Examples with RowSet org.apache.drill.test.rowSet.RowSet used on opensource projects

Example 11 with RowSet

use of org.apache.drill.test.rowSet.RowSet in project drill by axbaretto.

the class TestRecordBatchSizer method testSizerRepeatedFixedWidth.

@Test
public void testSizerRepeatedFixedWidth() {
    BatchSchema schema = new SchemaBuilder().addArray("a", MinorType.BIGINT).addArray("b", MinorType.FLOAT8).build();
    RowSetBuilder builder = fixture.rowSetBuilder(schema);
    for (long i = 0; i < 10; i++) {
        builder.addRow(new long[] { 1, 2, 3, 4, 5 }, new double[] { i * 0.1, i * 0.1, i * 0.1, i * 0.2, i * 0.3 });
    }
    RowSet rows = builder.build();
    // Run the record batch sizer on the resulting batch.
    RecordBatchSizer sizer = new RecordBatchSizer(rows.container());
    assertEquals(2, sizer.columns().size());
    /**
     * stdDataSize:8*10, stdNetSize:8*10+4, dataSizePerEntry:5*8, netSizePerEntry:5*8+4,
     * totalDataSize:5*8*10, totalNetSize:5*8*10+5*8, valueCount:10,
     * elementCount:50, estElementCountPerArray:5, isVariableWidth:false
     */
    verifyColumnValues(sizer.columns().get("a"), 80, 84, 40, 44, 400, 440, 10, 50, 5, false);
    verifyColumnValues(sizer.columns().get("b"), 80, 84, 40, 44, 400, 440, 10, 50, 5, false);
    SingleRowSet empty = fixture.rowSet(schema);
    VectorAccessible accessible = empty.vectorAccessible();
    UInt4Vector offsetVector;
    ValueVector dataVector;
    for (VectorWrapper<?> vw : accessible) {
        ValueVector v = vw.getValueVector();
        RecordBatchSizer.ColumnSize colSize = sizer.getColumn(v.getField().getName());
        // Allocates to nearest power of two
        colSize.allocateVector(v, testRowCount);
        offsetVector = ((RepeatedValueVector) v).getOffsetVector();
        assertEquals((Integer.highestOneBit(testRowCount) << 1), offsetVector.getValueCapacity());
        dataVector = ((RepeatedValueVector) v).getDataVector();
        assertEquals(Integer.highestOneBit((testRowCount * 5) << 1), dataVector.getValueCapacity());
        v.clear();
        // Allocates the same as value passed since it is already power of two.
        // -1 is done for adjustment needed for offset vector.
        colSize.allocateVector(v, testRowCountPowerTwo - 1);
        offsetVector = ((RepeatedValueVector) v).getOffsetVector();
        assertEquals(testRowCountPowerTwo, offsetVector.getValueCapacity());
        dataVector = ((RepeatedValueVector) v).getDataVector();
        assertEquals(Integer.highestOneBit((testRowCountPowerTwo - 1) * 5) << 1, dataVector.getValueCapacity());
        v.clear();
        // Allocate for max rows.
        colSize.allocateVector(v, ValueVector.MAX_ROW_COUNT - 1);
        offsetVector = ((RepeatedValueVector) v).getOffsetVector();
        assertEquals(ValueVector.MAX_ROW_COUNT, offsetVector.getValueCapacity());
        dataVector = ((RepeatedValueVector) v).getDataVector();
        assertEquals(Integer.highestOneBit(((ValueVector.MAX_ROW_COUNT - 1) * 5) << 1), dataVector.getValueCapacity());
        v.clear();
        // Allocate for 0 rows. should atleast do allocation for 1 row.
        colSize.allocateVector(v, 0);
        offsetVector = ((RepeatedValueVector) v).getOffsetVector();
        assertEquals(ValueVector.MIN_ROW_COUNT + 1, offsetVector.getValueCapacity());
        dataVector = ((RepeatedValueVector) v).getDataVector();
        assertEquals(ValueVector.MIN_ROW_COUNT, dataVector.getValueCapacity());
        v.clear();
    }
    empty.clear();
    rows.clear();
}

Example 12 with RowSet

use of org.apache.drill.test.rowSet.RowSet in project drill by axbaretto.

the class TestRecordBatchSizer method testSizerNullableFixedWidth.

