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

Example 76 with VectorizedRowBatch

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

the class TestVectorMathFunctions method testVectorSqrt.

@Test
public void testVectorSqrt() throws HiveException {
    VectorizedRowBatch b = getVectorizedRowBatchDoubleInDoubleOut();
    DoubleColumnVector resultV = (DoubleColumnVector) b.cols[1];
    b.cols[0].noNulls = true;
    VectorExpression expr = new FuncSqrtDoubleToDouble(0, 1);
    expr.evaluate(b);
    Assert.assertEquals(Math.sqrt(0.5d), resultV.vector[4], Double.MIN_VALUE);
}

Example 77 with VectorizedRowBatch

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

the class TestVectorMathFunctions method testVectorLn.

@Test
public void testVectorLn() throws HiveException {
    // test double->double version
    VectorizedRowBatch b = getVectorizedRowBatchDoubleInDoubleOut();
    DoubleColumnVector resultV = (DoubleColumnVector) b.cols[1];
    b.cols[0].noNulls = true;
    VectorExpression expr = new FuncLnDoubleToDouble(0, 1);
    expr.evaluate(b);
    Assert.assertEquals(Math.log(0.5), resultV.vector[4], Double.MIN_VALUE);
    // test long->double version
    b = getVectorizedRowBatchLongInDoubleOut();
    resultV = (DoubleColumnVector) b.cols[1];
    b.cols[0].noNulls = true;
    expr = new FuncLnLongToDouble(0, 1);
    expr.evaluate(b);
    Assert.assertEquals(Math.log(2), resultV.vector[4], Double.MIN_VALUE);
}

Example 78 with VectorizedRowBatch

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

the class TestVectorNull method doVectorCastTest.

private boolean doVectorCastTest(TypeInfo typeInfo, boolean isFilter, List<String> columns, String[] columnNames, TypeInfo[] typeInfos, DataTypePhysicalVariation[] dataTypePhysicalVariations, List<ExprNodeDesc> children, GenericUDF udf, ExprNodeGenericFuncDesc exprDesc, NullTestMode nullTestMode, VectorRandomBatchSource batchSource, ObjectInspector objectInspector, TypeInfo outputTypeInfo, Object[] resultObjects) throws Exception {
    HiveConf hiveConf = new HiveConf();
    if (nullTestMode == NullTestMode.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, (isFilter ? VectorExpressionDescriptor.Mode.FILTER : VectorExpressionDescriptor.Mode.PROJECTION));
    vectorExpression.transientInit(hiveConf);
    if (nullTestMode == NullTestMode.VECTOR_EXPRESSION && vectorExpression instanceof VectorUDFAdaptor) {
        System.out.println("*NO NATIVE VECTOR EXPRESSION* typeInfo " + typeInfo.toString() + " nullTestMode " + nullTestMode + " isFilter " + isFilter + " vectorExpression " + vectorExpression.toString());
    }
    // System.out.println("*VECTOR EXPRESSION* " + vectorExpression.getClass().getSimpleName());
    /*
    System.out.println(
        "*DEBUG* typeInfo " + typeInfo.toString() +
        " nullTestMode " + nullTestMode +
        " isFilter " + isFilter +
        " vectorExpression " + vectorExpression.toString());
    */
    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 = null;
    Object[] scrqtchRow = null;
    if (!isFilter) {
        resultVectorExtractRow = new VectorExtractRow();
        final int outputColumnNum = vectorExpression.getOutputColumnNum();
        resultVectorExtractRow.init(new TypeInfo[] { outputTypeInfo }, new int[] { outputColumnNum });
        scrqtchRow = new Object[1];
    }
    boolean copySelectedInUse = false;
    int[] copySelected = new int[VectorizedRowBatch.DEFAULT_SIZE];
    batchSource.resetBatchIteration();
    int rowIndex = 0;
    while (true) {
        if (!batchSource.fillNextBatch(batch)) {
            break;
        }
        final int originalBatchSize = batch.size;
        if (isFilter) {
            copySelectedInUse = batch.selectedInUse;
            if (batch.selectedInUse) {
                System.arraycopy(batch.selected, 0, copySelected, 0, originalBatchSize);
            }
        }
        // In filter mode, the batch size can be made smaller.
        vectorExpression.evaluate(batch);
        if (!isFilter) {
            extractResultObjects(batch, rowIndex, resultVectorExtractRow, scrqtchRow, objectInspector, resultObjects);
        } else {
            final int currentBatchSize = batch.size;
            if (copySelectedInUse && batch.selectedInUse) {
                int selectIndex = 0;
                for (int i = 0; i < originalBatchSize; i++) {
                    final int originalBatchIndex = copySelected[i];
                    final boolean booleanResult;
                    if (selectIndex < currentBatchSize && batch.selected[selectIndex] == originalBatchIndex) {
                        booleanResult = true;
                        selectIndex++;
                    } else {
                        booleanResult = false;
                    }
                    resultObjects[rowIndex + i] = new BooleanWritable(booleanResult);
                }
            } else if (batch.selectedInUse) {
                int selectIndex = 0;
                for (int i = 0; i < originalBatchSize; i++) {
                    final boolean booleanResult;
                    if (selectIndex < currentBatchSize && batch.selected[selectIndex] == i) {
                        booleanResult = true;
                        selectIndex++;
                    } else {
                        booleanResult = false;
                    }
                    resultObjects[rowIndex + i] = new BooleanWritable(booleanResult);
                }
            } else if (currentBatchSize == 0) {
                // Whole batch got zapped.
                for (int i = 0; i < originalBatchSize; i++) {
                    resultObjects[rowIndex + i] = new BooleanWritable(false);
                }
            } else {
                // Every row kept.
                for (int i = 0; i < originalBatchSize; i++) {
                    resultObjects[rowIndex + i] = new BooleanWritable(true);
                }
            }
        }
        rowIndex += originalBatchSize;
    }
    return true;
}

