Search in sources :

Example 6 with VectorRandomBatchSource

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

the class TestVectorDateAddSub method doDateAddSubTestsWithDiffColumnScalar.

private void doDateAddSubTestsWithDiffColumnScalar(Random random, String dateTimeStringTypeName, String integerTypeName, ColumnScalarMode columnScalarMode, boolean isAdd) throws Exception {
    TypeInfo dateTimeStringTypeInfo = TypeInfoUtils.getTypeInfoFromTypeString(dateTimeStringTypeName);
    PrimitiveCategory dateTimeStringPrimitiveCategory = ((PrimitiveTypeInfo) dateTimeStringTypeInfo).getPrimitiveCategory();
    boolean isStringFamily = (dateTimeStringPrimitiveCategory == PrimitiveCategory.STRING || dateTimeStringPrimitiveCategory == PrimitiveCategory.CHAR || dateTimeStringPrimitiveCategory == PrimitiveCategory.VARCHAR);
    TypeInfo integerTypeInfo = TypeInfoUtils.getTypeInfoFromTypeString(integerTypeName);
    PrimitiveCategory integerPrimitiveCategory = ((PrimitiveTypeInfo) integerTypeInfo).getPrimitiveCategory();
    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) {
        if (!isStringFamily) {
            generationSpecList.add(GenerationSpec.createSameType(dateTimeStringTypeInfo));
        } else {
            generationSpecList.add(GenerationSpec.createStringFamilyOtherTypeValue(dateTimeStringTypeInfo, TypeInfoFactory.dateTypeInfo));
        }
        explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
        String columnName = "col" + (columnNum++);
        col1Expr = new ExprNodeColumnDesc(dateTimeStringTypeInfo, columnName, "table", false);
        columns.add(columnName);
    } else {
        Object scalar1Object;
        if (!isStringFamily) {
            scalar1Object = VectorRandomRowSource.randomPrimitiveObject(random, (PrimitiveTypeInfo) dateTimeStringTypeInfo);
        } else {
            scalar1Object = VectorRandomRowSource.randomStringFamilyOtherTypeValue(random, dateTimeStringTypeInfo, TypeInfoFactory.dateTypeInfo, false);
        }
        col1Expr = new ExprNodeConstantDesc(dateTimeStringTypeInfo, scalar1Object);
    }
    ExprNodeDesc col2Expr;
    if (columnScalarMode == ColumnScalarMode.COLUMN_COLUMN || columnScalarMode == ColumnScalarMode.SCALAR_COLUMN) {
        generationSpecList.add(GenerationSpec.createSameType(integerTypeInfo));
        explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
        String columnName = "col" + (columnNum++);
        col2Expr = new ExprNodeColumnDesc(integerTypeInfo, columnName, "table", false);
        columns.add(columnName);
    } else {
        Object scalar2Object = VectorRandomRowSource.randomPrimitiveObject(random, (PrimitiveTypeInfo) integerTypeInfo);
        scalar2Object = smallerRange(random, integerPrimitiveCategory, /* wantWritable */
        false);
        col2Expr = new ExprNodeConstantDesc(integerTypeInfo, 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);
    if (columnScalarMode == ColumnScalarMode.COLUMN_COLUMN || columnScalarMode == ColumnScalarMode.SCALAR_COLUMN) {
        // Fixup numbers to limit the range to 0 ... N-1.
        for (int i = 0; i < randomRows.length; i++) {
            Object[] row = randomRows[i];
            if (row[columnNum - 2] != null) {
                row[columnNum - 2] = smallerRange(random, integerPrimitiveCategory, /* wantWritable */
                true);
            }
        }
    }
    VectorRandomBatchSource batchSource = VectorRandomBatchSource.createInterestingBatches(random, rowSource, randomRows, null);
    String[] outputScratchTypeNames = new String[] { "date" };
    VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx(columnNames, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), /* dataColumnNums */
    null, /* partitionColumnCount */
    0, /* virtualColumnCount */
    0, /* neededVirtualColumns */
    null, outputScratchTypeNames, null);
    final int rowCount = randomRows.length;
    Object[][] resultObjectsArray = new Object[DateAddSubTestMode.count][];
    for (int i = 0; i < DateAddSubTestMode.count; i++) {
        Object[] resultObjects = new Object[rowCount];
        resultObjectsArray[i] = resultObjects;
        GenericUDF udf = (isAdd ? new GenericUDFDateAdd() : new GenericUDFDateSub());
        ExprNodeGenericFuncDesc exprDesc = new ExprNodeGenericFuncDesc(TypeInfoFactory.dateTypeInfo, udf, children);
        DateAddSubTestMode dateAddSubTestMode = DateAddSubTestMode.values()[i];
        switch(dateAddSubTestMode) {
            case ROW_MODE:
                doRowDateAddSubTest(dateTimeStringTypeInfo, integerTypeInfo, columns, children, isAdd, exprDesc, randomRows, columnScalarMode, rowSource.rowStructObjectInspector(), resultObjects);
                break;
            case ADAPTOR:
            case VECTOR_EXPRESSION:
                doVectorDateAddSubTest(dateTimeStringTypeInfo, integerTypeInfo, columns, rowSource.typeInfos(), children, isAdd, exprDesc, dateAddSubTestMode, columnScalarMode, batchSource, batchContext, resultObjects);
                break;
            default:
                throw new RuntimeException("Unexpected IF statement test mode " + dateAddSubTestMode);
        }
    }
    for (int i = 0; i < rowCount; i++) {
        // Row-mode is the expected value.
        Object expectedResult = resultObjectsArray[0][i];
        for (int v = 1; v < DateAddSubTestMode.count; v++) {
            Object vectorResult = resultObjectsArray[v][i];
            if (expectedResult == null || vectorResult == null) {
                if (expectedResult != null || vectorResult != null) {
                    Assert.fail("Row " + i + " " + DateAddSubTestMode.values()[v] + " isAdd " + isAdd + " " + 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 + " " + DateAddSubTestMode.values()[v] + " isAdd " + isAdd + " " + 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]));
                }
            }
        }
    }
}
Also used : ArrayList(java.util.ArrayList) GenericUDFDateAdd(org.apache.hadoop.hive.ql.udf.generic.GenericUDFDateAdd) PrimitiveTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo) DataTypePhysicalVariation(org.apache.hadoop.hive.common.type.DataTypePhysicalVariation) ExprNodeColumnDesc(org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) PrimitiveCategory(org.apache.hadoop.hive.serde2.objectinspector.PrimitiveObjectInspector.PrimitiveCategory) ExprNodeConstantDesc(org.apache.hadoop.hive.ql.plan.ExprNodeConstantDesc) VectorRandomBatchSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomBatchSource) GenericUDFDateSub(org.apache.hadoop.hive.ql.udf.generic.GenericUDFDateSub) ExprNodeGenericFuncDesc(org.apache.hadoop.hive.ql.plan.ExprNodeGenericFuncDesc) PrimitiveTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo) DecimalTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.DecimalTypeInfo) TypeInfo(org.apache.hadoop.hive.serde2.typeinfo.TypeInfo) GenerationSpec(org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource.GenerationSpec) VectorizedRowBatchCtx(org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx) GenericUDF(org.apache.hadoop.hive.ql.udf.generic.GenericUDF) VectorRandomRowSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource)

