Search in sources :

Example 1 with SparkHashTableSinkDesc

use of org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc in project hive by apache.

the class Vectorizer method validateSparkHashTableSinkOperator.

private boolean validateSparkHashTableSinkOperator(SparkHashTableSinkOperator op) {
    SparkHashTableSinkDesc desc = op.getConf();
    byte tag = desc.getTag();
    // it's essentially a MapJoinDesc
    List<ExprNodeDesc> filterExprs = desc.getFilters().get(tag);
    List<ExprNodeDesc> keyExprs = desc.getKeys().get(tag);
    List<ExprNodeDesc> valueExprs = desc.getExprs().get(tag);
    return validateExprNodeDesc(filterExprs, "Filter", VectorExpressionDescriptor.Mode.FILTER, /* allowComplex */
    true) && validateExprNodeDesc(keyExprs, "Key") && validateExprNodeDesc(valueExprs, "Value");
}
Also used : SparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc) VectorSparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkHashTableSinkDesc) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc)

Example 2 with SparkHashTableSinkDesc

use of org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc in project hive by apache.

the class SparkReduceSinkMapJoinProc method process.

/* (non-Javadoc)
   * This processor addresses the RS-MJ case that occurs in spark on the small/hash
   * table side of things. The work that RS will be a part of must be connected
   * to the MJ work via be a broadcast edge.
   * We should not walk down the tree when we encounter this pattern because:
   * the type of work (map work or reduce work) needs to be determined
   * on the basis of the big table side because it may be a mapwork (no need for shuffle)
   * or reduce work.
   */
@SuppressWarnings("unchecked")
@Override
public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procContext, Object... nodeOutputs) throws SemanticException {
    GenSparkProcContext context = (GenSparkProcContext) procContext;
    if (!nd.getClass().equals(MapJoinOperator.class)) {
        return null;
    }
    MapJoinOperator mapJoinOp = (MapJoinOperator) nd;
    if (stack.size() < 2 || !(stack.get(stack.size() - 2) instanceof ReduceSinkOperator)) {
        context.currentMapJoinOperators.add(mapJoinOp);
        return null;
    }
    context.preceedingWork = null;
    context.currentRootOperator = null;
    ReduceSinkOperator parentRS = (ReduceSinkOperator) stack.get(stack.size() - 2);
    // remove the tag for in-memory side of mapjoin
    parentRS.getConf().setSkipTag(true);
    parentRS.setSkipTag(true);
    // remember the original parent list before we start modifying it.
    if (!context.mapJoinParentMap.containsKey(mapJoinOp)) {
        List<Operator<?>> parents = new ArrayList<Operator<?>>(mapJoinOp.getParentOperators());
        context.mapJoinParentMap.put(mapJoinOp, parents);
    }
    List<BaseWork> mapJoinWork;
    /*
     *  If there was a pre-existing work generated for the big-table mapjoin side,
     *  we need to hook the work generated for the RS (associated with the RS-MJ pattern)
     *  with the pre-existing work.
     *
     *  Otherwise, we need to associate that the mapjoin op
     *  to be linked to the RS work (associated with the RS-MJ pattern).
     *
     */
    mapJoinWork = context.mapJoinWorkMap.get(mapJoinOp);
    int workMapSize = context.childToWorkMap.get(parentRS).size();
    Preconditions.checkArgument(workMapSize == 1, "AssertionError: expected context.childToWorkMap.get(parentRS).size() to be 1, but was " + workMapSize);
    BaseWork parentWork = context.childToWorkMap.get(parentRS).get(0);
    // set the link between mapjoin and parent vertex
    int pos = context.mapJoinParentMap.get(mapJoinOp).indexOf(parentRS);
    if (pos == -1) {
        throw new SemanticException("Cannot find position of parent in mapjoin");
    }
    LOG.debug("Mapjoin " + mapJoinOp + ", pos: " + pos + " --> " + parentWork.getName());
    mapJoinOp.getConf().getParentToInput().put(pos, parentWork.getName());
    SparkEdgeProperty edgeProp = new SparkEdgeProperty(SparkEdgeProperty.SHUFFLE_NONE);
    if (mapJoinWork != null) {
        for (BaseWork myWork : mapJoinWork) {
            // link the work with the work associated with the reduce sink that triggered this rule
            SparkWork sparkWork = context.currentTask.getWork();
            LOG.debug("connecting " + parentWork.getName() + " with " + myWork.getName());
            sparkWork.connect(parentWork, myWork, edgeProp);
        }
    }
    // remember in case we need to connect additional work later
    Map<BaseWork, SparkEdgeProperty> linkWorkMap = null;
    if (context.linkOpWithWorkMap.containsKey(mapJoinOp)) {
        linkWorkMap = context.linkOpWithWorkMap.get(mapJoinOp);
    } else {
        linkWorkMap = new HashMap<BaseWork, SparkEdgeProperty>();
    }
    linkWorkMap.put(parentWork, edgeProp);
    context.linkOpWithWorkMap.put(mapJoinOp, linkWorkMap);
    List<ReduceSinkOperator> reduceSinks = context.linkWorkWithReduceSinkMap.get(parentWork);
    if (reduceSinks == null) {
        reduceSinks = new ArrayList<ReduceSinkOperator>();
    }
    reduceSinks.add(parentRS);
    context.linkWorkWithReduceSinkMap.put(parentWork, reduceSinks);
    // create the dummy operators
    List<Operator<?>> dummyOperators = new ArrayList<Operator<?>>();
    // create an new operator: HashTableDummyOperator, which share the table desc
    HashTableDummyDesc desc = new HashTableDummyDesc();
    HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(mapJoinOp.getCompilationOpContext(), desc);
    TableDesc tbl;
    // need to create the correct table descriptor for key/value
    RowSchema rowSchema = parentRS.getParentOperators().get(0).getSchema();
    tbl = PlanUtils.getReduceValueTableDesc(PlanUtils.getFieldSchemasFromRowSchema(rowSchema, ""));
    dummyOp.getConf().setTbl(tbl);
    Map<Byte, List<ExprNodeDesc>> keyExprMap = mapJoinOp.getConf().getKeys();
    List<ExprNodeDesc> keyCols = keyExprMap.get(Byte.valueOf((byte) 0));
    StringBuilder keyOrder = new StringBuilder();
    StringBuilder keyNullOrder = new StringBuilder();
    for (int i = 0; i < keyCols.size(); i++) {
        keyOrder.append("+");
        keyNullOrder.append(NullOrdering.defaultNullOrder(context.conf));
    }
    TableDesc keyTableDesc = PlanUtils.getReduceKeyTableDesc(PlanUtils.getFieldSchemasFromColumnList(keyCols, "mapjoinkey"), keyOrder.toString(), keyNullOrder.toString());
    mapJoinOp.getConf().setKeyTableDesc(keyTableDesc);
    // let the dummy op be the parent of mapjoin op
    mapJoinOp.replaceParent(parentRS, dummyOp);
    List<Operator<? extends OperatorDesc>> dummyChildren = new ArrayList<Operator<? extends OperatorDesc>>();
    dummyChildren.add(mapJoinOp);
    dummyOp.setChildOperators(dummyChildren);
    dummyOperators.add(dummyOp);
    // cut the operator tree so as to not retain connections from the parent RS downstream
    List<Operator<? extends OperatorDesc>> childOperators = parentRS.getChildOperators();
    int childIndex = childOperators.indexOf(mapJoinOp);
    childOperators.remove(childIndex);
    // at task startup
    if (mapJoinWork != null) {
        for (BaseWork myWork : mapJoinWork) {
            myWork.addDummyOp(dummyOp);
        }
    }
    if (context.linkChildOpWithDummyOp.containsKey(mapJoinOp)) {
        for (Operator<?> op : context.linkChildOpWithDummyOp.get(mapJoinOp)) {
            dummyOperators.add(op);
        }
    }
    context.linkChildOpWithDummyOp.put(mapJoinOp, dummyOperators);
    // replace ReduceSinkOp with HashTableSinkOp for the RSops which are parents of MJop
    MapJoinDesc mjDesc = mapJoinOp.getConf();
    HiveConf conf = context.conf;
    // Unlike in MR, we may call this method multiple times, for each
    // small table HTS. But, since it's idempotent, it should be OK.
    mjDesc.resetOrder();
    float hashtableMemoryUsage;
    if (hasGroupBy(mapJoinOp, context)) {
        hashtableMemoryUsage = conf.getFloatVar(HiveConf.ConfVars.HIVEHASHTABLEFOLLOWBYGBYMAXMEMORYUSAGE);
    } else {
        hashtableMemoryUsage = conf.getFloatVar(HiveConf.ConfVars.HIVEHASHTABLEMAXMEMORYUSAGE);
    }
    mjDesc.setHashTableMemoryUsage(hashtableMemoryUsage);
    SparkHashTableSinkDesc hashTableSinkDesc = new SparkHashTableSinkDesc(mjDesc);
    SparkHashTableSinkOperator hashTableSinkOp = (SparkHashTableSinkOperator) OperatorFactory.get(mapJoinOp.getCompilationOpContext(), hashTableSinkDesc);
    byte tag = (byte) pos;
    int[] valueIndex = mjDesc.getValueIndex(tag);
    if (valueIndex != null) {
        List<ExprNodeDesc> newValues = new ArrayList<ExprNodeDesc>();
        List<ExprNodeDesc> values = hashTableSinkDesc.getExprs().get(tag);
        for (int index = 0; index < values.size(); index++) {
            if (valueIndex[index] < 0) {
                newValues.add(values.get(index));
            }
        }
        hashTableSinkDesc.getExprs().put(tag, newValues);
    }
    // get all parents of reduce sink
    List<Operator<? extends OperatorDesc>> rsParentOps = parentRS.getParentOperators();
    for (Operator<? extends OperatorDesc> parent : rsParentOps) {
        parent.replaceChild(parentRS, hashTableSinkOp);
    }
    hashTableSinkOp.setParentOperators(rsParentOps);
    hashTableSinkOp.getConf().setTag(tag);
    return true;
}
Also used : ReduceSinkOperator(org.apache.hadoop.hive.ql.exec.ReduceSinkOperator) MapJoinOperator(org.apache.hadoop.hive.ql.exec.MapJoinOperator) GroupByOperator(org.apache.hadoop.hive.ql.exec.GroupByOperator) Operator(org.apache.hadoop.hive.ql.exec.Operator) HashTableDummyOperator(org.apache.hadoop.hive.ql.exec.HashTableDummyOperator) SparkHashTableSinkOperator(org.apache.hadoop.hive.ql.exec.SparkHashTableSinkOperator) ArrayList(java.util.ArrayList) ArrayList(java.util.ArrayList) List(java.util.List) HiveConf(org.apache.hadoop.hive.conf.HiveConf) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) BaseWork(org.apache.hadoop.hive.ql.plan.BaseWork) SemanticException(org.apache.hadoop.hive.ql.parse.SemanticException) MapJoinOperator(org.apache.hadoop.hive.ql.exec.MapJoinOperator) SparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc) HashTableDummyDesc(org.apache.hadoop.hive.ql.plan.HashTableDummyDesc) RowSchema(org.apache.hadoop.hive.ql.exec.RowSchema) MapJoinDesc(org.apache.hadoop.hive.ql.plan.MapJoinDesc) SparkWork(org.apache.hadoop.hive.ql.plan.SparkWork) HashTableDummyOperator(org.apache.hadoop.hive.ql.exec.HashTableDummyOperator) SparkEdgeProperty(org.apache.hadoop.hive.ql.plan.SparkEdgeProperty) ReduceSinkOperator(org.apache.hadoop.hive.ql.exec.ReduceSinkOperator) GenSparkProcContext(org.apache.hadoop.hive.ql.parse.spark.GenSparkProcContext) TableDesc(org.apache.hadoop.hive.ql.plan.TableDesc) OperatorDesc(org.apache.hadoop.hive.ql.plan.OperatorDesc) SparkHashTableSinkOperator(org.apache.hadoop.hive.ql.exec.SparkHashTableSinkOperator)

