoop.hive.ql.plan.ExprNodeGenericFuncDesc) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException)

Aggregations

SemanticException (org.apache.hadoop.hive.ql.parse.SemanticException)131 ArrayList (java.util.ArrayList)64 ExprNodeDesc (org.apache.hadoop.hive.ql.plan.ExprNodeDesc)36 HashMap (java.util.HashMap)30 HiveException (org.apache.hadoop.hive.ql.metadata.HiveException)27 Path (org.apache.hadoop.fs.Path)22 IOException (java.io.IOException)20 LinkedHashMap (java.util.LinkedHashMap)19 List (java.util.List)18 TableScanOperator (org.apache.hadoop.hive.ql.exec.TableScanOperator)18 Node (org.apache.hadoop.hive.ql.lib.Node)17 ExprNodeGenericFuncDesc (org.apache.hadoop.hive.ql.plan.ExprNodeGenericFuncDesc)17 ReduceSinkOperator (org.apache.hadoop.hive.ql.exec.ReduceSinkOperator)16 DefaultRuleDispatcher (org.apache.hadoop.hive.ql.lib.DefaultRuleDispatcher)16 Dispatcher (org.apache.hadoop.hive.ql.lib.Dispatcher)16 GraphWalker (org.apache.hadoop.hive.ql.lib.GraphWalker)16 ExprNodeColumnDesc (org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc)16 Operator (org.apache.hadoop.hive.ql.exec.Operator)15 DefaultGraphWalker (org.apache.hadoop.hive.ql.lib.DefaultGraphWalker)15 Table (org.apache.hadoop.hive.ql.metadata.Table)14