Example 7 with VectorRandomBatchSource

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

the class TestVectorStructField method doOneStructFieldTest.

private void doOneStructFieldTest(Random random, StructTypeInfo structTypeInfo, String structTypeName, int fieldIndex) throws Exception {
    List<GenerationSpec> generationSpecList = new ArrayList<GenerationSpec>();
    List<DataTypePhysicalVariation> explicitDataTypePhysicalVariationList = new ArrayList<DataTypePhysicalVariation>();
    List<String> columns = new ArrayList<String>();
    int columnNum = 1;
    generationSpecList.add(GenerationSpec.createSameType(structTypeInfo));
    explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
    ExprNodeDesc col1Expr;
    String columnName = "col" + (columnNum++);
    col1Expr = new ExprNodeColumnDesc(structTypeInfo, columnName, "table", false);
    columns.add(columnName);
    ObjectInspector structObjectInspector = VectorRandomRowSource.getObjectInspector(structTypeInfo);
    List<ObjectInspector> objectInspectorList = new ArrayList<ObjectInspector>();
    objectInspectorList.add(structObjectInspector);
    List<ExprNodeDesc> children = new ArrayList<ExprNodeDesc>();
    children.add(col1Expr);
    // ----------------------------------------------------------------------------------------------
    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);
    List<String> fieldNameList = structTypeInfo.getAllStructFieldNames();
    List<TypeInfo> fieldTypeInfoList = structTypeInfo.getAllStructFieldTypeInfos();
    String randomFieldName = fieldNameList.get(fieldIndex);
    TypeInfo outputTypeInfo = fieldTypeInfoList.get(fieldIndex);
    ExprNodeFieldDesc exprNodeFieldDesc = new ExprNodeFieldDesc(outputTypeInfo, col1Expr, randomFieldName, /* isList */
    false);
    final int rowCount = randomRows.length;
    Object[][] resultObjectsArray = new Object[StructFieldTestMode.count][];
    for (int i = 0; i < StructFieldTestMode.count; i++) {
        Object[] resultObjects = new Object[rowCount];
        resultObjectsArray[i] = resultObjects;
        StructFieldTestMode negativeTestMode = StructFieldTestMode.values()[i];
        switch(negativeTestMode) {
            case ROW_MODE:
                doRowStructFieldTest(structTypeInfo, columns, children, exprNodeFieldDesc, randomRows, rowSource.rowStructObjectInspector(), outputTypeInfo, resultObjects);
                break;
            case VECTOR_EXPRESSION:
                doVectorStructFieldTest(structTypeInfo, columns, columnNames, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), children, exprNodeFieldDesc, negativeTestMode, batchSource, exprNodeFieldDesc.getWritableObjectInspector(), outputTypeInfo, resultObjects);
                break;
            default:
                throw new RuntimeException("Unexpected Negative operator test mode " + negativeTestMode);
        }
    }
    for (int i = 0; i < rowCount; i++) {
        // Row-mode is the expected value.
        Object expectedResult = resultObjectsArray[0][i];
        for (int v = 1; v < StructFieldTestMode.count; v++) {
            Object vectorResult = resultObjectsArray[v][i];
            if (expectedResult == null || vectorResult == null) {
                if (expectedResult != null || vectorResult != null) {
                    Assert.fail("Row " + i + " structTypeName " + structTypeName + " outputTypeName " + outputTypeInfo.getTypeName() + " " + StructFieldTestMode.values()[v] + " 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 + " structTypeName " + structTypeName + " outputTypeName " + outputTypeInfo.getTypeName() + " " + StructFieldTestMode.values()[v] + " result " + vectorResult.toString() + " (" + vectorResult.getClass().getSimpleName() + ")" + " does not match row-mode expected result " + expectedResult.toString() + " (" + expectedResult.getClass().getSimpleName() + ")" + " row values " + Arrays.toString(randomRows[i]));
                }
            }
        }
    }
}
Also used : ObjectInspector(org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector) VectorRandomBatchSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomBatchSource) ArrayList(java.util.ArrayList) StructTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.StructTypeInfo) PrimitiveTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo) DecimalTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.DecimalTypeInfo) TypeInfo(org.apache.hadoop.hive.serde2.typeinfo.TypeInfo) GenerationSpec(org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource.GenerationSpec) DataTypePhysicalVariation(org.apache.hadoop.hive.common.type.DataTypePhysicalVariation) ExprNodeFieldDesc(org.apache.hadoop.hive.ql.plan.ExprNodeFieldDesc) ExprNodeColumnDesc(org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) VectorRandomRowSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource)

Example 8 with VectorRandomBatchSource

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

the class TestMapJoinOperator method createExpectedTestRowMultiSet.