Example 3 with SparkHashTableSinkDesc

use of org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc in project hive by apache.

the class Vectorizer method validateAndVectorizeOperator.

public Operator<? extends OperatorDesc> validateAndVectorizeOperator(Operator<? extends OperatorDesc> op, VectorizationContext vContext, boolean isReduce, boolean isTezOrSpark, VectorTaskColumnInfo vectorTaskColumnInfo) throws HiveException, VectorizerCannotVectorizeException {
    Operator<? extends OperatorDesc> vectorOp = null;
    // This "global" allows various validation methods to set the "not vectorized" reason.
    currentOperator = op;
    boolean isNative;
    try {
        switch(op.getType()) {
            case MAPJOIN:
                {
                    if (op instanceof MapJoinOperator) {
                        if (!validateMapJoinOperator((MapJoinOperator) op)) {
                            throw new VectorizerCannotVectorizeException();
                        }
                    } else if (op instanceof SMBMapJoinOperator) {
                        if (!validateSMBMapJoinOperator((SMBMapJoinOperator) op)) {
                            throw new VectorizerCannotVectorizeException();
                        }
                    } else {
                        setOperatorNotSupported(op);
                        throw new VectorizerCannotVectorizeException();
                    }
                    if (op instanceof MapJoinOperator) {
                        MapJoinDesc desc = (MapJoinDesc) op.getConf();
                        int joinType = desc.getConds()[0].getType();
                        VectorMapJoinDesc vectorMapJoinDesc = new VectorMapJoinDesc();
                        boolean specialize = canSpecializeMapJoin(op, desc, isTezOrSpark, vContext, vectorMapJoinDesc);
                        if (!specialize) {
                            Class<? extends Operator<?>> opClass = null;
                            // *NON-NATIVE* vector map differences for LEFT OUTER JOIN and Filtered...
                            List<ExprNodeDesc> bigTableFilters = desc.getFilters().get((byte) desc.getPosBigTable());
                            boolean isOuterAndFiltered = (!desc.isNoOuterJoin() && bigTableFilters.size() > 0);
                            if (!isOuterAndFiltered) {
                                opClass = VectorMapJoinOperator.class;
                            } else {
                                if (joinType == JoinDesc.FULL_OUTER_JOIN) {
                                    setOperatorIssue("Vectorized & filtered full-outer joins not supported");
                                    throw new VectorizerCannotVectorizeException();
                                }
                                opClass = VectorMapJoinOuterFilteredOperator.class;
                            }
                            vectorOp = OperatorFactory.getVectorOperator(opClass, op.getCompilationOpContext(), desc, vContext, vectorMapJoinDesc);
                            isNative = false;
                        } else {
                            // TEMPORARY Until Native Vector Map Join with Hybrid passes tests...
                            // HiveConf.setBoolVar(physicalContext.getConf(),
                            // HiveConf.ConfVars.HIVEUSEHYBRIDGRACEHASHJOIN, false);
                            vectorOp = specializeMapJoinOperator(op, vContext, desc, vectorMapJoinDesc);
                            isNative = true;
                            if (vectorTaskColumnInfo != null) {
                                if (usesVectorUDFAdaptor(vectorMapJoinDesc.getAllBigTableKeyExpressions())) {
                                    vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                                }
                                if (usesVectorUDFAdaptor(vectorMapJoinDesc.getAllBigTableValueExpressions())) {
                                    vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                                }
                            }
                        }
                    } else {
                        Preconditions.checkState(op instanceof SMBMapJoinOperator);
                        SMBJoinDesc smbJoinSinkDesc = (SMBJoinDesc) op.getConf();
                        // Check additional constraint.
                        if (smbJoinSinkDesc.getFilterMap() != null) {
                            setOperatorIssue("FilterMaps not supported for Vector Pass-Thru SMB MapJoin");
                            throw new VectorizerCannotVectorizeException();
                        }
                        VectorSMBJoinDesc vectorSMBJoinDesc = new VectorSMBJoinDesc();
                        vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), smbJoinSinkDesc, vContext, vectorSMBJoinDesc);
                        isNative = false;
                    }
                }
                break;
            case REDUCESINK:
                {
                    if (!validateReduceSinkOperator((ReduceSinkOperator) op)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    ReduceSinkDesc reduceDesc = (ReduceSinkDesc) op.getConf();
                    VectorReduceSinkDesc vectorReduceSinkDesc = new VectorReduceSinkDesc();
                    boolean specialize = canSpecializeReduceSink(reduceDesc, isTezOrSpark, vContext, vectorReduceSinkDesc);
                    if (!specialize) {
                        vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), reduceDesc, vContext, vectorReduceSinkDesc);
                        isNative = false;
                    } else {
                        vectorOp = specializeReduceSinkOperator(op, vContext, reduceDesc, vectorReduceSinkDesc);
                        isNative = true;
                        if (vectorTaskColumnInfo != null) {
                            VectorReduceSinkInfo vectorReduceSinkInfo = vectorReduceSinkDesc.getVectorReduceSinkInfo();
                            if (usesVectorUDFAdaptor(vectorReduceSinkInfo.getReduceSinkKeyExpressions())) {
                                vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                            }
                            if (usesVectorUDFAdaptor(vectorReduceSinkInfo.getReduceSinkValueExpressions())) {
                                vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                            }
                        }
                    }
                }
                break;
            case FILTER:
                {
