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

Example 21 with SchemaBuilder

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

the class TestResultSetLoaderMapArray method testNestedArray.

@Test
public void testNestedArray() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("c", MinorType.INT).addArray("d", MinorType.VARCHAR).resumeSchema().buildSchema();
    ResultSetLoaderImpl.ResultSetOptions options = new OptionBuilder().setSchema(schema).build();
    ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    RowSetLoader rootWriter = rsLoader.writer();
    // Write a couple of rows with arrays within arrays.
    // (And, of course, the Varchar is actually an array of
    // bytes, so that's three array levels.)
    rsLoader.startBatch();
    rootWriter.addRow(10, objArray(objArray(110, strArray("d1.1.1", "d1.1.2")), objArray(120, strArray("d1.2.1", "d1.2.2")))).addRow(20, objArray()).addRow(30, objArray(objArray(310, strArray("d3.1.1", "d3.2.2")), objArray(320, strArray()), objArray(330, strArray("d3.3.1", "d1.2.2"))));
    // Verify the batch
    RowSet actual = fixture.wrap(rsLoader.harvest());
    SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(10, objArray(objArray(110, strArray("d1.1.1", "d1.1.2")), objArray(120, strArray("d1.2.1", "d1.2.2")))).addRow(20, objArray()).addRow(30, objArray(objArray(310, strArray("d3.1.1", "d3.2.2")), objArray(320, strArray()), objArray(330, strArray("d3.3.1", "d1.2.2")))).build();
    new RowSetComparison(expected).verifyAndClearAll(actual);
    rsLoader.close();
}

Example 22 with SchemaBuilder

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

the class TestResultSetLoaderMapArray method testOverwriteRow.

/**
 * Version of the {#link TestResultSetLoaderProtocol#testOverwriteRow()} test
 * that uses nested columns inside an array of maps. Here we must call
 * <tt>start()</tt> to reset the array back to the initial start position after
 * each "discard."
 */
@Test
public void testOverwriteRow() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("b", MinorType.INT).add("c", MinorType.VARCHAR).resumeSchema().buildSchema();
    ResultSetLoaderImpl.ResultSetOptions options = new OptionBuilder().setSchema(schema).setRowCountLimit(ValueVector.MAX_ROW_COUNT).build();
    ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    RowSetLoader rootWriter = rsLoader.writer();
    // Can't use the shortcut to populate rows when doing overwrites.
    ScalarWriter aWriter = rootWriter.scalar("a");
    ArrayWriter maWriter = rootWriter.array("m");
    TupleWriter mWriter = maWriter.tuple();
    ScalarWriter bWriter = mWriter.scalar("b");
    ScalarWriter cWriter = mWriter.scalar("c");
    // Write 100,000 rows, overwriting 99% of them. This will cause vector
    // overflow and data corruption if overwrite does not work; but will happily
    // produce the correct result if everything works as it should.
    byte[] value = new byte[512];
    Arrays.fill(value, (byte) 'X');
    int count = 0;
    rsLoader.startBatch();
    while (count < 10_000) {
        rootWriter.start();
        count++;
        aWriter.setInt(count);
        for (int i = 0; i < 10; i++) {
            bWriter.setInt(count * 10 + i);
            cWriter.setBytes(value, value.length);
            maWriter.save();
        }
        if (count % 100 == 0) {
            rootWriter.save();
        }
    }
    // Verify using a reader.
    RowSet result = fixture.wrap(rsLoader.harvest());
    assertEquals(count / 100, result.rowCount());
    RowSetReader reader = result.reader();
    ArrayReader maReader = reader.array("m");
    TupleReader mReader = maReader.tuple();
    int rowId = 1;
    while (reader.next()) {
        assertEquals(rowId * 100, reader.scalar("a").getInt());
        assertEquals(10, maReader.size());
        for (int i = 0; i < 10; i++) {
            maReader.setPosn(i);
            assertEquals(rowId * 1000 + i, mReader.scalar("b").getInt());
            assertTrue(Arrays.equals(value, mReader.scalar("c").getBytes()));
        }
        rowId++;
    }
    result.clear();
    rsLoader.close();
}

Example 23 with SchemaBuilder

use of org.apache.drill.test.rowSet.schema.SchemaBuilder 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 24 with SchemaBuilder

use of org.apache.drill.test.rowSet.schema.SchemaBuilder 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 25 with SchemaBuilder

use of org.apache.drill.test.rowSet.schema.SchemaBuilder 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();
}