/*
   * Simulate the join by driving the test big table data by our test small table HashMap and
   * create the expected output as a multi-set of TestRow (i.e. TestRow and occurrence count).
   */
private RowTestObjectsMultiSet createExpectedTestRowMultiSet(MapJoinTestDescription testDesc, MapJoinTestData testData) throws HiveException {
    RowTestObjectsMultiSet expectedTestRowMultiSet = new RowTestObjectsMultiSet();
    VectorExtractRow vectorExtractRow = new VectorExtractRow();
    vectorExtractRow.init(testDesc.bigTableTypeInfos);
    final int bigTableColumnCount = testDesc.bigTableTypeInfos.length;
    Object[] bigTableRowObjects = new Object[bigTableColumnCount];
    final int bigTableKeyColumnCount = testDesc.bigTableKeyTypeInfos.length;
    Object[] bigTableKeyObjects = new Object[bigTableKeyColumnCount];
    VectorRandomBatchSource bigTableBatchSource = testData.getBigTableBatchSource();
    VectorizedRowBatch batch = testData.getBigTableBatch();
    bigTableBatchSource.resetBatchIteration();
    while (bigTableBatchSource.fillNextBatch(batch)) {
        final int size = testData.bigTableBatch.size;
        for (int r = 0; r < size; r++) {
            vectorExtractRow.extractRow(testData.bigTableBatch, r, bigTableRowObjects);
            // Form key object array
            // NULLs may be present in {FULL|LEFT|RIGHT} OUTER joins.
            boolean hasAnyNulls = false;
            for (int k = 0; k < bigTableKeyColumnCount; k++) {
                int keyColumnNum = testDesc.bigTableKeyColumnNums[k];
                Object keyObject = bigTableRowObjects[keyColumnNum];
                if (keyObject == null) {
                    hasAnyNulls = true;
                }
                bigTableKeyObjects[k] = keyObject;
                bigTableKeyObjects[k] = ((PrimitiveObjectInspector) testDesc.bigTableObjectInspectors[keyColumnNum]).copyObject(bigTableKeyObjects[k]);
            }
            RowTestObjects testKey = new RowTestObjects(bigTableKeyObjects);
            if (testData.smallTableKeyHashMap.containsKey(testKey) && !hasAnyNulls) {
                int smallTableKeyIndex = testData.smallTableKeyHashMap.get(testKey);
                switch(testDesc.vectorMapJoinVariation) {
                    case INNER:
                    case OUTER:
                    case FULL_OUTER:
                        {
                            // One row per value.
                            ArrayList<RowTestObjects> valueList = testData.smallTableValues.get(smallTableKeyIndex);
                            final int valueCount = valueList.size();
                            for (int v = 0; v < valueCount; v++) {
                                Object[] outputObjects = new Object[testDesc.outputColumnNames.length];
                                addBigTableRetained(testDesc, bigTableRowObjects, outputObjects);
                                int outputColumnNum = testDesc.bigTableRetainColumnNums.length;
                                final int smallTableRetainKeyColumnNumsLength = testDesc.smallTableRetainKeyColumnNums.length;
                                for (int o = 0; o < smallTableRetainKeyColumnNumsLength; o++) {
                                    outputObjects[outputColumnNum++] = bigTableKeyObjects[testDesc.smallTableRetainKeyColumnNums[o]];
                                }
                                Object[] valueRow = valueList.get(v).getRow();
                                final int smallTableRetainValueColumnNumsLength = testDesc.smallTableRetainValueColumnNums.length;
                                for (int o = 0; o < smallTableRetainValueColumnNumsLength; o++) {
                                    outputObjects[outputColumnNum++] = valueRow[testDesc.smallTableRetainValueColumnNums[o]];
                                }
                                addToOutput(testDesc, expectedTestRowMultiSet, outputObjects, RowTestObjectsMultiSet.RowFlag.REGULAR);
                            }
                        }
                        break;
                    case INNER_BIG_ONLY:
                    case LEFT_SEMI:
                    case LEFT_ANTI:
                        {
                            Object[] outputObjects = new Object[testDesc.outputColumnNames.length];
                            addBigTableRetained(testDesc, bigTableRowObjects, outputObjects);
                            int outputColumnNum = testDesc.bigTableRetainColumnNums.length;
                            final int smallTableRetainKeyColumnNumsLength = testDesc.smallTableRetainKeyColumnNums.length;
                            for (int o = 0; o < smallTableRetainKeyColumnNumsLength; o++) {
