Examples with VectorizedRowBatchCtx org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx used on opensource projects

Example 31 with VectorizedRowBatchCtx

use of org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx in project hive by apache.

the class TestVectorStringConcat method doStringConcatTestsWithDiffColumnScalar.

private void doStringConcatTestsWithDiffColumnScalar(Random random, String stringTypeName1, String stringTypeName2, ColumnScalarMode columnScalarMode) throws Exception {
    TypeInfo stringTypeInfo1 = TypeInfoUtils.getTypeInfoFromTypeString(stringTypeName1);
    PrimitiveCategory stringPrimitiveCategory1 = ((PrimitiveTypeInfo) stringTypeInfo1).getPrimitiveCategory();
    TypeInfo stringTypeInfo2 = TypeInfoUtils.getTypeInfoFromTypeString(stringTypeName2);
    PrimitiveCategory stringPrimitiveCategory2 = ((PrimitiveTypeInfo) stringTypeInfo2).getPrimitiveCategory();
    String functionName = "concat";
    List<GenerationSpec> generationSpecList = new ArrayList<GenerationSpec>();
    List<DataTypePhysicalVariation> explicitDataTypePhysicalVariationList = new ArrayList<DataTypePhysicalVariation>();
    List<String> columns = new ArrayList<String>();
    int columnNum = 1;
    ExprNodeDesc col1Expr;
    if (columnScalarMode == ColumnScalarMode.COLUMN_COLUMN || columnScalarMode == ColumnScalarMode.COLUMN_SCALAR) {
        generationSpecList.add(GenerationSpec.createSameType(stringTypeInfo1));
        explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
        String columnName = "col" + (columnNum++);
        col1Expr = new ExprNodeColumnDesc(stringTypeInfo1, columnName, "table", false);
        columns.add(columnName);
    } else {
        Object scalar1Object = VectorRandomRowSource.randomPrimitiveObject(random, (PrimitiveTypeInfo) stringTypeInfo1);
        col1Expr = new ExprNodeConstantDesc(stringTypeInfo1, scalar1Object);
    }
    ExprNodeDesc col2Expr;
    if (columnScalarMode == ColumnScalarMode.COLUMN_COLUMN || columnScalarMode == ColumnScalarMode.SCALAR_COLUMN) {
        generationSpecList.add(GenerationSpec.createSameType(stringTypeInfo2));
        explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
        String columnName = "col" + (columnNum++);
        col2Expr = new ExprNodeColumnDesc(stringTypeInfo2, columnName, "table", false);
        columns.add(columnName);
    } else {
        Object scalar2Object = VectorRandomRowSource.randomPrimitiveObject(random, (PrimitiveTypeInfo) stringTypeInfo2);
        col2Expr = new ExprNodeConstantDesc(stringTypeInfo2, scalar2Object);
    }
    List<ExprNodeDesc> children = new ArrayList<ExprNodeDesc>();
    children.add(col1Expr);
    children.add(col2Expr);
    // ----------------------------------------------------------------------------------------------
    String[] columnNames = columns.toArray(new String[0]);
    VectorRandomRowSource rowSource = new VectorRandomRowSource();
    rowSource.initGenerationSpecSchema(random, generationSpecList, /* maxComplexDepth */
    0, /* allowNull */
    true, /* isUnicodeOk */
    true, explicitDataTypePhysicalVariationList);
    Object[][] randomRows = rowSource.randomRows(100000);
    VectorRandomBatchSource batchSource = VectorRandomBatchSource.createInterestingBatches(random, rowSource, randomRows, null);
    String[] outputScratchTypeNames = new String[] { "string" };
    VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx(columnNames, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), /* dataColumnNums */
    null, /* partitionColumnCount */
    0, /* virtualColumnCount */
    0, /* neededVirtualColumns */
    null, outputScratchTypeNames, null);
    GenericUDF genericUdf;
    FunctionInfo funcInfo = null;
    try {
        funcInfo = FunctionRegistry.getFunctionInfo(functionName);
    } catch (SemanticException e) {
        Assert.fail("Failed to load " + functionName + " " + e);
    }
    genericUdf = funcInfo.getGenericUDF();
    final int rowCount = randomRows.length;
    Object[][] resultObjectsArray = new Object[StringConcatTestMode.count][];
    for (int i = 0; i < StringConcatTestMode.count; i++) {
        Object[] resultObjects = new Object[rowCount];
        resultObjectsArray[i] = resultObjects;
        StringConcatTestMode stringConcatTestMode = StringConcatTestMode.values()[i];
        switch(stringConcatTestMode) {
            case ROW_MODE:
                doRowStringConcatTest(stringTypeInfo1, stringTypeInfo2, columns, children, randomRows, columnScalarMode, rowSource.rowStructObjectInspector(), genericUdf, resultObjects);
                break;
            case ADAPTOR:
            case VECTOR_EXPRESSION:
                doVectorStringConcatTest(stringTypeInfo1, stringTypeInfo2, columns, rowSource.typeInfos(), children, stringConcatTestMode, columnScalarMode, batchSource, batchContext, rowSource.rowStructObjectInspector(), genericUdf, resultObjects);
                break;
            default:
                throw new RuntimeException("Unexpected IF statement test mode " + stringConcatTestMode);
        }
    }
    for (int i = 0; i < rowCount; i++) {
        // Row-mode is the expected value.
        Object expectedResult = resultObjectsArray[0][i];
        for (int v = 1; v < StringConcatTestMode.count; v++) {
            Object vectorResult = resultObjectsArray[v][i];
            if (expectedResult == null || vectorResult == null) {
                if (expectedResult != null || vectorResult != null) {
                    Assert.fail("Row " + i + " " + StringConcatTestMode.values()[v] + " " + columnScalarMode + " result is NULL " + (vectorResult == null) + " does not match row-mode expected result is NULL " + (expectedResult == null) + " row values " + Arrays.toString(randomRows[i]));
                }
            } else {
                if (!expectedResult.equals(vectorResult)) {
                    Assert.fail("Row " + i + " " + StringConcatTestMode.values()[v] + " " + columnScalarMode + " result \"" + vectorResult.toString() + "\"" + " (" + vectorResult.getClass().getSimpleName() + ")" + " does not match row-mode expected result \"" + expectedResult.toString() + "\"" + " (" + expectedResult.getClass().getSimpleName() + ")" + " row values " + Arrays.toString(randomRows[i]));
                }
            }
        }
    }
}

