Search in sources :

Example 91 with TupleMetadata

use of org.apache.drill.exec.record.metadata.TupleMetadata in project drill by apache.

the class ParquetTableMetadataUtils method getRowGroupMetadata.

/**
 * Returns {@link RowGroupMetadata} instance converted from specified parquet {@code rowGroupMetadata}.
 *
 * @param tableMetadata    table metadata which contains row group metadata to convert
 * @param rowGroupMetadata row group metadata to convert
 * @param rgIndexInFile    index of current row group within the file
 * @param location         location of file with current row group
 * @return {@link RowGroupMetadata} instance converted from specified parquet {@code rowGroupMetadata}
 */
public static RowGroupMetadata getRowGroupMetadata(MetadataBase.ParquetTableMetadataBase tableMetadata, MetadataBase.RowGroupMetadata rowGroupMetadata, int rgIndexInFile, Path location) {
    Map<SchemaPath, ColumnStatistics<?>> columnsStatistics = getRowGroupColumnStatistics(tableMetadata, rowGroupMetadata);
    List<StatisticsHolder<?>> rowGroupStatistics = new ArrayList<>();
    rowGroupStatistics.add(new StatisticsHolder<>(rowGroupMetadata.getRowCount(), TableStatisticsKind.ROW_COUNT));
    rowGroupStatistics.add(new StatisticsHolder<>(rowGroupMetadata.getStart(), new BaseStatisticsKind<>(ExactStatisticsConstants.START, true)));
    rowGroupStatistics.add(new StatisticsHolder<>(rowGroupMetadata.getLength(), new BaseStatisticsKind<>(ExactStatisticsConstants.LENGTH, true)));
    Map<SchemaPath, TypeProtos.MajorType> columns = getRowGroupFields(tableMetadata, rowGroupMetadata);
    Map<SchemaPath, TypeProtos.MajorType> intermediateColumns = getIntermediateFields(tableMetadata, rowGroupMetadata);
    TupleMetadata schema = new TupleSchema();
    columns.forEach((schemaPath, majorType) -> SchemaPathUtils.addColumnMetadata(schema, schemaPath, majorType, intermediateColumns));
    MetadataInfo metadataInfo = MetadataInfo.builder().type(MetadataType.ROW_GROUP).build();
    return RowGroupMetadata.builder().tableInfo(TableInfo.UNKNOWN_TABLE_INFO).metadataInfo(metadataInfo).schema(schema).columnsStatistics(columnsStatistics).metadataStatistics(rowGroupStatistics).hostAffinity(rowGroupMetadata.getHostAffinity()).rowGroupIndex(rgIndexInFile).path(location).build();
}
Also used : ColumnStatistics(org.apache.drill.metastore.statistics.ColumnStatistics) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) ArrayList(java.util.ArrayList) BaseStatisticsKind(org.apache.drill.metastore.statistics.BaseStatisticsKind) TupleSchema(org.apache.drill.exec.record.metadata.TupleSchema) StatisticsHolder(org.apache.drill.metastore.statistics.StatisticsHolder) SchemaPath(org.apache.drill.common.expression.SchemaPath) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata)

Example 92 with TupleMetadata

use of org.apache.drill.exec.record.metadata.TupleMetadata in project drill by apache.

the class TestResultSetLoaderRepeatedList method test3DEarlySchema.

// Adapted from TestRepeatedListAccessors.testSchema3DWriterReader
// That test exercises the low-level schema and writer mechanisms.
// Here we simply ensure that the 3D case continues to work when
// wrapped in the Result Set Loader
@Test
public void test3DEarlySchema() {
    final TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addArray("cube", MinorType.VARCHAR, 3).buildSchema();
    final ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).build();
    final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    rsLoader.startBatch();
    final RowSetLoader writer = rsLoader.writer();
    writer.addRow(1, objArray(objArray(strArray("a", "b"), strArray("c")), objArray(strArray("d", "e", "f"), null), null, objArray())).addRow(2, null).addRow(3, objArray()).addRow(4, objArray(objArray())).addRow(5, singleObjArray(objArray(strArray("g", "h"), strArray("i"))));
    final SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(1, objArray(objArray(strArray("a", "b"), strArray("c")), objArray(strArray("d", "e", "f"), strArray()), objArray(), objArray())).addRow(2, objArray()).addRow(3, objArray()).addRow(4, objArray(objArray())).addRow(5, singleObjArray(objArray(strArray("g", "h"), strArray("i")))).build();
    RowSetUtilities.verify(expected, fixture.wrap(rsLoader.harvest()));
}
Also used : SingleRowSet(org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet) ResultSetLoader(org.apache.drill.exec.physical.resultSet.ResultSetLoader) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) RowSetLoader(org.apache.drill.exec.physical.resultSet.RowSetLoader) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 93 with TupleMetadata