                    if (!validateFilterOperator((FilterOperator) op)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    VectorFilterDesc vectorFilterDesc = new VectorFilterDesc();
                    vectorOp = vectorizeFilterOperator(op, vContext, vectorFilterDesc);
                    isNative = true;
                    if (vectorTaskColumnInfo != null) {
                        VectorExpression vectorPredicateExpr = vectorFilterDesc.getPredicateExpression();
                        if (usesVectorUDFAdaptor(vectorPredicateExpr)) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                    }
                }
                break;
            case TOPNKEY:
                {
                    if (!validateTopNKeyOperator((TopNKeyOperator) op)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    VectorTopNKeyDesc vectorTopNKeyDesc = new VectorTopNKeyDesc();
                    vectorOp = vectorizeTopNKeyOperator(op, vContext, vectorTopNKeyDesc);
                    isNative = true;
                    if (vectorTaskColumnInfo != null) {
                        VectorExpression[] keyExpressions = vectorTopNKeyDesc.getKeyExpressions();
                        if (usesVectorUDFAdaptor(keyExpressions)) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                    }
                }
                break;
            case SELECT:
                {
                    if (!validateSelectOperator((SelectOperator) op)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    VectorSelectDesc vectorSelectDesc = new VectorSelectDesc();
                    vectorOp = vectorizeSelectOperator(op, vContext, vectorSelectDesc);
                    isNative = true;
                    if (vectorTaskColumnInfo != null) {
                        VectorExpression[] vectorSelectExprs = vectorSelectDesc.getSelectExpressions();
                        if (usesVectorUDFAdaptor(vectorSelectExprs)) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                    }
                }
                break;
            case GROUPBY:
                {
                    // The validateGroupByOperator method will update vectorGroupByDesc.
                    VectorGroupByDesc vectorGroupByDesc = new VectorGroupByDesc();
                    if (!validateGroupByOperator((GroupByOperator) op, isReduce, isTezOrSpark, vectorGroupByDesc)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    ImmutablePair<Operator<? extends OperatorDesc>, String> pair = doVectorizeGroupByOperator(op, vContext, vectorGroupByDesc);
                    if (pair.left == null) {
                        setOperatorIssue(pair.right);
                        throw new VectorizerCannotVectorizeException();
                    }
                    vectorOp = pair.left;
                    isNative = false;
                    if (vectorTaskColumnInfo != null) {
                        VectorExpression[] vecKeyExpressions = vectorGroupByDesc.getKeyExpressions();
                        if (usesVectorUDFAdaptor(vecKeyExpressions)) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                        VectorAggregationDesc[] vecAggrDescs = vectorGroupByDesc.getVecAggrDescs();
                        for (VectorAggregationDesc vecAggrDesc : vecAggrDescs) {
                            if (usesVectorUDFAdaptor(vecAggrDesc.getInputExpression())) {
                                vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                            }
                        }
                    }
                }
                break;
            case FILESINK:
                {
                    if (!validateFileSinkOperator((FileSinkOperator) op)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    FileSinkDesc fileSinkDesc = (FileSinkDesc) op.getConf();
                    VectorFileSinkDesc vectorFileSinkDesc = new VectorFileSinkDesc();
                    boolean isArrowSpecialization = checkForArrowFileSink(fileSinkDesc, isTezOrSpark, vContext, vectorFileSinkDesc);
                    if (isArrowSpecialization) {
                        vectorOp = specializeArrowFileSinkOperator(op, vContext, fileSinkDesc, vectorFileSinkDesc);
                        isNative = true;
                    } else {
                        vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), fileSinkDesc, vContext, vectorFileSinkDesc);
                        isNative = false;
                    }
                }
                break;
            case LIMIT:
                {
                    // No validation.
                    LimitDesc limitDesc = (LimitDesc) op.getConf();
                    VectorLimitDesc vectorLimitDesc = new VectorLimitDesc();
                    vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), limitDesc, vContext, vectorLimitDesc);
                    isNative = true;
                }
                break;
            case EVENT:
                {
                    // No validation.
                    AppMasterEventDesc eventDesc = (AppMasterEventDesc) op.getConf();
                    VectorAppMasterEventDesc vectorEventDesc = new VectorAppMasterEventDesc();
                    vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), eventDesc, vContext, vectorEventDesc);
                    isNative = true;
                }
                break;
            case PTF:
                {
                    // The validatePTFOperator method will update vectorPTFDesc.
                    VectorPTFDesc vectorPTFDesc = new VectorPTFDesc();
                    if (!validatePTFOperator((PTFOperator) op, vContext, vectorPTFDesc)) {
                        throw new VectorizerCannotVectorizeException();
                    }
                    vectorOp = vectorizePTFOperator(op, vContext, vectorPTFDesc);
                    isNative = true;
                }
                break;
            case HASHTABLESINK:
                {
                    // No validation.
                    SparkHashTableSinkDesc sparkHashTableSinkDesc = (SparkHashTableSinkDesc) op.getConf();
                    VectorSparkHashTableSinkDesc vectorSparkHashTableSinkDesc = new VectorSparkHashTableSinkDesc();
                    vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), sparkHashTableSinkDesc, vContext, vectorSparkHashTableSinkDesc);
                    isNative = true;
                }
                break;
            case SPARKPRUNINGSINK:
                {
                    // No validation.
                    SparkPartitionPruningSinkDesc sparkPartitionPruningSinkDesc = (SparkPartitionPruningSinkDesc) op.getConf();
