Search in sources :

Example 1 with TupleWriter

use of org.apache.drill.exec.vector.accessor.TupleWriter in project drill by axbaretto.

the class TestResultSetLoaderMapArray method testBasics.

@Test
public void testBasics() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("c", MinorType.INT).add("d", MinorType.VARCHAR).resumeSchema().buildSchema();
    ResultSetLoaderImpl.ResultSetOptions options = new OptionBuilder().setSchema(schema).build();
    ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    RowSetLoader rootWriter = rsLoader.writer();
    // Verify structure and schema
    TupleMetadata actualSchema = rootWriter.schema();
    assertEquals(2, actualSchema.size());
    assertTrue(actualSchema.metadata(1).isArray());
    assertTrue(actualSchema.metadata(1).isMap());
    assertEquals(2, actualSchema.metadata("m").mapSchema().size());
    assertEquals(2, actualSchema.column("m").getChildren().size());
    // Write a couple of rows with arrays.
    rsLoader.startBatch();
    rootWriter.addRow(10, objArray(objArray(110, "d1.1"), objArray(120, "d2.2"))).addRow(20, objArray()).addRow(30, objArray(objArray(310, "d3.1"), objArray(320, "d3.2"), objArray(330, "d3.3")));
    // Verify the first batch
    RowSet actual = fixture.wrap(rsLoader.harvest());
    SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(10, objArray(objArray(110, "d1.1"), objArray(120, "d2.2"))).addRow(20, objArray()).addRow(30, objArray(objArray(310, "d3.1"), objArray(320, "d3.2"), objArray(330, "d3.3"))).build();
    new RowSetComparison(expected).verifyAndClearAll(actual);
    // In the second, create a row, then add a map member.
    // Should be back-filled to empty for the first row.
    rsLoader.startBatch();
    rootWriter.addRow(40, objArray(objArray(410, "d4.1"), objArray(420, "d4.2")));
    TupleWriter mapWriter = rootWriter.array("m").tuple();
    mapWriter.addColumn(SchemaBuilder.columnSchema("e", MinorType.VARCHAR, DataMode.OPTIONAL));
    rootWriter.addRow(50, objArray(objArray(510, "d5.1", "e5.1"), objArray(520, "d5.2", null))).addRow(60, objArray(objArray(610, "d6.1", "e6.1"), objArray(620, "d6.2", null), objArray(630, "d6.3", "e6.3")));
    // Verify the second batch
    actual = fixture.wrap(rsLoader.harvest());
    TupleMetadata expectedSchema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m").add("c", MinorType.INT).add("d", MinorType.VARCHAR).addNullable("e", MinorType.VARCHAR).resumeSchema().buildSchema();
    expected = fixture.rowSetBuilder(expectedSchema).addRow(40, objArray(objArray(410, "d4.1", null), objArray(420, "d4.2", null))).addRow(50, objArray(objArray(510, "d5.1", "e5.1"), objArray(520, "d5.2", null))).addRow(60, objArray(objArray(610, "d6.1", "e6.1"), objArray(620, "d6.2", null), objArray(630, "d6.3", "e6.3"))).build();
    new RowSetComparison(expected).verifyAndClearAll(actual);
    rsLoader.close();
}
Also used : SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSetComparison(org.apache.drill.test.rowSet.RowSetComparison) ResultSetLoader(org.apache.drill.exec.physical.rowSet.ResultSetLoader) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.test.rowSet.schema.SchemaBuilder) SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.test.rowSet.RowSet) RowSetLoader(org.apache.drill.exec.physical.rowSet.RowSetLoader) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 2 with TupleWriter

use of org.apache.drill.exec.vector.accessor.TupleWriter 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();
}
Also used : TupleReader(org.apache.drill.exec.vector.accessor.TupleReader) SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.test.rowSet.RowSet) ArrayReader(org.apache.drill.exec.vector.accessor.ArrayReader) ResultSetLoader(org.apache.drill.exec.physical.rowSet.ResultSetLoader) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.test.rowSet.schema.SchemaBuilder) RowSetLoader(org.apache.drill.exec.physical.rowSet.RowSetLoader) ArrayWriter(org.apache.drill.exec.vector.accessor.ArrayWriter) RowSetReader(org.apache.drill.test.rowSet.RowSetReader) ScalarWriter(org.apache.drill.exec.vector.accessor.ScalarWriter) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 3 with TupleWriter