                                outputObjects[outputColumnNum++] = bigTableKeyObjects[testDesc.smallTableRetainKeyColumnNums[o]];
                            }
                            addToOutput(testDesc, expectedTestRowMultiSet, outputObjects, RowTestObjectsMultiSet.RowFlag.REGULAR);
                        }
                        break;
                    default:
                        throw new RuntimeException("Unknown operator variation " + testDesc.vectorMapJoinVariation);
                }
            } else {
                if (testDesc.vectorMapJoinVariation == VectorMapJoinVariation.OUTER || testDesc.vectorMapJoinVariation == VectorMapJoinVariation.FULL_OUTER) {
                    // We need to add a non-match row with nulls for small table values.
                    Object[] outputObjects = new Object[testDesc.outputColumnNames.length];
                    addBigTableRetained(testDesc, bigTableRowObjects, outputObjects);
                    int outputColumnNum = testDesc.bigTableRetainColumnNums.length;
                    final int smallTableRetainKeyColumnNumsLength = testDesc.smallTableRetainKeyColumnNums.length;
                    for (int o = 0; o < smallTableRetainKeyColumnNumsLength; o++) {
                        outputObjects[outputColumnNum++] = null;
                    }
                    final int smallTableRetainValueColumnNumsLength = testDesc.smallTableRetainValueColumnNums.length;
                    for (int o = 0; o < smallTableRetainValueColumnNumsLength; o++) {
                        outputObjects[outputColumnNum++] = null;
                    }
                    addToOutput(testDesc, expectedTestRowMultiSet, outputObjects, RowTestObjectsMultiSet.RowFlag.LEFT_OUTER);
                }
            }
        }
    }
    if (testDesc.vectorMapJoinVariation == VectorMapJoinVariation.FULL_OUTER) {
        System.out.println("*BENCHMARK* ----------------------------------------------------------------------");
        System.out.println("*BENCHMARK* FULL OUTER non-match key count " + testData.fullOuterAdditionalSmallTableKeys.size());
        // Fill in non-match Small Table key results.
        for (RowTestObjects smallTableKey : testData.fullOuterAdditionalSmallTableKeys) {
            // System.out.println(
            // "*BENCHMARK* fullOuterAdditionalSmallTableKey " + smallTableKey.toString());
            int smallTableKeyIndex = testData.smallTableKeyHashMap.get(smallTableKey);
            // One row per value.
            ArrayList<RowTestObjects> valueList = testData.smallTableValues.get(smallTableKeyIndex);
            final int valueCount = valueList.size();
            for (int v = 0; v < valueCount; v++) {
                Object[] outputObjects = new Object[testDesc.outputColumnNames.length];
                // Non-match Small Table keys produce NULL Big Table columns.
                final int bigTableRetainColumnNumsLength = testDesc.bigTableRetainColumnNums.length;
                for (int o = 0; o < bigTableRetainColumnNumsLength; o++) {
                    outputObjects[o] = null;
                }
                int outputColumnNum = testDesc.bigTableRetainColumnNums.length;
                // The output result may include 0, 1, or more small key columns...
                Object[] smallKeyObjects = smallTableKey.getRow();
                final int smallTableRetainKeyColumnNumsLength = testDesc.smallTableRetainKeyColumnNums.length;
                for (int o = 0; o < smallTableRetainKeyColumnNumsLength; o++) {
                    outputObjects[outputColumnNum++] = smallKeyObjects[testDesc.smallTableRetainKeyColumnNums[o]];
                }
                Object[] valueRow = valueList.get(v).getRow();
                final int smallTableRetainValueColumnNumsLength = testDesc.smallTableRetainValueColumnNums.length;
                for (int o = 0; o < smallTableRetainValueColumnNumsLength; o++) {
                    outputObjects[outputColumnNum++] = valueRow[testDesc.smallTableRetainValueColumnNums[o]];
                }
                addToOutput(testDesc, expectedTestRowMultiSet, outputObjects, RowTestObjectsMultiSet.RowFlag.FULL_OUTER);
            }
        }
    }
    return expectedTestRowMultiSet;
}
Also used : VectorizedRowBatch(org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatch) VectorRandomBatchSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomBatchSource) ArrayList(java.util.ArrayList) RowTestObjectsMultiSet(org.apache.hadoop.hive.ql.exec.util.rowobjects.RowTestObjectsMultiSet) VectorExtractRow(org.apache.hadoop.hive.ql.exec.vector.VectorExtractRow) RowTestObjects(org.apache.hadoop.hive.ql.exec.util.rowobjects.RowTestObjects)

