Examples with ReduceSinkDesc org.apache.hadoop.hive.ql.plan.ReduceSinkDesc used on opensource projects

Example 21 with ReduceSinkDesc

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

the class TestExecDriver method populateMapRedPlan5.

@SuppressWarnings("unchecked")
private void populateMapRedPlan5(Table src) throws SemanticException {
    // map-side work
    ArrayList<String> outputColumns = new ArrayList<String>();
    for (int i = 0; i < 2; i++) {
        outputColumns.add("_col" + i);
    }
    Operator<ReduceSinkDesc> op0 = OperatorFactory.get(ctx, PlanUtils.getReduceSinkDesc(Utilities.makeList(getStringColumn("0")), Utilities.makeList(getStringColumn("0"), getStringColumn("1")), outputColumns, false, -1, 1, -1, AcidUtils.Operation.NOT_ACID, NullOrdering.NULLS_LAST));
    Operator<SelectDesc> op4 = OperatorFactory.get(new SelectDesc(Utilities.makeList(getStringColumn("key"), getStringColumn("value")), outputColumns), op0);
    addMapWork(mr, src, "a", op4);
    ReduceWork rWork = new ReduceWork();
    mr.setReduceWork(rWork);
    rWork.setNumReduceTasks(Integer.valueOf(1));
    rWork.setKeyDesc(op0.getConf().getKeySerializeInfo());
    rWork.getTagToValueDesc().add(op0.getConf().getValueSerializeInfo());
    // reduce side work
    Operator<FileSinkDesc> op3 = OperatorFactory.get(ctx, new FileSinkDesc(new Path(TMPDIR + File.separator + "mapredplan5.out"), Utilities.defaultTd, false));
    List<ExprNodeDesc> cols = new ArrayList<ExprNodeDesc>();
    cols.add(getStringColumn(Utilities.ReduceField.KEY + ".reducesinkkey" + 0));
    cols.add(getStringColumn(Utilities.ReduceField.VALUE.toString() + "." + outputColumns.get(1)));
    Operator<SelectDesc> op2 = OperatorFactory.get(new SelectDesc(cols, outputColumns), op3);
    rWork.setReducer(op2);
}

Example 22 with ReduceSinkDesc

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

the class TestExecDriver method populateMapRedPlan2.

@SuppressWarnings("unchecked")
private void populateMapRedPlan2(Table src) throws Exception {
    ArrayList<String> outputColumns = new ArrayList<String>();
    for (int i = 0; i < 2; i++) {
        outputColumns.add("_col" + i);
    }
    // map-side work
    Operator<ReduceSinkDesc> op1 = OperatorFactory.get(ctx, PlanUtils.getReduceSinkDesc(Utilities.makeList(getStringColumn("key")), Utilities.makeList(getStringColumn("key"), getStringColumn("value")), outputColumns, false, -1, 1, -1, AcidUtils.Operation.NOT_ACID, NullOrdering.NULLS_LAST));
    addMapWork(mr, src, "a", op1);
    ReduceWork rWork = new ReduceWork();
    rWork.setNumReduceTasks(Integer.valueOf(1));
    rWork.setKeyDesc(op1.getConf().getKeySerializeInfo());
    rWork.getTagToValueDesc().add(op1.getConf().getValueSerializeInfo());
    mr.setReduceWork(rWork);
    // reduce side work
    Operator<FileSinkDesc> op4 = OperatorFactory.get(ctx, new FileSinkDesc(new Path(TMPDIR + File.separator + "mapredplan2.out"), Utilities.defaultTd, false));
    Operator<FilterDesc> op3 = OperatorFactory.get(getTestFilterDesc("0"), op4);
    List<ExprNodeDesc> cols = new ArrayList<ExprNodeDesc>();
    cols.add(getStringColumn(Utilities.ReduceField.KEY + ".reducesinkkey" + 0));
    cols.add(getStringColumn(Utilities.ReduceField.VALUE.toString() + "." + outputColumns.get(1)));
    Operator<SelectDesc> op2 = OperatorFactory.get(new SelectDesc(cols, outputColumns), op3);
    rWork.setReducer(op2);
}

Example 23 with ReduceSinkDesc

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

the class TestExecDriver method populateMapRedPlan1.