Example 32 with VectorizedRowBatchCtx

use of org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx in project hive by apache.

the class TestVectorStringUnary method doTests.

private void doTests(Random random, String typeName, String functionName) throws Exception {
    TypeInfo typeInfo = TypeInfoUtils.getTypeInfoFromTypeString(typeName);
    TypeInfo targetTypeInfo;
    if (functionName.equals("char_length") || functionName.equals("length") || functionName.equals("octet_length")) {
        targetTypeInfo = TypeInfoFactory.intTypeInfo;
    } else {
        targetTypeInfo = typeInfo;
    }
    List<GenerationSpec> generationSpecList = new ArrayList<GenerationSpec>();
    List<DataTypePhysicalVariation> explicitDataTypePhysicalVariationList = new ArrayList<DataTypePhysicalVariation>();
    List<String> columns = new ArrayList<String>();
    int columnNum = 1;
    ExprNodeDesc col1Expr;
    StringGenerationOption stringGenerationOption = new StringGenerationOption(true, true);
    generationSpecList.add(GenerationSpec.createStringFamily(typeInfo, stringGenerationOption));
    explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
    String columnName = "col" + (columnNum++);
    col1Expr = new ExprNodeColumnDesc(typeInfo, columnName, "table", false);
    columns.add(columnName);
    VectorRandomRowSource rowSource = new VectorRandomRowSource();
    rowSource.initGenerationSpecSchema(random, generationSpecList, /* maxComplexDepth */
    0, /* allowNull */
    true, /* isUnicodeOk */
    true, explicitDataTypePhysicalVariationList);
    List<ExprNodeDesc> children = new ArrayList<ExprNodeDesc>();
    children.add(col1Expr);
    // ----------------------------------------------------------------------------------------------
    String[] columnNames = columns.toArray(new String[0]);
    String[] outputScratchTypeNames = new String[] { targetTypeInfo.getTypeName() };
    DataTypePhysicalVariation[] outputDataTypePhysicalVariations = new DataTypePhysicalVariation[] { DataTypePhysicalVariation.NONE };
    VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx(columnNames, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), /* dataColumnNums */
    null, /* partitionColumnCount */
    0, /* virtualColumnCount */
    0, /* neededVirtualColumns */
    null, outputScratchTypeNames, outputDataTypePhysicalVariations);
    Object[][] randomRows = rowSource.randomRows(100000);
    VectorRandomBatchSource batchSource = VectorRandomBatchSource.createInterestingBatches(random, rowSource, randomRows, null);
    GenericUDF genericUdf;
    FunctionInfo funcInfo = null;
    try {
        funcInfo = FunctionRegistry.getFunctionInfo(functionName);
    } catch (SemanticException e) {
        Assert.fail("Failed to load " + functionName + " " + e);
    }
    genericUdf = funcInfo.getGenericUDF();
    final int rowCount = randomRows.length;
    Object[][] resultObjectsArray = new Object[StringUnaryTestMode.count][];
    for (int i = 0; i < StringUnaryTestMode.count; i++) {
        Object[] resultObjects = new Object[rowCount];
        resultObjectsArray[i] = resultObjects;
        StringUnaryTestMode stringUnaryTestMode = StringUnaryTestMode.values()[i];
        switch(stringUnaryTestMode) {
            case ROW_MODE:
                doRowIfTest(typeInfo, targetTypeInfo, columns, children, randomRows, rowSource.rowStructObjectInspector(), genericUdf, resultObjects);
                break;
            case ADAPTOR:
            case VECTOR_EXPRESSION:
                doVectorIfTest(typeInfo, targetTypeInfo, columns, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), children, stringUnaryTestMode, batchSource, batchContext, genericUdf, resultObjects);
                break;
            default:
                throw new RuntimeException("Unexpected STRING Unary test mode " + stringUnaryTestMode);
        }
    }
    for (int i = 0; i < rowCount; i++) {
        // Row-mode is the expected value.
        Object expectedResult = resultObjectsArray[0][i];
        for (int v = 1; v < StringUnaryTestMode.count; v++) {