Example 9 with VectorRandomBatchSource

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

the class TestMapJoinOperator method generateBigAndSmallTableRowLogLines.

private void generateBigAndSmallTableRowLogLines(MapJoinTestDescription testDesc, MapJoinTestData testData) throws HiveException {
    // Generate Big Table rows log lines...
    VectorExtractRow vectorExtractRow = new VectorExtractRow();
    vectorExtractRow.init(testDesc.bigTableTypeInfos);
    final int bigTableColumnCount = testDesc.bigTableTypeInfos.length;
    Object[] bigTableRowObjects = new Object[bigTableColumnCount];
    /*
    PrintStream big_ps;
    try {
      big_ps = new PrintStream("/Users/mmccline/VecFullOuterRefresh/out_big");
    } catch (Exception e) {
      throw new HiveException(e);
    }
    */
    VectorRandomBatchSource bigTableBatchSource = testData.getBigTableBatchSource();
    VectorizedRowBatch batch = testData.getBigTableBatch();
    bigTableBatchSource.resetBatchIteration();
    while (bigTableBatchSource.fillNextBatch(batch)) {
        final int size = testData.bigTableBatch.size;
        for (int r = 0; r < size; r++) {
            vectorExtractRow.extractRow(testData.bigTableBatch, r, bigTableRowObjects);
        // big_ps.println(rowToCsvString(bigTableRowObjects));
        }
    }
    // big_ps.close();
    /*
    PrintStream small_ps;
    try {
      small_ps = new PrintStream("/Users/mmccline/VecFullOuterRefresh/out_small");
    } catch (Exception e) {
      throw new HiveException(e);
    }
    */
    // Generate Small Table rows log lines...
    final int keyKeyColumnNumsLength = testDesc.bigTableKeyColumnNums.length;
    final int smallTableRetainValueLength = testDesc.smallTableRetainValueColumnNums.length;
    final int smallTableLength = keyKeyColumnNumsLength + smallTableRetainValueLength;
    for (Entry<RowTestObjects, Integer> entry : testData.smallTableKeyHashMap.entrySet()) {
        if (smallTableRetainValueLength == 0) {
            Object[] smallTableRowObjects = entry.getKey().getRow();
        // small_ps.println(rowToCsvString(smallTableRowObjects));
        } else {
            Integer valueIndex = entry.getValue();
            ArrayList<RowTestObjects> valueList = testData.smallTableValues.get(valueIndex);
            final int valueCount = valueList.size();
            for (int v = 0; v < valueCount; v++) {
                Object[] smallTableRowObjects = new Object[smallTableLength];
                System.arraycopy(entry.getKey().getRow(), 0, smallTableRowObjects, 0, keyKeyColumnNumsLength);
                int outputColumnNum = keyKeyColumnNumsLength;
                Object[] valueRow = valueList.get(v).getRow();
                for (int o = 0; o < smallTableRetainValueLength; o++) {
                    smallTableRowObjects[outputColumnNum++] = valueRow[testDesc.smallTableRetainValueColumnNums[o]];
                }
            // small_ps.println(rowToCsvString(smallTableRowObjects));
            }
        }
    }
// small_ps.close();
}
Also used : VectorizedRowBatch(org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatch) VectorRandomBatchSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomBatchSource) VectorExtractRow(org.apache.hadoop.hive.ql.exec.vector.VectorExtractRow) RowTestObjects(org.apache.hadoop.hive.ql.exec.util.rowobjects.RowTestObjects)

Example 10 with VectorRandomBatchSource

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

the class TestVectorDateDiff method doDateDiffTestsWithDiffColumnScalar.