use of org.apache.drill.exec.vector.accessor.TupleWriter 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();
}
Also used : SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSetComparison(org.apache.drill.test.rowSet.RowSetComparison) ResultSetLoader(org.apache.drill.exec.physical.rowSet.ResultSetLoader) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.test.rowSet.schema.SchemaBuilder) SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.test.rowSet.RowSet) RowSetLoader(org.apache.drill.exec.physical.rowSet.RowSetLoader) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 4 with TupleWriter

use of org.apache.drill.exec.vector.accessor.TupleWriter in project drill by axbaretto.

the class TestResultSetLoaderMaps method testOverwriteRow.

/**
 * Version of the {#link TestResultSetLoaderProtocol#testOverwriteRow()} test
 * that uses nested columns.
 */
@Test
public void testOverwriteRow() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMap("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");
    TupleWriter mWriter = rootWriter.tuple("m");
    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 < 100_000) {
        rootWriter.start();
        count++;
        aWriter.setInt(count);
        bWriter.setInt(count * 10);
        cWriter.setBytes(value, value.length);
        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();
    TupleReader mReader = reader.tuple("m");
    int rowId = 1;
    while (reader.next()) {
        assertEquals(rowId * 100, reader.scalar("a").getInt());
        assertEquals(rowId * 1000, mReader.scalar("b").getInt());
        assertTrue(Arrays.equals(value, mReader.scalar("c").getBytes()));
        rowId++;
    }
    result.clear();
    rsLoader.close();
}
Also used : TupleReader(org.apache.drill.exec.vector.accessor.TupleReader) SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.test.rowSet.RowSet) ResultSetLoader(org.apache.drill.exec.physical.rowSet.ResultSetLoader) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.test.rowSet.schema.SchemaBuilder) RowSetLoader(org.apache.drill.exec.physical.rowSet.RowSetLoader) RowSetReader(org.apache.drill.test.rowSet.RowSetReader) ScalarWriter(org.apache.drill.exec.vector.accessor.ScalarWriter) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 5 with TupleWriter

use of org.apache.drill.exec.vector.accessor.TupleWriter in project drill by axbaretto.

the class RowSetTest method testMapStructure.