            Object vectorResult = resultObjectsArray[v][i];
            if (expectedResult == null || vectorResult == null) {
                if (expectedResult != null || vectorResult != null) {
                    Assert.fail("Row " + i + " " + StringUnaryTestMode.values()[v] + " typeName " + typeName + " result is NULL " + (vectorResult == null) + " does not match row-mode expected result is NULL " + (expectedResult == null) + " functionName " + functionName + " genericUdf " + genericUdf.getClass().getSimpleName());
                }
            } else {
                if (!expectedResult.equals(vectorResult)) {
                    Assert.fail("Row " + i + " " + StringUnaryTestMode.values()[v] + " typeName " + typeName + " result \"" + vectorResult.toString() + "\"" + " (" + vectorResult.getClass().getSimpleName() + ")" + " does not match row-mode expected result \"" + expectedResult.toString() + "\"" + " (" + expectedResult.getClass().getSimpleName() + ")" + " functionName " + functionName + " genericUdf " + genericUdf.getClass().getSimpleName());
                }
            }
        }
    }
}

Example 33 with VectorizedRowBatchCtx

use of org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx in project hive by apache.

the class TestVectorTimestampExtract method doVectorCastTest.

private boolean doVectorCastTest(TypeInfo dateTimeStringTypeInfo, List<String> columns, String[] columnNames, TypeInfo[] typeInfos, DataTypePhysicalVariation[] dataTypePhysicalVariations, List<ExprNodeDesc> children, ExprNodeGenericFuncDesc exprDesc, TimestampExtractTestMode timestampExtractTestMode, VectorRandomBatchSource batchSource, Object[] resultObjects) throws Exception {
    HiveConf hiveConf = new HiveConf();
    if (timestampExtractTestMode == TimestampExtractTestMode.ADAPTOR) {
        hiveConf.setBoolVar(HiveConf.ConfVars.HIVE_TEST_VECTOR_ADAPTOR_OVERRIDE, true);
    }
    VectorizationContext vectorizationContext = new VectorizationContext("name", columns, Arrays.asList(typeInfos), Arrays.asList(dataTypePhysicalVariations), hiveConf);
    VectorExpression vectorExpression = vectorizationContext.getVectorExpression(exprDesc);
    vectorExpression.transientInit(hiveConf);
    if (timestampExtractTestMode == TimestampExtractTestMode.VECTOR_EXPRESSION && vectorExpression instanceof VectorUDFAdaptor) {
        System.out.println("*NO NATIVE VECTOR EXPRESSION* dateTimeStringTypeInfo " + dateTimeStringTypeInfo.toString() + " timestampExtractTestMode " + timestampExtractTestMode + " vectorExpression " + vectorExpression.toString());
    }
    // System.out.println("*VECTOR EXPRESSION* " + vectorExpression.getClass().getSimpleName());
    /*
    System.out.println(
        "*DEBUG* dateTimeStringTypeInfo " + dateTimeStringTypeInfo.toString() +
        " timestampExtractTestMode " + timestampExtractTestMode +
        " vectorExpression " + vectorExpression.getClass().getSimpleName());
    */
    VectorRandomRowSource rowSource = batchSource.getRowSource();
    VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx(columnNames, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), /* dataColumnNums */
    null, /* partitionColumnCount */
    0, /* virtualColumnCount */
    0, /* neededVirtualColumns */
    null, vectorizationContext.getScratchColumnTypeNames(), vectorizationContext.getScratchDataTypePhysicalVariations());
    VectorizedRowBatch batch = batchContext.createVectorizedRowBatch();
    VectorExtractRow resultVectorExtractRow = new VectorExtractRow();
    resultVectorExtractRow.init(new TypeInfo[] { TypeInfoFactory.intTypeInfo }, new int[] { vectorExpression.getOutputColumnNum() });
    Object[] scrqtchRow = new Object[1];
    batchSource.resetBatchIteration();
    int rowIndex = 0;
    while (true) {
        if (!batchSource.fillNextBatch(batch)) {
            break;
        }
        vectorExpression.evaluate(batch);
        extractResultObjects(batch, rowIndex, resultVectorExtractRow, scrqtchRow, TypeInfoFactory.intTypeInfo, resultObjects);
        rowIndex += batch.size;
    }
    return true;
}