private void doDateDiffTestsWithDiffColumnScalar(Random random, String dateTimeStringTypeName1, String dateTimeStringTypeName2, ColumnScalarMode columnScalarMode) throws Exception {
    TypeInfo dateTimeStringTypeInfo1 = TypeInfoUtils.getTypeInfoFromTypeString(dateTimeStringTypeName1);
    PrimitiveCategory dateTimeStringPrimitiveCategory1 = ((PrimitiveTypeInfo) dateTimeStringTypeInfo1).getPrimitiveCategory();
    boolean isStringFamily1 = (dateTimeStringPrimitiveCategory1 == PrimitiveCategory.STRING || dateTimeStringPrimitiveCategory1 == PrimitiveCategory.CHAR || dateTimeStringPrimitiveCategory1 == PrimitiveCategory.VARCHAR);
    TypeInfo dateTimeStringTypeInfo2 = TypeInfoUtils.getTypeInfoFromTypeString(dateTimeStringTypeName2);
    PrimitiveCategory dateTimeStringPrimitiveCategory2 = ((PrimitiveTypeInfo) dateTimeStringTypeInfo2).getPrimitiveCategory();
    boolean isStringFamily2 = (dateTimeStringPrimitiveCategory2 == PrimitiveCategory.STRING || dateTimeStringPrimitiveCategory2 == PrimitiveCategory.CHAR || dateTimeStringPrimitiveCategory2 == PrimitiveCategory.VARCHAR);
    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) {
        if (!isStringFamily1) {
            generationSpecList.add(GenerationSpec.createSameType(dateTimeStringTypeInfo1));
        } else {
            generationSpecList.add(GenerationSpec.createStringFamilyOtherTypeValue(dateTimeStringTypeInfo1, TypeInfoFactory.dateTypeInfo));
        }
        explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
        String columnName = "col" + (columnNum++);
        col1Expr = new ExprNodeColumnDesc(dateTimeStringTypeInfo1, columnName, "table", false);
        columns.add(columnName);
    } else {
        Object scalar1Object;
        if (!isStringFamily1) {
            scalar1Object = VectorRandomRowSource.randomPrimitiveObject(random, (PrimitiveTypeInfo) dateTimeStringTypeInfo1);
        } else {
            scalar1Object = VectorRandomRowSource.randomStringFamilyOtherTypeValue(random, dateTimeStringTypeInfo1, TypeInfoFactory.dateTypeInfo, false);
        }
        col1Expr = new ExprNodeConstantDesc(dateTimeStringTypeInfo1, scalar1Object);
    }
    ExprNodeDesc col2Expr;
    if (columnScalarMode == ColumnScalarMode.COLUMN_COLUMN || columnScalarMode == ColumnScalarMode.SCALAR_COLUMN) {
        if (!isStringFamily2) {
            generationSpecList.add(GenerationSpec.createSameType(dateTimeStringTypeInfo2));
        } else {
            generationSpecList.add(GenerationSpec.createStringFamilyOtherTypeValue(dateTimeStringTypeInfo2, TypeInfoFactory.dateTypeInfo));
        }
        explicitDataTypePhysicalVariationList.add(DataTypePhysicalVariation.NONE);
        String columnName = "col" + (columnNum++);
        col2Expr = new ExprNodeColumnDesc(dateTimeStringTypeInfo2, columnName, "table", false);
        columns.add(columnName);
    } else {
        Object scalar2Object;
        if (!isStringFamily2) {
            scalar2Object = VectorRandomRowSource.randomPrimitiveObject(random, (PrimitiveTypeInfo) dateTimeStringTypeInfo2);
        } else {
            scalar2Object = VectorRandomRowSource.randomStringFamilyOtherTypeValue(random, dateTimeStringTypeInfo2, TypeInfoFactory.dateTypeInfo, false);
        }
        col2Expr = new ExprNodeConstantDesc(dateTimeStringTypeInfo2, 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[] { "date" };