@Test
public void testSizerNullableFixedWidth() {
    BatchSchema schema = new SchemaBuilder().addNullable("a", MinorType.BIGINT).addNullable("b", MinorType.FLOAT8).build();
    RowSetBuilder builder = fixture.rowSetBuilder(schema);
    for (long i = 0; i < 10; i++) {
        builder.addRow(i, i * 0.1);
    }
    RowSet rows = builder.build();
    // Run the record batch sizer on the resulting batch.
    RecordBatchSizer sizer = new RecordBatchSizer(rows.container());
    assertEquals(2, sizer.columns().size());
    ColumnSize aColumn = sizer.columns().get("a");
    ColumnSize bColumn = sizer.columns().get("b");
    /**
     * stdDataSize:8, stdNetSize:8+1, dataSizePerEntry:8, netSizePerEntry:8+1,
     * totalDataSize:8*10, totalNetSize:(8+1)*10, valueCount:10,
     * elementCount:10, estElementCountPerArray:1, isVariableWidth:false
     */
    verifyColumnValues(aColumn, 8, 9, 8, 9, 80, 90, 10, 10, 1, false);
    verifyColumnValues(bColumn, 8, 9, 8, 9, 80, 90, 10, 10, 1, false);
    SingleRowSet empty = fixture.rowSet(schema);
    VectorAccessible accessible = empty.vectorAccessible();
    ValueVector bitVector, valueVector;
    for (VectorWrapper<?> vw : accessible) {
        ValueVector v = vw.getValueVector();
        RecordBatchSizer.ColumnSize colSize = sizer.getColumn(v.getField().getName());
        // Allocates to nearest power of two
        colSize.allocateVector(v, testRowCount);
        bitVector = ((NullableVector) v).getBitsVector();
        assertEquals((Integer.highestOneBit(testRowCount) << 1), bitVector.getValueCapacity());
        valueVector = ((NullableVector) v).getValuesVector();
        assertEquals(Integer.highestOneBit(testRowCount << 1), valueVector.getValueCapacity());
        v.clear();
        // Allocates the same as value passed since it is already power of two.
        colSize.allocateVector(v, testRowCountPowerTwo);
        bitVector = ((NullableVector) v).getBitsVector();
        assertEquals(testRowCountPowerTwo, bitVector.getValueCapacity());
        valueVector = ((NullableVector) v).getValuesVector();
        assertEquals(testRowCountPowerTwo, valueVector.getValueCapacity());
        v.clear();
        // Allocate for max rows.
        colSize.allocateVector(v, ValueVector.MAX_ROW_COUNT);
        bitVector = ((NullableVector) v).getBitsVector();
        assertEquals(ValueVector.MAX_ROW_COUNT, bitVector.getValueCapacity());
        valueVector = ((NullableVector) v).getValuesVector();
        assertEquals(ValueVector.MAX_ROW_COUNT, valueVector.getValueCapacity());
        v.clear();
        // Allocate for 0 rows. should atleast do allocation for 1 row.
        colSize.allocateVector(v, 0);
        bitVector = ((NullableVector) v).getBitsVector();
        assertEquals(ValueVector.MIN_ROW_COUNT, bitVector.getValueCapacity());
        valueVector = ((NullableVector) v).getValuesVector();
        assertEquals(ValueVector.MIN_ROW_COUNT, valueVector.getValueCapacity());
        v.clear();
    }
    empty.clear();
    rows.clear();
}

Example 13 with RowSet

use of org.apache.drill.test.rowSet.RowSet in project drill by axbaretto.

the class TestRecordBatchSizer method testSizerMap.

