Search in sources :

Example 16 with TupleReader

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

the class TestResultSetLoaderMaps method testMapOverflowWithNewColumn.

/**
 * Test the case in which a new column is added during the overflow row. Unlike
 * the top-level schema case, internally we must create a copy of the map, and
 * move vectors across only when the result is to include the schema version
 * of the target column. For overflow, the new column is added after the
 * first batch; it is added in the second batch that contains the overflow
 * row in which the column was added.
 */
@Test
public void testMapOverflowWithNewColumn() {
    final TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMap("m").add("b", MinorType.INT).add("c", MinorType.VARCHAR).resumeSchema().buildSchema();
    final ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).rowCountLimit(ValueVector.MAX_ROW_COUNT).build();
    final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    assertEquals(4, rsLoader.schemaVersion());
    final RowSetLoader rootWriter = rsLoader.writer();
    // Can't use the shortcut to populate rows when doing a schema
    // change.
    final ScalarWriter aWriter = rootWriter.scalar("a");
    final TupleWriter mWriter = rootWriter.tuple("m");
    final ScalarWriter bWriter = mWriter.scalar("b");
    final ScalarWriter cWriter = mWriter.scalar("c");
    final byte[] value = new byte[512];
    Arrays.fill(value, (byte) 'X');
    int count = 0;
    rsLoader.startBatch();
    while (!rootWriter.isFull()) {
        rootWriter.start();
        aWriter.setInt(count);
        bWriter.setInt(count * 10);
        cWriter.setBytes(value, value.length);
        if (rootWriter.isFull()) {
            // Overflow just occurred. Add another column.
            mWriter.addColumn(SchemaBuilder.columnSchema("d", MinorType.INT, DataMode.OPTIONAL));
            mWriter.scalar("d").setInt(count * 100);
        }
        rootWriter.save();
        count++;
    }
    // Result set should include the original columns, but not d.
    RowSet result = fixture.wrap(rsLoader.harvest());
    assertEquals(4, rsLoader.schemaVersion());
    assertTrue(schema.isEquivalent(result.schema()));
    final BatchSchema expectedSchema = new BatchSchema(SelectionVectorMode.NONE, schema.toFieldList());
    assertTrue(expectedSchema.isEquivalent(result.batchSchema()));
    // Use a reader to validate row-by-row. Too large to create an expected
    // result set.
    RowSetReader reader = result.reader();
    TupleReader mapReader = reader.tuple("m");
    int rowId = 0;
    while (reader.next()) {
        assertEquals(rowId, reader.scalar("a").getInt());
        assertEquals(rowId * 10, mapReader.scalar("b").getInt());
        assertTrue(Arrays.equals(value, mapReader.scalar("c").getBytes()));
        rowId++;
    }
    result.clear();
    // Next batch should start with the overflow row
    rsLoader.startBatch();
    assertEquals(1, rootWriter.rowCount());
    result = fixture.wrap(rsLoader.harvest());
    assertEquals(1, result.rowCount());
    reader = result.reader();
    mapReader = reader.tuple("m");
    while (reader.next()) {
        assertEquals(rowId, reader.scalar("a").getInt());
        assertEquals(rowId * 10, mapReader.scalar("b").getInt());
        assertTrue(Arrays.equals(value, mapReader.scalar("c").getBytes()));
        assertEquals(rowId * 100, mapReader.scalar("d").getInt());
    }
    result.clear();
    rsLoader.close();
}
Also used : TupleReader(org.apache.drill.exec.vector.accessor.TupleReader) SingleRowSet(org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.exec.physical.rowSet.RowSet) ResultSetLoader(org.apache.drill.exec.physical.resultSet.ResultSetLoader) BatchSchema(org.apache.drill.exec.record.BatchSchema) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) RowSetLoader(org.apache.drill.exec.physical.resultSet.RowSetLoader) RowSetReader(org.apache.drill.exec.physical.rowSet.RowSetReader) ScalarWriter(org.apache.drill.exec.vector.accessor.ScalarWriter) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 17 with TupleReader

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

the class TestResultSetLoaderMaps method testMapOverflowWithNewColumn.