@SuppressWarnings("unchecked")
private void populateMapRedPlan1(Table src) throws SemanticException {
    ArrayList<String> outputColumns = new ArrayList<String>();
    for (int i = 0; i < 2; i++) {
        outputColumns.add("_col" + i);
    }
    // map-side work
    Operator<ReduceSinkDesc> op1 = OperatorFactory.get(ctx, PlanUtils.getReduceSinkDesc(Utilities.makeList(getStringColumn("key")), Utilities.makeList(getStringColumn("value")), outputColumns, true, -1, 1, -1, AcidUtils.Operation.NOT_ACID, NullOrdering.NULLS_LAST));
    addMapWork(mr, src, "a", op1);
    ReduceWork rWork = new ReduceWork();
    rWork.setNumReduceTasks(Integer.valueOf(1));
    rWork.setKeyDesc(op1.getConf().getKeySerializeInfo());
    rWork.getTagToValueDesc().add(op1.getConf().getValueSerializeInfo());
    mr.setReduceWork(rWork);
    // reduce side work
    Operator<FileSinkDesc> op3 = OperatorFactory.get(ctx, new FileSinkDesc(new Path(TMPDIR + File.separator + "mapredplan1.out"), Utilities.defaultTd, false));
    List<ExprNodeDesc> cols = new ArrayList<ExprNodeDesc>();
    cols.add(getStringColumn(Utilities.ReduceField.VALUE.toString() + "." + outputColumns.get(1)));
    List<String> colNames = new ArrayList<String>();
    colNames.add(HiveConf.getColumnInternalName(2));
    Operator<SelectDesc> op2 = OperatorFactory.get(new SelectDesc(cols, colNames), op3);
    rWork.setReducer(op2);
}

Example 24 with ReduceSinkDesc

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

the class TestExecDriver method populateMapRedPlan4.

@SuppressWarnings("unchecked")
private void populateMapRedPlan4(Table src) throws SemanticException {
    // map-side work
    ArrayList<String> outputColumns = new ArrayList<String>();
    for (int i = 0; i < 2; i++) {
        outputColumns.add("_col" + i);
    }
    Operator<ReduceSinkDesc> op1 = OperatorFactory.get(ctx, PlanUtils.getReduceSinkDesc(Utilities.makeList(getStringColumn("tkey")), Utilities.makeList(getStringColumn("tkey"), getStringColumn("tvalue")), outputColumns, false, -1, 1, -1, AcidUtils.Operation.NOT_ACID, NullOrdering.NULLS_LAST));
    Operator<ScriptDesc> op0 = OperatorFactory.get(new ScriptDesc("cat", PlanUtils.getDefaultTableDesc("" + Utilities.tabCode, "key,value"), TextRecordWriter.class, PlanUtils.getDefaultTableDesc("" + Utilities.tabCode, "tkey,tvalue"), TextRecordReader.class, TextRecordReader.class, PlanUtils.getDefaultTableDesc("" + Utilities.tabCode, "key")), op1);
    Operator<SelectDesc> op4 = OperatorFactory.get(new SelectDesc(Utilities.makeList(getStringColumn("key"), getStringColumn("value")), outputColumns), op0);
    addMapWork(mr, src, "a", op4);
    ReduceWork rWork = new ReduceWork();
    rWork.setKeyDesc(op1.getConf().getKeySerializeInfo());
    rWork.getTagToValueDesc().add(op1.getConf().getValueSerializeInfo());
    rWork.setNumReduceTasks(Integer.valueOf(1));
    mr.setReduceWork(rWork);
    // reduce side work
    Operator<FileSinkDesc> op3 = OperatorFactory.get(ctx, new FileSinkDesc(new Path(TMPDIR + File.separator + "mapredplan4.out"), Utilities.defaultTd, false));
    List<ExprNodeDesc> cols = new ArrayList<ExprNodeDesc>();
    cols.add(getStringColumn(Utilities.ReduceField.KEY + ".reducesinkkey" + 0));
    cols.add(getStringColumn(Utilities.ReduceField.VALUE.toString() + "." + outputColumns.get(1)));
    Operator<SelectDesc> op2 = OperatorFactory.get(new SelectDesc(cols, outputColumns), op3);
    rWork.setReducer(op2);
}

Example 25 with ReduceSinkDesc

use of org.apache.hadoop.hive.ql.plan.ReduceSinkDesc 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;
}