                    VectorSparkPartitionPruningSinkDesc vectorSparkPartitionPruningSinkDesc = new VectorSparkPartitionPruningSinkDesc();
                    vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), sparkPartitionPruningSinkDesc, vContext, vectorSparkPartitionPruningSinkDesc);
                    // need to maintain the unique ID so that target map works can
                    // read the output
                    ((SparkPartitionPruningSinkOperator) vectorOp).setUniqueId(((SparkPartitionPruningSinkOperator) op).getUniqueId());
                    isNative = true;
                }
                break;
            default:
                setOperatorNotSupported(op);
                throw new VectorizerCannotVectorizeException();
        }
    } catch (HiveException e) {
        setOperatorIssue(e.getMessage());
        throw new VectorizerCannotVectorizeException();
    }
    Preconditions.checkState(vectorOp != null);
    if (vectorTaskColumnInfo != null && !isNative) {
        vectorTaskColumnInfo.setAllNative(false);
    }
    LOG.debug("vectorizeOperator " + vectorOp.getClass().getName());
    LOG.debug("vectorizeOperator " + vectorOp.getConf().getClass().getName());
    // These operators need to be linked to enable runtime statistics to be gathered/used correctly
    planMapper.link(op, vectorOp);
    return vectorOp;
}
Also used : VectorFileSinkArrowOperator(org.apache.hadoop.hive.ql.exec.vector.filesink.VectorFileSinkArrowOperator) VectorMapJoinAntiJoinMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinAntiJoinMultiKeyOperator) VectorMapJoinInnerBigOnlyMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyMultiKeyOperator) VectorMapJoinLeftSemiMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiMultiKeyOperator) VectorMapJoinOuterFilteredOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOuterFilteredOperator) SparkPartitionPruningSinkOperator(org.apache.hadoop.hive.ql.parse.spark.SparkPartitionPruningSinkOperator) VectorizationOperator(org.apache.hadoop.hive.ql.exec.vector.VectorizationOperator) VectorMapJoinInnerMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerMultiKeyOperator) VectorMapJoinFullOuterStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinFullOuterStringOperator) VectorMapJoinOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator) VectorPTFOperator(org.apache.hadoop.hive.ql.exec.vector.ptf.VectorPTFOperator) VectorMapJoinInnerStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerStringOperator) VectorMapJoinOuterLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinOuterLongOperator) VectorMapJoinLeftSemiStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiStringOperator) VectorMapJoinLeftSemiLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiLongOperator) VectorMapJoinFullOuterLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinFullOuterLongOperator) VectorMapJoinFullOuterMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinFullOuterMultiKeyOperator) VectorMapJoinInnerBigOnlyLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyLongOperator) VectorMapJoinAntiJoinStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinAntiJoinStringOperator) VectorMapJoinInnerBigOnlyStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyStringOperator) VectorMapJoinOuterStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinOuterStringOperator) VectorMapJoinInnerLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerLongOperator) VectorMapJoinOuterMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinOuterMultiKeyOperator) VectorMapJoinAntiJoinLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinAntiJoinLongOperator) VectorAppMasterEventDesc(org.apache.hadoop.hive.ql.plan.VectorAppMasterEventDesc) AppMasterEventDesc(org.apache.hadoop.hive.ql.plan.AppMasterEventDesc) VectorMapJoinDesc(org.apache.hadoop.hive.ql.plan.VectorMapJoinDesc) HiveException(org.apache.hadoop.hive.ql.metadata.HiveException) SMBJoinDesc(org.apache.hadoop.hive.ql.plan.SMBJoinDesc) VectorSMBJoinDesc(org.apache.hadoop.hive.ql.plan.VectorSMBJoinDesc) VectorFileSinkDesc(org.apache.hadoop.hive.ql.plan.VectorFileSinkDesc) FileSinkDesc(org.apache.hadoop.hive.ql.plan.FileSinkDesc) VectorMapJoinOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator) VectorFileSinkDesc(org.apache.hadoop.hive.ql.plan.VectorFileSinkDesc) VectorReduceSinkInfo(org.apache.hadoop.hive.ql.plan.VectorReduceSinkInfo) VectorSparkPartitionPruningSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkPartitionPruningSinkDesc) SparkPartitionPruningSinkDesc(org.apache.hadoop.hive.ql.optimizer.spark.SparkPartitionPruningSinkDesc) VectorAppMasterEventDesc(org.apache.hadoop.hive.ql.plan.VectorAppMasterEventDesc) ArrayList(java.util.ArrayList) List(java.util.List) VectorSelectDesc(org.apache.hadoop.hive.ql.plan.VectorSelectDesc) VectorReduceSinkDesc(org.apache.hadoop.hive.ql.plan.VectorReduceSinkDesc) ReduceSinkDesc(org.apache.hadoop.hive.ql.plan.ReduceSinkDesc) SparkPartitionPruningSinkOperator(org.apache.hadoop.hive.ql.parse.spark.SparkPartitionPruningSinkOperator) VectorMapJoinOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator) VectorFilterDesc(org.apache.hadoop.hive.ql.plan.VectorFilterDesc) SparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc) VectorSparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkHashTableSinkDesc) VectorMapJoinDesc(org.apache.hadoop.hive.ql.plan.VectorMapJoinDesc) MapJoinDesc(org.apache.hadoop.hive.ql.plan.MapJoinDesc) VectorSparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkHashTableSinkDesc) VectorSparkPartitionPruningSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkPartitionPruningSinkDesc) VectorSMBJoinDesc(org.apache.hadoop.hive.ql.plan.VectorSMBJoinDesc) VectorReduceSinkDesc(org.apache.hadoop.hive.ql.plan.VectorReduceSinkDesc) VectorLimitDesc(org.apache.hadoop.hive.ql.plan.VectorLimitDesc) VectorAggregationDesc(org.apache.hadoop.hive.ql.exec.vector.VectorAggregationDesc) VectorLimitDesc(org.apache.hadoop.hive.ql.plan.VectorLimitDesc) LimitDesc(org.apache.hadoop.hive.ql.plan.LimitDesc) VectorTopNKeyDesc(org.apache.hadoop.hive.ql.plan.VectorTopNKeyDesc) ImmutablePair(org.apache.commons.lang3.tuple.ImmutablePair) VectorMapJoinOuterFilteredOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOuterFilteredOperator) VectorGroupByDesc(org.apache.hadoop.hive.ql.plan.VectorGroupByDesc) VectorPTFDesc(org.apache.hadoop.hive.ql.plan.VectorPTFDesc) VectorExpression(org.apache.hadoop.hive.ql.exec.vector.expressions.VectorExpression)