    VectorizedRowBatchCtx batchContext = new VectorizedRowBatchCtx(columnNames, rowSource.typeInfos(), rowSource.dataTypePhysicalVariations(), /* dataColumnNums */
    null, /* partitionColumnCount */
    0, /* virtualColumnCount */
    0, /* neededVirtualColumns */
    null, outputScratchTypeNames, null);
    final int rowCount = randomRows.length;
    Object[][] resultObjectsArray = new Object[DateDiffTestMode.count][];
    for (int i = 0; i < DateDiffTestMode.count; i++) {
        Object[] resultObjects = new Object[rowCount];
        resultObjectsArray[i] = resultObjects;
        GenericUDF udf = new GenericUDFDateDiff();
        ExprNodeGenericFuncDesc exprDesc = new ExprNodeGenericFuncDesc(TypeInfoFactory.intTypeInfo, udf, children);
        DateDiffTestMode dateDiffTestMode = DateDiffTestMode.values()[i];
        switch(dateDiffTestMode) {
            case ROW_MODE:
                doRowDateAddSubTest(dateTimeStringTypeInfo1, dateTimeStringTypeInfo2, columns, children, exprDesc, randomRows, columnScalarMode, rowSource.rowStructObjectInspector(), resultObjects);
                break;
            case ADAPTOR:
            case VECTOR_EXPRESSION:
                doVectorDateAddSubTest(dateTimeStringTypeInfo1, dateTimeStringTypeInfo2, columns, rowSource.typeInfos(), children, exprDesc, dateDiffTestMode, columnScalarMode, batchSource, batchContext, resultObjects);
                break;
            default:
                throw new RuntimeException("Unexpected IF statement test mode " + dateDiffTestMode);
        }
    }
    for (int i = 0; i < rowCount; i++) {
        // Row-mode is the expected value.
        Object expectedResult = resultObjectsArray[0][i];
        for (int v = 1; v < DateDiffTestMode.count; v++) {
            Object vectorResult = resultObjectsArray[v][i];
            if (expectedResult == null || vectorResult == null) {
                if (expectedResult != null || vectorResult != null) {
                    Assert.fail("Row " + i + " " + DateDiffTestMode.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 + " " + DateDiffTestMode.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]));
                }
            }
        }
    }
}
Also used : ArrayList(java.util.ArrayList) PrimitiveTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo) DataTypePhysicalVariation(org.apache.hadoop.hive.common.type.DataTypePhysicalVariation) ExprNodeColumnDesc(org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc) ExprNodeDesc(org.apache.hadoop.hive.ql.plan.ExprNodeDesc) PrimitiveCategory(org.apache.hadoop.hive.serde2.objectinspector.PrimitiveObjectInspector.PrimitiveCategory) ExprNodeConstantDesc(org.apache.hadoop.hive.ql.plan.ExprNodeConstantDesc) VectorRandomBatchSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomBatchSource) ExprNodeGenericFuncDesc(org.apache.hadoop.hive.ql.plan.ExprNodeGenericFuncDesc) PrimitiveTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo) DecimalTypeInfo(org.apache.hadoop.hive.serde2.typeinfo.DecimalTypeInfo) TypeInfo(org.apache.hadoop.hive.serde2.typeinfo.TypeInfo) GenerationSpec(org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource.GenerationSpec) GenericUDFDateDiff(org.apache.hadoop.hive.ql.udf.generic.GenericUDFDateDiff) VectorizedRowBatchCtx(org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx) GenericUDF(org.apache.hadoop.hive.ql.udf.generic.GenericUDF) VectorRandomRowSource(org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource)