@Test
public void testSizerMap() {
    BatchSchema schema = new SchemaBuilder().addMap("map").add("key", MinorType.INT).add("value", MinorType.VARCHAR).resumeSchema().build();
    RowSetBuilder builder = fixture.rowSetBuilder(schema);
    for (int i = 0; i < 10; i++) {
        builder.addRow((Object) (new Object[] { 10, "a" }));
    }
    RowSet rows = builder.build();
    // Run the record batch sizer on the resulting batch.
    RecordBatchSizer sizer = new RecordBatchSizer(rows.container());
    assertEquals(1, sizer.columns().size());
    /**
     * stdDataSize:50+4, stdNetSize:50+4+4, dataSizePerEntry:4+1,
     * netSizePerEntry: 4+1+4,
     * totalDataSize:5*10, totalNetSize:4*10+4*10+1*10,
     * valueCount:10,
     * elementCount:10, estElementCountPerArray:1, isVariableWidth:true
     */
    verifyColumnValues(sizer.columns().get("map"), 54, 58, 5, 9, 50, 90, 10, 10, 1, false);
    SingleRowSet empty = fixture.rowSet(schema);
    VectorAccessible accessible = empty.vectorAccessible();
    for (VectorWrapper<?> vw : accessible) {
        ValueVector v = vw.getValueVector();
        RecordBatchSizer.ColumnSize colSize = sizer.getColumn(v.getField().getName());
        // Allocates to nearest power of two
        colSize.allocateVector(v, testRowCount);
        MapVector mapVector = (MapVector) v;
        ValueVector keyVector = mapVector.getChild("key");
        ValueVector valueVector1 = mapVector.getChild("value");
        assertEquals((Integer.highestOneBit(testRowCount) << 1), keyVector.getValueCapacity());
        UInt4Vector offsetVector = ((VariableWidthVector) valueVector1).getOffsetVector();
        assertEquals((Integer.highestOneBit(testRowCount) << 1), offsetVector.getValueCapacity());
        assertEquals(Integer.highestOneBit(testRowCount << 1) - 1, valueVector1.getValueCapacity());
        // Allocates the same as value passed since it is already power of two.
        colSize.allocateVector(v, testRowCountPowerTwo - 1);
        mapVector = (MapVector) v;
        keyVector = mapVector.getChild("key");
        valueVector1 = mapVector.getChild("value");
        assertEquals((Integer.highestOneBit(testRowCountPowerTwo - 1) << 1), keyVector.getValueCapacity());
        offsetVector = ((VariableWidthVector) valueVector1).getOffsetVector();
        assertEquals(testRowCountPowerTwo, offsetVector.getValueCapacity());
        assertEquals(Integer.highestOneBit(testRowCountPowerTwo) - 1, valueVector1.getValueCapacity());
        // Allocate for max rows.
        colSize.allocateVector(v, ValueVector.MAX_ROW_COUNT - 1);
        mapVector = (MapVector) v;
        keyVector = mapVector.getChild("key");
        valueVector1 = mapVector.getChild("value");
        assertEquals(ValueVector.MAX_ROW_COUNT, keyVector.getValueCapacity());
        offsetVector = ((VariableWidthVector) valueVector1).getOffsetVector();
        assertEquals(ValueVector.MAX_ROW_COUNT, offsetVector.getValueCapacity());
        assertEquals(ValueVector.MAX_ROW_COUNT - 1, valueVector1.getValueCapacity());
        // Allocate for 0 rows. should atleast do allocation for 1 row.
        colSize.allocateVector(v, 0);
        mapVector = (MapVector) v;
        keyVector = mapVector.getChild("key");
        valueVector1 = mapVector.getChild("value");
        assertEquals(ValueVector.MIN_ROW_COUNT, keyVector.getValueCapacity());
        offsetVector = ((VariableWidthVector) valueVector1).getOffsetVector();
        assertEquals(ValueVector.MIN_ROW_COUNT + 1, offsetVector.getValueCapacity());
        assertEquals(ValueVector.MIN_ROW_COUNT, valueVector1.getValueCapacity());
        v.clear();
    }
    empty.clear();
    rows.clear();
}

Example 14 with RowSet

use of org.apache.drill.test.rowSet.RowSet in project drill by axbaretto.

the class TestRecordBatchSizer method testEmptyVariableWidthVector.