Example 4 with SparkHashTableSinkDesc

use of org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc in project hive by apache.

the class Vectorizer method vectorizeOperator.

public Operator<? extends OperatorDesc> vectorizeOperator(Operator<? extends OperatorDesc> op, VectorizationContext vContext, boolean isTezOrSpark, VectorTaskColumnInfo vectorTaskColumnInfo) throws HiveException {
    Operator<? extends OperatorDesc> vectorOp = null;
    boolean isNative;
    switch(op.getType()) {
        case TABLESCAN:
            vectorOp = vectorizeTableScanOperator(op, vContext);
            isNative = true;
            break;
        case MAPJOIN:
            {
                if (op instanceof MapJoinOperator) {
                    VectorMapJoinInfo vectorMapJoinInfo = new VectorMapJoinInfo();
                    MapJoinDesc desc = (MapJoinDesc) op.getConf();
                    boolean specialize = canSpecializeMapJoin(op, desc, isTezOrSpark, vContext, vectorMapJoinInfo);
                    if (!specialize) {
                        Class<? extends Operator<?>> opClass = null;
                        // *NON-NATIVE* vector map differences for LEFT OUTER JOIN and Filtered...
                        List<ExprNodeDesc> bigTableFilters = desc.getFilters().get((byte) desc.getPosBigTable());
                        boolean isOuterAndFiltered = (!desc.isNoOuterJoin() && bigTableFilters.size() > 0);
                        if (!isOuterAndFiltered) {
                            opClass = VectorMapJoinOperator.class;
                        } else {
                            opClass = VectorMapJoinOuterFilteredOperator.class;
                        }
                        vectorOp = OperatorFactory.getVectorOperator(opClass, op.getCompilationOpContext(), op.getConf(), vContext);
                        isNative = false;
                    } else {
                        // TEMPORARY Until Native Vector Map Join with Hybrid passes tests...
                        // HiveConf.setBoolVar(physicalContext.getConf(),
                        //    HiveConf.ConfVars.HIVEUSEHYBRIDGRACEHASHJOIN, false);
                        vectorOp = specializeMapJoinOperator(op, vContext, desc, vectorMapJoinInfo);
                        isNative = true;
                        if (vectorTaskColumnInfo != null) {
                            if (usesVectorUDFAdaptor(vectorMapJoinInfo.getBigTableKeyExpressions())) {
                                vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                            }
                            if (usesVectorUDFAdaptor(vectorMapJoinInfo.getBigTableValueExpressions())) {
                                vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                            }
                        }
                    }
                } else {
                    Preconditions.checkState(op instanceof SMBMapJoinOperator);
                    SMBJoinDesc smbJoinSinkDesc = (SMBJoinDesc) op.getConf();
                    VectorSMBJoinDesc vectorSMBJoinDesc = new VectorSMBJoinDesc();
                    smbJoinSinkDesc.setVectorDesc(vectorSMBJoinDesc);
                    vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), smbJoinSinkDesc, vContext);
                    isNative = false;
                }
            }
            break;
        case REDUCESINK:
            {
                VectorReduceSinkInfo vectorReduceSinkInfo = new VectorReduceSinkInfo();
                ReduceSinkDesc desc = (ReduceSinkDesc) op.getConf();
                boolean specialize = canSpecializeReduceSink(desc, isTezOrSpark, vContext, vectorReduceSinkInfo);
                if (!specialize) {
                    vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), op.getConf(), vContext);
                    isNative = false;
                } else {
                    vectorOp = specializeReduceSinkOperator(op, vContext, desc, vectorReduceSinkInfo);
                    isNative = true;
                    if (vectorTaskColumnInfo != null) {
                        if (usesVectorUDFAdaptor(vectorReduceSinkInfo.getReduceSinkKeyExpressions())) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                        if (usesVectorUDFAdaptor(vectorReduceSinkInfo.getReduceSinkValueExpressions())) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                    }
                }
            }
            break;
        case FILTER:
            {
                vectorOp = vectorizeFilterOperator(op, vContext);
                isNative = true;
                if (vectorTaskColumnInfo != null) {
                    VectorFilterDesc vectorFilterDesc = (VectorFilterDesc) ((AbstractOperatorDesc) vectorOp.getConf()).getVectorDesc();
                    VectorExpression vectorPredicateExpr = vectorFilterDesc.getPredicateExpression();
                    if (usesVectorUDFAdaptor(vectorPredicateExpr)) {
                        vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                    }
                }
            }
            break;
        case SELECT:
            {
                vectorOp = vectorizeSelectOperator(op, vContext);
                isNative = true;
                if (vectorTaskColumnInfo != null) {
                    VectorSelectDesc vectorSelectDesc = (VectorSelectDesc) ((AbstractOperatorDesc) vectorOp.getConf()).getVectorDesc();
                    VectorExpression[] vectorSelectExprs = vectorSelectDesc.getSelectExpressions();
                    if (usesVectorUDFAdaptor(vectorSelectExprs)) {
                        vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                    }
                }
            }
            break;
        case GROUPBY:
            {
                vectorOp = vectorizeGroupByOperator(op, vContext);
                isNative = false;
                if (vectorTaskColumnInfo != null) {
                    VectorGroupByDesc vectorGroupByDesc = (VectorGroupByDesc) ((AbstractOperatorDesc) vectorOp.getConf()).getVectorDesc();
                    if (!vectorGroupByDesc.isVectorOutput()) {
                        vectorTaskColumnInfo.setGroupByVectorOutput(false);
                    }
                    VectorExpression[] vecKeyExpressions = vectorGroupByDesc.getKeyExpressions();
                    if (usesVectorUDFAdaptor(vecKeyExpressions)) {
                        vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                    }
                    VectorAggregateExpression[] vecAggregators = vectorGroupByDesc.getAggregators();
                    for (VectorAggregateExpression vecAggr : vecAggregators) {
                        if (usesVectorUDFAdaptor(vecAggr.inputExpression())) {
                            vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
                        }
                    }
                }
            }
            break;
        case FILESINK:
            {
                FileSinkDesc fileSinkDesc = (FileSinkDesc) op.getConf();