use of org.apache.drill.exec.record.metadata.TupleMetadata in project drill by apache.

the class TestResultSetLoaderRepeatedList method test2DEarlySchema.

@Test
public void test2DEarlySchema() {
    final TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addRepeatedList("list2").addArray(MinorType.VARCHAR).resumeSchema().buildSchema();
    final ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().readerSchema(schema).build();
    final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    do2DTest(schema, rsLoader);
    rsLoader.close();
}
Also used : ResultSetLoader(org.apache.drill.exec.physical.resultSet.ResultSetLoader) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Example 94 with TupleMetadata

use of org.apache.drill.exec.record.metadata.TupleMetadata in project drill by apache.

the class TestResultSetLoaderRepeatedList method test2DOverflow.

@Test
public void test2DOverflow() {
    final TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addRepeatedList("list2").addArray(MinorType.VARCHAR).resumeSchema().buildSchema();
    final ResultSetLoaderImpl.ResultSetOptions options = new ResultSetOptionBuilder().rowCountLimit(ValueVector.MAX_ROW_COUNT).readerSchema(schema).build();
    final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator(), options);
    final RowSetLoader writer = rsLoader.writer();
    // Fill the batch with enough data to cause overflow.
    // Data must be large enough to cause overflow before 64K rows
    // Make a bit bigger to overflow early.
    final int outerSize = 7;
    final int innerSize = 5;
    final int strLength = ValueVector.MAX_BUFFER_SIZE / ValueVector.MAX_ROW_COUNT / outerSize / innerSize + 20;
    final byte[] value = new byte[strLength - 6];
    Arrays.fill(value, (byte) 'X');
    final String strValue = new String(value, Charsets.UTF_8);
    int rowCount = 0;
    int elementCount = 0;
    final ArrayWriter outerWriter = writer.array(1);
    final ArrayWriter innerWriter = outerWriter.array();
    final ScalarWriter elementWriter = innerWriter.scalar();
    rsLoader.startBatch();
    while (!writer.isFull()) {
        writer.start();
        writer.scalar(0).setInt(rowCount);
        for (int j = 0; j < outerSize; j++) {
            for (int k = 0; k < innerSize; k++) {
                elementWriter.setString(String.format("%s%06d", strValue, elementCount));
                elementCount++;
            }
            outerWriter.save();
        }
        writer.save();
        rowCount++;
    }
    // Number of rows should be driven by vector size.
    // Our row count should include the overflow row
    final int expectedCount = ValueVector.MAX_BUFFER_SIZE / (strLength * innerSize * outerSize);
    assertEquals(expectedCount + 1, rowCount);
    // Loader's row count should include only "visible" rows
    assertEquals(expectedCount, writer.rowCount());
    // Total count should include invisible and look-ahead rows.
    assertEquals(expectedCount + 1, rsLoader.totalRowCount());
    // Result should exclude the overflow row
    RowSet result = fixture.wrap(rsLoader.harvest());
    assertEquals(expectedCount, result.rowCount());
    // Verify the data.
    RowSetReader reader = result.reader();
    ArrayReader outerReader = reader.array(1);
    ArrayReader innerReader = outerReader.array();
    ScalarReader strReader = innerReader.scalar();
    int readRowCount = 0;
    int readElementCount = 0;
    while (reader.next()) {
        assertEquals(readRowCount, reader.scalar(0).getInt());
        for (int i = 0; i < outerSize; i++) {
            assertTrue(outerReader.next());
            for (int j = 0; j < innerSize; j++) {
                assertTrue(innerReader.next());
                assertEquals(String.format("%s%06d", strValue, readElementCount), strReader.getString());
                readElementCount++;
            }
            assertFalse(innerReader.next());
        }
        assertFalse(outerReader.next());
        readRowCount++;
    }
    assertEquals(readRowCount, result.rowCount());
    result.clear();
    // Write a few more rows to verify the overflow row.
    rsLoader.startBatch();
    for (int i = 0; i < 1000; i++) {
        writer.start();
        writer.scalar(0).setInt(rowCount);
        for (int j = 0; j < outerSize; j++) {
            for (int k = 0; k < innerSize; k++) {
                elementWriter.setString(String.format("%s%06d", strValue, elementCount));
                elementCount++;
            }
            outerWriter.save();
        }
        writer.save();
        rowCount++;
    }
    result = fixture.wrap(rsLoader.harvest());
    assertEquals(1001, result.rowCount());
    final int startCount = readRowCount;
    reader = result.reader();
    outerReader = reader.array(1);
    innerReader = outerReader.array();
    strReader = innerReader.scalar();
    while (reader.next()) {
        assertEquals(readRowCount, reader.scalar(0).getInt());
        for (int i = 0; i < outerSize; i++) {
            assertTrue(outerReader.next());
            for (int j = 0; j < innerSize; j++) {
                assertTrue(innerReader.next());
                elementWriter.setString(String.format("%s%06d", strValue, readElementCount));
                assertEquals(String.format("%s%06d", strValue, readElementCount), strReader.getString());
                readElementCount++;
            }
            assertFalse(innerReader.next());
        }
        assertFalse(outerReader.next());
        readRowCount++;
    }
    assertEquals(readRowCount - startCount, result.rowCount());
    result.clear();
    rsLoader.close();
}
Also used : SingleRowSet(org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.exec.physical.rowSet.RowSet) ScalarReader(org.apache.drill.exec.vector.accessor.ScalarReader) ArrayReader(org.apache.drill.exec.vector.accessor.ArrayReader) ResultSetLoader(org.apache.drill.exec.physical.resultSet.ResultSetLoader) 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)