/**
 * Test the case in which a new column is added during the overflow row. Unlike
 * the top-level schema case, internally we must create a copy of the map, and
 * move vectors across only when the result is to include the schema version
 * of the target column. For overflow, the new column is added after the
 * first batch; it is added in the second batch that contains the overflow
 * row in which the column was added.
 */
@Test
public void testMapOverflowWithNewColumn() {
    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);
    assertEquals(4, rsLoader.schemaVersion());
    RowSetLoader rootWriter = rsLoader.writer();
    // Can't use the shortcut to populate rows when doing a schema
    // change.
    ScalarWriter aWriter = rootWriter.scalar("a");
    TupleWriter mWriter = rootWriter.tuple("m");
    ScalarWriter bWriter = mWriter.scalar("b");
    ScalarWriter cWriter = mWriter.scalar("c");
    byte[] value = new byte[512];
    Arrays.fill(value, (byte) 'X');
    int count = 0;
    rsLoader.startBatch();
    while (!rootWriter.isFull()) {
        rootWriter.start();
        aWriter.setInt(count);
        bWriter.setInt(count * 10);
        cWriter.setBytes(value, value.length);
        if (rootWriter.isFull()) {
            // Overflow just occurred. Add another column.
            mWriter.addColumn(SchemaBuilder.columnSchema("d", MinorType.INT, DataMode.OPTIONAL));
            mWriter.scalar("d").setInt(count * 100);
        }
        rootWriter.save();
        count++;
    }
    // Result set should include the original columns, but not d.
    RowSet result = fixture.wrap(rsLoader.harvest());
    assertEquals(4, rsLoader.schemaVersion());
    assertTrue(schema.isEquivalent(result.schema()));
    BatchSchema expectedSchema = new BatchSchema(SelectionVectorMode.NONE, schema.toFieldList());
    assertTrue(expectedSchema.isEquivalent(result.batchSchema()));
    // Use a reader to validate row-by-row. Too large to create an expected
    // result set.
    RowSetReader reader = result.reader();
    TupleReader mapReader = reader.tuple("m");
    int rowId = 0;
    while (reader.next()) {
        assertEquals(rowId, reader.scalar("a").getInt());
        assertEquals(rowId * 10, mapReader.scalar("b").getInt());
        assertTrue(Arrays.equals(value, mapReader.scalar("c").getBytes()));
        rowId++;
    }
    result.clear();
    // Next batch should start with the overflow row
    rsLoader.startBatch();
    assertEquals(1, rootWriter.rowCount());
    result = fixture.wrap(rsLoader.harvest());
    assertEquals(1, result.rowCount());
    reader = result.reader();
    mapReader = reader.tuple("m");
    while (reader.next()) {
        assertEquals(rowId, reader.scalar("a").getInt());
        assertEquals(rowId * 10, mapReader.scalar("b").getInt());
        assertTrue(Arrays.equals(value, mapReader.scalar("c").getBytes()));
        assertEquals(rowId * 100, mapReader.scalar("d").getInt());
    }
    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) BatchSchema(org.apache.drill.exec.record.BatchSchema) 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 18 with TupleReader

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

the class TestResultSetLoaderMapArray method testOmittedValues.

/**
 * Check that the "fill-empties" logic descends down into
 * a repeated map.
 */
@Test
public void testOmittedValues() {
    TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addMapArray("m").addNullable("a", MinorType.INT).addNullable("b", 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();
    int mapSkip = 5;
    int entrySkip = 3;
    int rowCount = 1000;
    int entryCount = 10;
    rsLoader.startBatch();
    ArrayWriter maWriter = rootWriter.array("m");
    TupleWriter mWriter = maWriter.tuple();
    for (int i = 0; i < rowCount; i++) {
        rootWriter.start();
        rootWriter.scalar(0).setInt(i);
        if (i % mapSkip != 0) {
            for (int j = 0; j < entryCount; j++) {
                if (j % entrySkip != 0) {
                    mWriter.scalar(0).setInt(i * entryCount + j);
                    mWriter.scalar(1).setString("b-" + i + "." + j);
                }
                maWriter.save();
            }
        }
        rootWriter.save();
    }
    RowSet result = fixture.wrap(rsLoader.harvest());
    assertEquals(rowCount, result.rowCount());
    RowSetReader reader = result.reader();
    ArrayReader maReader = reader.array("m");
    TupleReader mReader = maReader.tuple();
    for (int i = 0; i < rowCount; i++) {
        assertTrue(reader.next());
        assertEquals(i, reader.scalar(0).getInt());
        if (i % mapSkip == 0) {
            assertEquals(0, maReader.size());
            continue;
        }
        assertEquals(entryCount, maReader.size());
        for (int j = 0; j < entryCount; j++) {
            maReader.setPosn(j);
            if (j % entrySkip == 0) {
                assertTrue(mReader.scalar(0).isNull());
                assertTrue(mReader.scalar(1).isNull());
            } else {
                assertFalse(mReader.scalar(0).isNull());
                assertFalse(mReader.scalar(1).isNull());
                assertEquals(i * entryCount + j, mReader.scalar(0).getInt());
                assertEquals("b-" + i + "." + j, mReader.scalar(1).getString());
            }
        }
    }
    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) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 19 with TupleReader

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

the class TestResultSetLoaderMapArray method testDoubleNestedArray.