/**
 * Test a simple map structure at the top level of a row.
 *
 * @throws VectorOverflowException should never occur
 */
@Test
public void testMapStructure() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMap("m").addArray("b", MinorType.INT).resumeSchema().buildSchema();
    ExtendableRowSet rowSet = fixture.rowSet(schema);
    RowSetWriter writer = rowSet.writer();
    // Map and Int
    // Test Invariants
    assertEquals(ObjectType.SCALAR, writer.column("a").type());
    assertEquals(ObjectType.SCALAR, writer.column(0).type());
    assertEquals(ObjectType.TUPLE, writer.column("m").type());
    assertEquals(ObjectType.TUPLE, writer.column(1).type());
    assertSame(writer.column(1).tuple(), writer.tuple(1));
    TupleWriter mapWriter = writer.column(1).tuple();
    assertEquals(ObjectType.SCALAR, mapWriter.column("b").array().entry().type());
    assertEquals(ObjectType.SCALAR, mapWriter.column("b").array().entryType());
    ScalarWriter aWriter = writer.column("a").scalar();
    ScalarWriter bWriter = writer.column("m").tuple().column("b").array().entry().scalar();
    assertSame(bWriter, writer.tuple(1).array(0).scalar());
    assertEquals(ValueType.INTEGER, bWriter.valueType());
    try {
        writer.column(1).scalar();
        fail();
    } catch (UnsupportedOperationException e) {
    // Expected
    }
    try {
        writer.column(1).array();
        fail();
    } catch (UnsupportedOperationException e) {
    // Expected
    }
    // Write data
    aWriter.setInt(10);
    bWriter.setInt(11);
    bWriter.setInt(12);
    writer.save();
    aWriter.setInt(20);
    bWriter.setInt(21);
    bWriter.setInt(22);
    writer.save();
    aWriter.setInt(30);
    bWriter.setInt(31);
    bWriter.setInt(32);
    writer.save();
    // Finish the row set and get a reader.
    SingleRowSet actual = writer.done();
    RowSetReader reader = actual.reader();
    assertEquals(ObjectType.SCALAR, reader.column("a").type());
    assertEquals(ObjectType.SCALAR, reader.column(0).type());
    assertEquals(ObjectType.TUPLE, reader.column("m").type());
    assertEquals(ObjectType.TUPLE, reader.column(1).type());
    assertSame(reader.column(1).tuple(), reader.tuple(1));
    ScalarReader aReader = reader.column(0).scalar();
    TupleReader mReader = reader.column(1).tuple();
    assertEquals(ObjectType.SCALAR, mReader.column("b").array().entryType());
    ScalarElementReader bReader = mReader.column(0).elements();
    assertEquals(ValueType.INTEGER, bReader.valueType());
    assertTrue(reader.next());
    assertEquals(10, aReader.getInt());
    assertEquals(11, bReader.getInt(0));
    assertEquals(12, bReader.getInt(1));
    assertTrue(reader.next());
    assertEquals(20, aReader.getInt());
    assertEquals(21, bReader.getInt(0));
    assertEquals(22, bReader.getInt(1));
    assertTrue(reader.next());
    assertEquals(30, aReader.getInt());
    assertEquals(31, bReader.getInt(0));
    assertEquals(32, bReader.getInt(1));
    assertFalse(reader.next());
    // Verify that the map accessor's value count was set.
    @SuppressWarnings("resource") MapVector mapVector = (MapVector) actual.container().getValueVector(1).getValueVector();
    assertEquals(actual.rowCount(), mapVector.getAccessor().getValueCount());
    SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(10, objArray(intArray(11, 12))).addRow(20, objArray(intArray(21, 22))).addRow(30, objArray(intArray(31, 32))).build();
    new RowSetComparison(expected).verifyAndClearAll(actual);
}
Also used : RowSetWriter(org.apache.drill.test.rowSet.RowSetWriter) TupleReader(org.apache.drill.exec.vector.accessor.TupleReader) SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) ScalarElementReader(org.apache.drill.exec.vector.accessor.ScalarElementReader) ScalarReader(org.apache.drill.exec.vector.accessor.ScalarReader) RowSetComparison(org.apache.drill.test.rowSet.RowSetComparison) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.test.rowSet.schema.SchemaBuilder) RowSetReader(org.apache.drill.test.rowSet.RowSetReader) ScalarWriter(org.apache.drill.exec.vector.accessor.ScalarWriter) ExtendableRowSet(org.apache.drill.test.rowSet.RowSet.ExtendableRowSet) RepeatedMapVector(org.apache.drill.exec.vector.complex.RepeatedMapVector) MapVector(org.apache.drill.exec.vector.complex.MapVector) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Aggregations

TupleWriter (org.apache.drill.exec.vector.accessor.TupleWriter)59 TupleMetadata (org.apache.drill.exec.record.metadata.TupleMetadata)52 SubOperatorTest (org.apache.drill.test.SubOperatorTest)50 Test (org.junit.Test)50 SchemaBuilder (org.apache.drill.exec.record.metadata.SchemaBuilder)35 SingleRowSet (org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet)33 RowSetLoader (org.apache.drill.exec.physical.resultSet.RowSetLoader)26 ResultSetLoader (org.apache.drill.exec.physical.resultSet.ResultSetLoader)25 ScalarWriter (org.apache.drill.exec.vector.accessor.ScalarWriter)23 TupleReader (org.apache.drill.exec.vector.accessor.TupleReader)20 RowSet (org.apache.drill.exec.physical.rowSet.RowSet)19 ArrayWriter (org.apache.drill.exec.vector.accessor.ArrayWriter)18 SchemaBuilder (org.apache.drill.test.rowSet.schema.SchemaBuilder)16 SingleRowSet (org.apache.drill.test.rowSet.RowSet.SingleRowSet)15 ResultSetLoader (org.apache.drill.exec.physical.rowSet.ResultSetLoader)14 RowSetLoader (org.apache.drill.exec.physical.rowSet.RowSetLoader)14 RowSet (org.apache.drill.test.rowSet.RowSet)13 ArrayReader (org.apache.drill.exec.vector.accessor.ArrayReader)12 RowSetComparison (org.apache.drill.test.rowSet.RowSetComparison)12 ScalarReader (org.apache.drill.exec.vector.accessor.ScalarReader)10