Example 95 with TupleMetadata

use of org.apache.drill.exec.record.metadata.TupleMetadata in project drill by apache.

the class TestResultSetLoaderUnions method testVariantListDynamic.

/**
 * Test a variant list created dynamically at load time.
 * The list starts with no type, at which time it can hold
 * only null values. Then we add a Varchar, and finally an
 * Int.
 * <p>
 * This test is superficial. There are many odd cases to consider.
 * <ul>
 * <li>Write nulls to a list with no type. (This test ensures that
 * adding a (nullable) scalar "does the right thing."</li>
 * <li>Add a map to the list. Maps carry no "bits" vector, so null
 * list entries to that point are lost. (For maps, we could go straight
 * to a union, with just a map, to preserve the null states. This whole
 * area is a huge mess...)</li>
 * <li>Do the type transitions when writing to a row. (The tests here
 * do the transition between rows.)</li>
 * </ul>
 *
 * The reason for the sparse coverage is that Drill barely supports lists
 * and unions; most code is just plain broken. Our goal here is not to fix
 * all those problems, just to leave things no more broken than before.
 */
@Test
public void testVariantListDynamic() {
    final ResultSetLoader rsLoader = new ResultSetLoaderImpl(fixture.allocator());
    final RowSetLoader writer = rsLoader.writer();
    // Can write a batch as if this was a repeated Varchar, except
    // that any value can also be null.
    rsLoader.startBatch();
    writer.addColumn(MaterializedField.create("id", Types.required(MinorType.INT)));
    writer.addColumn(MaterializedField.create("list", Types.optional(MinorType.LIST)));
    // Sanity check: should be an array of variants because we said the
    // types within the list are expandable (which is the default.)
    final ArrayWriter arrWriter = writer.array("list");
    assertEquals(ObjectType.VARIANT, arrWriter.entryType());
    final VariantWriter variant = arrWriter.variant();
    // We need to verify that the internal state is what we expect, so
    // the next assertion peeks inside the private bits of the union
    // writer. No client code should ever need to do this, of course.
    assertTrue(((UnionWriterImpl) variant).shim() instanceof EmptyListShim);
    // No types, so all we can do is add a null list, or a list of nulls.
    writer.addRow(1, null).addRow(2, variantArray()).addRow(3, variantArray(null, null));
    // Add a String. Now we can create a list of strings and/or nulls.
    variant.addMember(MinorType.VARCHAR);
    assertTrue(variant.hasType(MinorType.VARCHAR));
    // Sanity check: sniff inside to ensure that the list contains a single
    // type.
    assertTrue(((UnionWriterImpl) variant).shim() instanceof SimpleListShim);
    assertTrue(((ListWriterImpl) arrWriter).vector().getDataVector() instanceof NullableVarCharVector);
    writer.addRow(4, variantArray("fred", null, "barney"));
    // Add an integer. The list vector should be promoted to union.
    // Now we can add both types.
    variant.addMember(MinorType.INT);