Example 34 with VectorizedRowBatchCtx

use of org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx in project hive by apache.

the class TestVectorNegative method doVectorArithmeticTest.

private void doVectorArithmeticTest(TypeInfo typeInfo, List<String> columns, String[] columnNames, TypeInfo[] typeInfos, DataTypePhysicalVariation[] dataTypePhysicalVariations, List<ExprNodeDesc> children, ExprNodeGenericFuncDesc exprDesc, NegativeTestMode negativeTestMode, VectorRandomBatchSource batchSource, ObjectInspector objectInspector, TypeInfo outputTypeInfo, Object[] resultObjects) throws Exception {
    HiveConf hiveConf = new HiveConf();
    if (negativeTestMode == NegativeTestMode.ADAPTOR) {
        hiveConf.setBoolVar(HiveConf.ConfVars.HIVE_TEST_VECTOR_ADAPTOR_OVERRIDE, true);
    }
    VectorizationContext vectorizationContext = new VectorizationContext("name", columns, Arrays.asList(typeInfos), Arrays.asList(dataTypePhysicalVariations), hiveConf);
    VectorExpression vectorExpression = vectorizationContext.getVectorExpression(exprDesc);
    vectorExpression.transientInit(hiveConf);
    if (negativeTestMode == NegativeTestMode.VECTOR_EXPRESSION && vectorExpression instanceof VectorUDFAdaptor) {
        System.out.println("*NO NATIVE VECTOR EXPRESSION* typeInfo " + typeInfo.toString() + " negativeTestMode " + negativeTestMode + " vectorExpression " + vectorExpression.toString());
    }
    String[] outputScratchTypeNames = vectorizationContext.getScratchColumnTypeNames();
    VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx(columnNames, typeInfos, dataTypePhysicalVariations, /* dataColumnNums */
    null, /* partitionColumnCount */
    0, /* virtualColumnCount */
    0, /* neededVirtualColumns */
    null, outputScratchTypeNames, null);
    VectorizedRowBatch batch = batchContext.createVectorizedRowBatch();
    VectorExtractRow resultVectorExtractRow = new VectorExtractRow();
    resultVectorExtractRow.init(new TypeInfo[] { outputTypeInfo }, new int[] { vectorExpression.getOutputColumnNum() });
    Object[] scrqtchRow = new Object[1];
    // System.out.println("*VECTOR EXPRESSION* " + vectorExpression.getClass().getSimpleName());
    /*
    System.out.println(
        "*DEBUG* typeInfo " + typeInfo.toString() +
        " negativeTestMode " + negativeTestMode +
        " vectorExpression " + vectorExpression.toString());
    */
    batchSource.resetBatchIteration();
    int rowIndex = 0;
    while (true) {
        if (!batchSource.fillNextBatch(batch)) {
            break;
        }
        vectorExpression.evaluate(batch);
        extractResultObjects(batch, rowIndex, resultVectorExtractRow, scrqtchRow, objectInspector, resultObjects);
        rowIndex += batch.size;
    }
}