                VectorFileSinkDesc vectorFileSinkDesc = new VectorFileSinkDesc();
                fileSinkDesc.setVectorDesc(vectorFileSinkDesc);
                vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), fileSinkDesc, vContext);
                isNative = false;
            }
            break;
        case LIMIT:
            {
                LimitDesc limitDesc = (LimitDesc) op.getConf();
                VectorLimitDesc vectorLimitDesc = new VectorLimitDesc();
                limitDesc.setVectorDesc(vectorLimitDesc);
                vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), limitDesc, vContext);
                isNative = true;
            }
            break;
        case EVENT:
            {
                AppMasterEventDesc eventDesc = (AppMasterEventDesc) op.getConf();
                VectorAppMasterEventDesc vectorEventDesc = new VectorAppMasterEventDesc();
                eventDesc.setVectorDesc(vectorEventDesc);
                vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), eventDesc, vContext);
                isNative = true;
            }
            break;
        case HASHTABLESINK:
            {
                SparkHashTableSinkDesc sparkHashTableSinkDesc = (SparkHashTableSinkDesc) op.getConf();
                VectorSparkHashTableSinkDesc vectorSparkHashTableSinkDesc = new VectorSparkHashTableSinkDesc();
                sparkHashTableSinkDesc.setVectorDesc(vectorSparkHashTableSinkDesc);
                vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), sparkHashTableSinkDesc, vContext);
                isNative = true;
            }
            break;
        case SPARKPRUNINGSINK:
            {
                SparkPartitionPruningSinkDesc sparkPartitionPruningSinkDesc = (SparkPartitionPruningSinkDesc) op.getConf();
                VectorSparkPartitionPruningSinkDesc vectorSparkPartitionPruningSinkDesc = new VectorSparkPartitionPruningSinkDesc();
                sparkPartitionPruningSinkDesc.setVectorDesc(vectorSparkPartitionPruningSinkDesc);
                vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), sparkPartitionPruningSinkDesc, vContext);
                isNative = true;
            }
            break;
        default:
            // These are children of GROUP BY operators with non-vector outputs.
            isNative = false;
            vectorOp = op;
            break;
    }
    Preconditions.checkState(vectorOp != null);
    if (vectorTaskColumnInfo != null && !isNative) {
        vectorTaskColumnInfo.setAllNative(false);
    }
    LOG.debug("vectorizeOperator " + vectorOp.getClass().getName());
    LOG.debug("vectorizeOperator " + vectorOp.getConf().getClass().getName());
    if (vectorOp != op) {
        fixupParentChildOperators(op, vectorOp);
        ((AbstractOperatorDesc) vectorOp.getConf()).setVectorMode(true);
    }
    return vectorOp;
}
Also used : VectorMapJoinInnerStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerStringOperator) VectorReduceSinkLongOperator(org.apache.hadoop.hive.ql.exec.vector.reducesink.VectorReduceSinkLongOperator) VectorMapJoinOuterLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinOuterLongOperator) VectorReduceSinkStringOperator(org.apache.hadoop.hive.ql.exec.vector.reducesink.VectorReduceSinkStringOperator) VectorMapJoinInnerBigOnlyMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyMultiKeyOperator) VectorMapJoinLeftSemiMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiMultiKeyOperator) VectorMapJoinLeftSemiStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiStringOperator) VectorMapJoinLeftSemiLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiLongOperator) VectorReduceSinkMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.reducesink.VectorReduceSinkMultiKeyOperator) VectorMapJoinOuterFilteredOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOuterFilteredOperator) VectorMapJoinInnerBigOnlyLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyLongOperator) VectorMapJoinInnerBigOnlyStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyStringOperator) VectorMapJoinInnerMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerMultiKeyOperator) VectorMapJoinOuterStringOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinOuterStringOperator) VectorMapJoinOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator) VectorMapJoinInnerLongOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerLongOperator) VectorMapJoinOuterMultiKeyOperator(org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinOuterMultiKeyOperator) AppMasterEventDesc(org.apache.hadoop.hive.ql.plan.AppMasterEventDesc) VectorAppMasterEventDesc(org.apache.hadoop.hive.ql.plan.VectorAppMasterEventDesc) SMBJoinDesc(org.apache.hadoop.hive.ql.plan.SMBJoinDesc) VectorSMBJoinDesc(org.apache.hadoop.hive.ql.plan.VectorSMBJoinDesc) VectorFileSinkDesc(org.apache.hadoop.hive.ql.plan.VectorFileSinkDesc) FileSinkDesc(org.apache.hadoop.hive.ql.plan.FileSinkDesc) VectorMapJoinOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator) VectorFileSinkDesc(org.apache.hadoop.hive.ql.plan.VectorFileSinkDesc) VectorReduceSinkInfo(org.apache.hadoop.hive.ql.plan.VectorReduceSinkInfo) VectorSparkPartitionPruningSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkPartitionPruningSinkDesc) SparkPartitionPruningSinkDesc(org.apache.hadoop.hive.ql.optimizer.spark.SparkPartitionPruningSinkDesc) VectorAppMasterEventDesc(org.apache.hadoop.hive.ql.plan.VectorAppMasterEventDesc) ArrayList(java.util.ArrayList) List(java.util.List) VectorSelectDesc(org.apache.hadoop.hive.ql.plan.VectorSelectDesc) VectorReduceSinkDesc(org.apache.hadoop.hive.ql.plan.VectorReduceSinkDesc) ReduceSinkDesc(org.apache.hadoop.hive.ql.plan.ReduceSinkDesc) VectorMapJoinOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator) VectorFilterDesc(org.apache.hadoop.hive.ql.plan.VectorFilterDesc) SparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc) VectorSparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkHashTableSinkDesc) AbstractOperatorDesc(org.apache.hadoop.hive.ql.plan.AbstractOperatorDesc) MapJoinDesc(org.apache.hadoop.hive.ql.plan.MapJoinDesc) VectorMapJoinDesc(org.apache.hadoop.hive.ql.plan.VectorMapJoinDesc) VectorSparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkHashTableSinkDesc) VectorSparkPartitionPruningSinkDesc(org.apache.hadoop.hive.ql.plan.VectorSparkPartitionPruningSinkDesc) VectorMapJoinInfo(org.apache.hadoop.hive.ql.plan.VectorMapJoinInfo) VectorSMBJoinDesc(org.apache.hadoop.hive.ql.plan.VectorSMBJoinDesc) VectorAggregateExpression(org.apache.hadoop.hive.ql.exec.vector.expressions.aggregates.VectorAggregateExpression) VectorLimitDesc(org.apache.hadoop.hive.ql.plan.VectorLimitDesc) LimitDesc(org.apache.hadoop.hive.ql.plan.LimitDesc) VectorLimitDesc(org.apache.hadoop.hive.ql.plan.VectorLimitDesc) VectorMapJoinOuterFilteredOperator(org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOuterFilteredOperator) VectorGroupByDesc(org.apache.hadoop.hive.ql.plan.VectorGroupByDesc) VectorExpression(org.apache.hadoop.hive.ql.exec.vector.expressions.VectorExpression)