    // Sanity check: sniff inside to ensure promotion to union occurred
    assertTrue(((UnionWriterImpl) variant).shim() instanceof UnionVectorShim);
    assertTrue(((ListWriterImpl) arrWriter).vector().getDataVector() instanceof UnionVector);
    writer.addRow(5, variantArray("wilma", null, 30));
    // Verify
    final RowSet result = fixture.wrap(rsLoader.harvest());
    final TupleMetadata schema = new SchemaBuilder().add("id", MinorType.INT).addList("list").addType(MinorType.VARCHAR).addType(MinorType.INT).resumeSchema().buildSchema();
    final SingleRowSet expected = fixture.rowSetBuilder(schema).addRow(1, null).addRow(2, variantArray()).addRow(3, variantArray(null, null)).addRow(4, variantArray("fred", null, "barney")).addRow(5, variantArray("wilma", null, 30)).build();
    RowSetUtilities.verify(expected, result);
}
Also used : VariantWriter(org.apache.drill.exec.vector.accessor.VariantWriter) SingleRowSet(org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet) SingleRowSet(org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet) RowSet(org.apache.drill.exec.physical.rowSet.RowSet) SimpleListShim(org.apache.drill.exec.vector.accessor.writer.SimpleListShim) UnionVector(org.apache.drill.exec.vector.complex.UnionVector) UnionWriterImpl(org.apache.drill.exec.vector.accessor.writer.UnionWriterImpl) UnionVectorShim(org.apache.drill.exec.vector.accessor.writer.UnionVectorShim) ResultSetLoader(org.apache.drill.exec.physical.resultSet.ResultSetLoader) NullableVarCharVector(org.apache.drill.exec.vector.NullableVarCharVector) EmptyListShim(org.apache.drill.exec.vector.accessor.writer.EmptyListShim) 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) SubOperatorTest(org.apache.drill.test.SubOperatorTest) Test(org.junit.Test)

Aggregations

TupleMetadata (org.apache.drill.exec.record.metadata.TupleMetadata)1235 Test (org.junit.Test)1126 SchemaBuilder (org.apache.drill.exec.record.metadata.SchemaBuilder)1008 RowSet (org.apache.drill.exec.physical.rowSet.RowSet)598 SubOperatorTest (org.apache.drill.test.SubOperatorTest)460 RowSetBuilder (org.apache.drill.exec.physical.rowSet.RowSetBuilder)293 SingleRowSet (org.apache.drill.exec.physical.rowSet.RowSet.SingleRowSet)264 ClusterTest (org.apache.drill.test.ClusterTest)261 EvfTest (org.apache.drill.categories.EvfTest)230 RowSetComparison (org.apache.drill.test.rowSet.RowSetComparison)211 ResultSetLoader (org.apache.drill.exec.physical.resultSet.ResultSetLoader)111 JsonTest (org.apache.drill.categories.JsonTest)110 DirectRowSet (org.apache.drill.exec.physical.rowSet.DirectRowSet)109 BaseTest (org.apache.drill.test.BaseTest)106 ColumnMetadata (org.apache.drill.exec.record.metadata.ColumnMetadata)100 RowSetLoader (org.apache.drill.exec.physical.resultSet.RowSetLoader)89 ScalarReader (org.apache.drill.exec.vector.accessor.ScalarReader)72 ScalarWriter (org.apache.drill.exec.vector.accessor.ScalarWriter)69 UserException (org.apache.drill.common.exceptions.UserException)67 SchemaBuilder (org.apache.drill.test.rowSet.schema.SchemaBuilder)65