Aggregations

VectorRandomBatchSource (org.apache.hadoop.hive.ql.exec.vector.VectorRandomBatchSource)22 ArrayList (java.util.ArrayList)21 DataTypePhysicalVariation (org.apache.hadoop.hive.common.type.DataTypePhysicalVariation)19 VectorRandomRowSource (org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource)19 ExprNodeColumnDesc (org.apache.hadoop.hive.ql.plan.ExprNodeColumnDesc)19 ExprNodeDesc (org.apache.hadoop.hive.ql.plan.ExprNodeDesc)19 TypeInfo (org.apache.hadoop.hive.serde2.typeinfo.TypeInfo)19 GenerationSpec (org.apache.hadoop.hive.ql.exec.vector.VectorRandomRowSource.GenerationSpec)18 DecimalTypeInfo (org.apache.hadoop.hive.serde2.typeinfo.DecimalTypeInfo)16 PrimitiveTypeInfo (org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo)16 GenericUDF (org.apache.hadoop.hive.ql.udf.generic.GenericUDF)14 ExprNodeConstantDesc (org.apache.hadoop.hive.ql.plan.ExprNodeConstantDesc)11 ExprNodeGenericFuncDesc (org.apache.hadoop.hive.ql.plan.ExprNodeGenericFuncDesc)11 ObjectInspector (org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector)11 PrimitiveCategory (org.apache.hadoop.hive.serde2.objectinspector.PrimitiveObjectInspector.PrimitiveCategory)8 SemanticException (org.apache.hadoop.hive.ql.parse.SemanticException)6 CharTypeInfo (org.apache.hadoop.hive.serde2.typeinfo.CharTypeInfo)6 VarcharTypeInfo (org.apache.hadoop.hive.serde2.typeinfo.VarcharTypeInfo)6 VectorizedRowBatchCtx (org.apache.hadoop.hive.ql.exec.vector.VectorizedRowBatchCtx)5 FunctionInfo (org.apache.hadoop.hive.ql.exec.FunctionInfo)4