/**
 * Test to verify that record batch sizer handles the actual empty vectors correctly. RowSetBuilder by default
 * allocates Drillbuf of 10bytes for each vector type which makes their capacity >0 and not ==0 which will be in
 * case of empty vectors.
 */
@Test
public void testEmptyVariableWidthVector() {
    final BatchSchema schema = new SchemaBuilder().add("key", MinorType.INT).add("value", MinorType.VARCHAR).build();
    final RowSetBuilder builder = fixture.rowSetBuilder(schema);
    final RowSet rows = builder.build();
    // Release the initial bytes allocated by RowSetBuilder
    VectorAccessibleUtilities.clear(rows.container());
    try {
        // Create RecordBatchSizer for this empty container and it should not fail
        final RecordBatchSizer sizer = new RecordBatchSizer(rows.container());
        int totalDataSize = 0;
        for (ColumnSize colSize : sizer.columns().values()) {
            totalDataSize += colSize.getTotalDataSize();
        }
        // Verify that the totalDataSize for all the columns is zero
        assertEquals(0, totalDataSize);
    } catch (Exception ex) {
        fail();
    } finally {
        rows.clear();
    }
}

Example 15 with RowSet

use of org.apache.drill.test.rowSet.RowSet in project drill by axbaretto.

the class TestResultSetLoaderMaps method testEmptyMapAddition.

/**
 * Test adding an empty map to a loader after writing the first row.
 * Then add columns in another batch. Yes, this is a bizarre condition,
 * but we must check it anyway for robustness.
 */
@Test
public void testEmptyMapAddition() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).buildSchema();
    ResultSetLoaderImpl.ResultSetOptions options = new OptionBuilder().setSchema(schema).build();
    ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    assertEquals(1, rsLoader.schemaVersion());
    RowSetLoader rootWriter = rsLoader.writer();
    // Start without the map. Add a map after the first row.
    rsLoader.startBatch();
    rootWriter.addRow(10);
    int mapIndex = rootWriter.addColumn(SchemaBuilder.columnSchema("m", MinorType.MAP, DataMode.REQUIRED));
    TupleWriter mapWriter = rootWriter.tuple(mapIndex);
    rootWriter.addRow(20, objArray()).addRow(30, objArray());
    RowSet actual = fixture.wrap(rsLoader.harvest());
    assertEquals(2, rsLoader.schemaVersion());
    assertEquals(3, actual.rowCount());
    // Validate first batch
    TupleMetadata expectedSchema = new SchemaBuilder().add("a", MinorType.INT).addMap("m").resumeSchema().buildSchema();
    SingleRowSet expected = fixture.rowSetBuilder(expectedSchema).addRow(10, objArray()).addRow(20, objArray()).addRow(30, objArray()).build();
    new RowSetComparison(expected).verifyAndClearAll(actual);
    // Now add another column to the map
    rsLoader.startBatch();
    mapWriter.addColumn(SchemaBuilder.columnSchema("a", MinorType.VARCHAR, DataMode.REQUIRED));
    rootWriter.addRow(40, objArray("fred")).addRow(50, objArray("barney"));
    actual = fixture.wrap(rsLoader.harvest());
    assertEquals(3, rsLoader.schemaVersion());
    assertEquals(2, actual.rowCount());
    // Validate first batch
    expectedSchema = new SchemaBuilder().add("a", MinorType.INT).addMap("m").add("a", MinorType.VARCHAR).resumeSchema().buildSchema();
    expected = fixture.rowSetBuilder(expectedSchema).addRow(40, objArray("fred")).addRow(50, objArray("barney")).build();
    new RowSetComparison(expected).verifyAndClearAll(actual);
    rsLoader.close();
}