Example 79 with VectorizedRowBatch

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

the class TestVectorStringConcat method doVectorStringConcatTest.

private void doVectorStringConcatTest(TypeInfo stringTypeInfo1, TypeInfo stringTypeInfo2, List<String> columns, TypeInfo[] typeInfos, List<ExprNodeDesc> children, StringConcatTestMode stringConcatTestMode, ColumnScalarMode columnScalarMode, VectorRandomBatchSource batchSource, VectorizedRowBatchCtx batchContext, ObjectInspector rowInspector, GenericUDF genericUdf, Object[] resultObjects) throws Exception {
    HiveConf hiveConf = new HiveConf();
    if (stringConcatTestMode == StringConcatTestMode.ADAPTOR) {
        hiveConf.setBoolVar(HiveConf.ConfVars.HIVE_TEST_VECTOR_ADAPTOR_OVERRIDE, true);
    }
    DataTypePhysicalVariation[] dataTypePhysicalVariations = new DataTypePhysicalVariation[2];
    Arrays.fill(dataTypePhysicalVariations, DataTypePhysicalVariation.NONE);
    ExprNodeGenericFuncDesc exprDesc = new ExprNodeGenericFuncDesc(TypeInfoFactory.stringTypeInfo, genericUdf, children);
    // ---------------------------------------
    // Just so we can get the output type...
    ExprNodeEvaluator evaluator = ExprNodeEvaluatorFactory.get(exprDesc, hiveConf);
    evaluator.initialize(rowInspector);
    ObjectInspector objectInspector = evaluator.getOutputOI();
    TypeInfo outputTypeInfo = TypeInfoUtils.getTypeInfoFromObjectInspector(objectInspector);
    /*
     * Again with correct output type...
     */
    exprDesc = new ExprNodeGenericFuncDesc(outputTypeInfo, genericUdf, children);
    // ---------------------------------------
    VectorizationContext vectorizationContext = new VectorizationContext("name", columns, Arrays.asList(typeInfos), Arrays.asList(dataTypePhysicalVariations), hiveConf);
    VectorExpression vectorExpression = vectorizationContext.getVectorExpression(exprDesc);
    vectorExpression.transientInit(hiveConf);
    if (stringConcatTestMode == StringConcatTestMode.VECTOR_EXPRESSION && vectorExpression instanceof VectorUDFAdaptor) {
        System.out.println("*NO NATIVE VECTOR EXPRESSION* stringTypeInfo1 " + stringTypeInfo1.toString() + " stringTypeInfo2 " + stringTypeInfo2.toString() + " stringConcatTestMode " + stringConcatTestMode + " columnScalarMode " + columnScalarMode + " vectorExpression " + vectorExpression.toString());
    }
    VectorizedRowBatch batch = batchContext.createVectorizedRowBatch();
    VectorExtractRow resultVectorExtractRow = new VectorExtractRow();
    resultVectorExtractRow.init(new TypeInfo[] { outputTypeInfo }, new int[] { columns.size() });
    Object[] scrqtchRow = new Object[1];
    // System.out.println("*VECTOR EXPRESSION* " + vectorExpression.getClass().getSimpleName());
    /*
    System.out.println(
        "*DEBUG* stringTypeInfo1 " + stringTypeInfo1.toString() +
        " stringTypeInfo2 " + stringTypeInfo2.toString() +
        " stringConcatTestMode " + stringConcatTestMode +
        " columnScalarMode " + columnScalarMode +
        " 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;
    }
}

Example 80 with VectorizedRowBatch

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

the class TestVectorFilterExpressions method testFilterLongColLessLongColumn.

@Test
public void testFilterLongColLessLongColumn() throws HiveException {
    int seed = 17;
    VectorizedRowBatch vrg = VectorizedRowGroupGenUtil.getVectorizedRowBatch(5, 3, seed);
    LongColumnVector lcv0 = (LongColumnVector) vrg.cols[0];
    LongColumnVector lcv1 = (LongColumnVector) vrg.cols[1];
    LongColumnVector lcv2 = (LongColumnVector) vrg.cols[2];
    FilterLongColLessLongColumn expr = new FilterLongColLessLongColumn(2, 1);
    LongColAddLongScalar childExpr = new LongColAddLongScalar(0, 10, 2);
    expr.setChildExpressions(new VectorExpression[] { childExpr });
    // Basic case
    lcv0.vector[0] = 10;
    lcv0.vector[1] = 20;
    lcv0.vector[2] = 9;
    lcv0.vector[3] = 20;
    lcv0.vector[4] = 10;
    lcv1.vector[0] = 20;
    lcv1.vector[1] = 10;
    lcv1.vector[2] = 20;
    lcv1.vector[3] = 10;
    lcv1.vector[4] = 20;
    expr.evaluate(vrg);
    assertEquals(1, vrg.size);
    assertEquals(2, vrg.selected[0]);
}