/**
 * Test a doubly-nested arrays of maps.
 */
@Test
public void testDoubleNestedArray() {
    TupleMetadata schema = new SchemaBuilder().add("a", MinorType.INT).addMapArray("m1").add("b", MinorType.INT).addMapArray("m2").add("c", MinorType.INT).addArray("d", MinorType.VARCHAR).resumeMap().resumeSchema().buildSchema();
    ResultSetLoaderImpl.ResultSetOptions options = new OptionBuilder().setSchema(schema).build();
    ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    RowSetLoader rootWriter = rsLoader.writer();
    rsLoader.startBatch();
    ScalarWriter aWriter = rootWriter.scalar("a");
    ArrayWriter a1Writer = rootWriter.array("m1");
    TupleWriter m1Writer = a1Writer.tuple();
    ScalarWriter bWriter = m1Writer.scalar("b");
    ArrayWriter a2Writer = m1Writer.array("m2");
    TupleWriter m2Writer = a2Writer.tuple();
    ScalarWriter cWriter = m2Writer.scalar("c");
    ScalarWriter dWriter = m2Writer.array("d").scalar();
    for (int i = 0; i < 5; i++) {
        rootWriter.start();
        aWriter.setInt(i);
        for (int j = 0; j < 4; j++) {
            int a1Key = i + 10 + j;
            bWriter.setInt(a1Key);
            for (int k = 0; k < 3; k++) {
                int a2Key = a1Key * 10 + k;
                cWriter.setInt(a2Key);
                for (int l = 0; l < 2; l++) {
                    dWriter.setString("d-" + (a2Key * 10 + l));
                }
                a2Writer.save();
            }
            a1Writer.save();
        }
        rootWriter.save();
    }
    RowSet results = fixture.wrap(rsLoader.harvest());
    RowSetReader reader = results.reader();
    ScalarReader aReader = reader.scalar("a");
    ArrayReader a1Reader = reader.array("m1");
    TupleReader m1Reader = a1Reader.tuple();
    ScalarReader bReader = m1Reader.scalar("b");
    ArrayReader a2Reader = m1Reader.array("m2");
    TupleReader m2Reader = a2Reader.tuple();
    ScalarReader cReader = m2Reader.scalar("c");
    ScalarElementReader dReader = m2Reader.elements("d");
    for (int i = 0; i < 5; i++) {
        reader.next();
        assertEquals(i, aReader.getInt());
        for (int j = 0; j < 4; j++) {
            a1Reader.setPosn(j);
            int a1Key = i + 10 + j;
            assertEquals(a1Key, bReader.getInt());
            for (int k = 0; k < 3; k++) {
                a2Reader.setPosn(k);
                int a2Key = a1Key * 10 + k;
                assertEquals(a2Key, cReader.getInt());
                for (int l = 0; l < 2; l++) {
                    assertEquals("d-" + (a2Key * 10 + l), dReader.getString(l));
                }
            }
        }
    }
    rsLoader.close();
}
Also used : TupleReader(org.apache.drill.exec.vector.accessor.TupleReader) ScalarElementReader(org.apache.drill.exec.vector.accessor.ScalarElementReader) SingleRowSet(org.apache.drill.test.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.test.rowSet.RowSet) ScalarReader(org.apache.drill.exec.vector.accessor.ScalarReader) 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 20 with TupleReader

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

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
 * {@code start()} 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 ResultSetOptionBuilder().readerSchema(schema).rowCountLimit(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++) {
            assert (maReader.next());
            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.exec.physical.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.exec.physical.rowSet.RowSet) ArrayReader(org.apache.drill.exec.vector.accessor.ArrayReader) ResultSetLoader(org.apache.drill.exec.physical.resultSet.ResultSetLoader) TupleWriter(org.apache.drill.exec.vector.accessor.TupleWriter) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) RowSetLoader(org.apache.drill.exec.physical.resultSet.RowSetLoader) ArrayWriter(org.apache.drill.exec.vector.accessor.ArrayWriter) RowSetReader(org.apache.drill.exec.physical.rowSet.RowSetReader) ScalarWriter(org.apache.drill.exec.vector.accessor.ScalarWriter) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Aggregations

TupleMetadata (org.apache.drill.exec.record.metadata.TupleMetadata)26 TupleReader (org.apache.drill.exec.vector.accessor.TupleReader)26 SubOperatorTest (org.apache.drill.test.SubOperatorTest)21 Test (org.junit.Test)21 TupleWriter (org.apache.drill.exec.vector.accessor.TupleWriter)20 ArrayReader (org.apache.drill.exec.vector.accessor.ArrayReader)17 ScalarWriter (org.apache.drill.exec.vector.accessor.ScalarWriter)17 SingleRowSet (org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet)14 SchemaBuilder (org.apache.drill.exec.record.metadata.SchemaBuilder)14 ArrayWriter (org.apache.drill.exec.vector.accessor.ArrayWriter)11 RowSetReader (org.apache.drill.exec.physical.rowSet.RowSetReader)10 BatchSchema (org.apache.drill.exec.record.BatchSchema)7 ObjectReader (org.apache.drill.exec.vector.accessor.ObjectReader)7 ScalarReader (org.apache.drill.exec.vector.accessor.ScalarReader)6 IOException (java.io.IOException)5 ArrayList (java.util.ArrayList)5 Collections (java.util.Collections)5 HashMap (java.util.HashMap)5 HashSet (java.util.HashSet)5 List (java.util.List)5