Example 5 with SparkHashTableSinkDesc

use of org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc in project hive by apache.

the class GenSparkSkewJoinProcessor method insertSHTS.

/**
 * Insert SparkHashTableSink and HashTableDummy between small dir TS and MJ.
 */
@SuppressWarnings("unchecked")
private static void insertSHTS(byte tag, TableScanOperator tableScan, MapWork bigMapWork) {
    Preconditions.checkArgument(tableScan.getChildOperators().size() == 1 && tableScan.getChildOperators().get(0) instanceof MapJoinOperator);
    HashTableDummyDesc desc = new HashTableDummyDesc();
    HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(tableScan.getCompilationOpContext(), desc);
    dummyOp.getConf().setTbl(tableScan.getTableDescSkewJoin());
    MapJoinOperator mapJoinOp = (MapJoinOperator) tableScan.getChildOperators().get(0);
    mapJoinOp.replaceParent(tableScan, dummyOp);
    List<Operator<? extends OperatorDesc>> mapJoinChildren = new ArrayList<Operator<? extends OperatorDesc>>();
    mapJoinChildren.add(mapJoinOp);
    dummyOp.setChildOperators(mapJoinChildren);
    bigMapWork.addDummyOp(dummyOp);
    MapJoinDesc mjDesc = mapJoinOp.getConf();
    // mapjoin should not be affected by join reordering
    mjDesc.resetOrder();
    SparkHashTableSinkDesc hashTableSinkDesc = new SparkHashTableSinkDesc(mjDesc);
    SparkHashTableSinkOperator hashTableSinkOp = (SparkHashTableSinkOperator) OperatorFactory.get(tableScan.getCompilationOpContext(), hashTableSinkDesc);
    int[] valueIndex = mjDesc.getValueIndex(tag);
    if (valueIndex != null) {
        List<ExprNodeDesc> newValues = new ArrayList<ExprNodeDesc>();
        List<ExprNodeDesc> values = hashTableSinkDesc.getExprs().get(tag);
        for (int index = 0; index < values.size(); index++) {
            if (valueIndex[index] < 0) {
                newValues.add(values.get(index));
            }
        }
        hashTableSinkDesc.getExprs().put(tag, newValues);
    }
    tableScan.replaceChild(mapJoinOp, hashTableSinkOp);
    List<Operator<? extends OperatorDesc>> tableScanParents = new ArrayList<Operator<? extends OperatorDesc>>();
    tableScanParents.add(tableScan);
    hashTableSinkOp.setParentOperators(tableScanParents);
    hashTableSinkOp.getConf().setTag(tag);
}
Also used : MapJoinOperator(org.apache.hadoop.hive.ql.exec.MapJoinOperator) MapJoinOperator(org.apache.hadoop.hive.ql.exec.MapJoinOperator) HashTableDummyOperator(org.apache.hadoop.hive.ql.exec.HashTableDummyOperator) SparkHashTableSinkOperator(org.apache.hadoop.hive.ql.exec.SparkHashTableSinkOperator) JoinOperator(org.apache.hadoop.hive.ql.exec.JoinOperator) TableScanOperator(org.apache.hadoop.hive.ql.exec.TableScanOperator) Operator(org.apache.hadoop.hive.ql.exec.Operator) SparkHashTableSinkDesc(org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc) HashTableDummyDesc(org.apache.hadoop.hive.ql.plan.HashTableDummyDesc) MapJoinDesc(org.apache.hadoop.hive.ql.plan.MapJoinDesc) ArrayList(java.util.ArrayList) HashTableDummyOperator(org.apache.hadoop.hive.ql.exec.HashTableDummyOperator) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) OperatorDesc(org.apache.hadoop.hive.ql.plan.OperatorDesc) SparkHashTableSinkOperator(org.apache.hadoop.hive.ql.exec.SparkHashTableSinkOperator)

Aggregations

SparkHashTableSinkDesc (org.apache.hadoop.hive.ql.plan.SparkHashTableSinkDesc)5 ArrayList (java.util.ArrayList)4 MapJoinDesc (org.apache.hadoop.hive.ql.plan.MapJoinDesc)4 List (java.util.List)3 ExprNodeDesc (org.apache.hadoop.hive.ql.plan.ExprNodeDesc)3 HashTableDummyOperator (org.apache.hadoop.hive.ql.exec.HashTableDummyOperator)2 MapJoinOperator (org.apache.hadoop.hive.ql.exec.MapJoinOperator)2 Operator (org.apache.hadoop.hive.ql.exec.Operator)2 SparkHashTableSinkOperator (org.apache.hadoop.hive.ql.exec.SparkHashTableSinkOperator)2 VectorMapJoinOperator (org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOperator)2 VectorMapJoinOuterFilteredOperator (org.apache.hadoop.hive.ql.exec.vector.VectorMapJoinOuterFilteredOperator)2 VectorExpression (org.apache.hadoop.hive.ql.exec.vector.expressions.VectorExpression)2 VectorMapJoinInnerBigOnlyLongOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyLongOperator)2 VectorMapJoinInnerBigOnlyMultiKeyOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyMultiKeyOperator)2 VectorMapJoinInnerBigOnlyStringOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerBigOnlyStringOperator)2 VectorMapJoinInnerLongOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerLongOperator)2 VectorMapJoinInnerMultiKeyOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerMultiKeyOperator)2 VectorMapJoinInnerStringOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinInnerStringOperator)2 VectorMapJoinLeftSemiLongOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiLongOperator)2 VectorMapJoinLeftSemiMultiKeyOperator (org.apache.hadoop.hive.ql.exec.vector.mapjoin.VectorMapJoinLeftSemiMultiKeyOperator)2