Search in sources :

Example 16 with MetadataInfo

use of org.apache.drill.metastore.metadata.MetadataInfo in project drill by apache.

the class TestTableMetadataUnitConversion method testSegmentMetadata.

@Test
public void testSegmentMetadata() {
    TableInfo tableInfo = data.basicTableInfo;
    MetadataInfo metadataInfo = MetadataInfo.builder().type(MetadataType.SEGMENT).key("part_int=3").identifier("part_int=3/part_varchar=g").build();
    Path path = new Path("/tmp/nation");
    String unitPath = path.toUri().getPath();
    Set<Path> locations = new HashSet<>();
    locations.add(new Path("part_int=3/part_varchar=g/0_0_0.parquet"));
    locations.add(new Path("part_int=3/part_varchar=g/0_0_1.parquet"));
    List<String> unitLocations = locations.stream().map(location -> location.toUri().getPath()).collect(Collectors.toList());
    // check required fields
    SegmentMetadata requiredFieldsMetadata = SegmentMetadata.builder().tableInfo(tableInfo).metadataInfo(metadataInfo).columnsStatistics(data.columnsStatistics).metadataStatistics(data.metadataStatistics).path(path).locations(locations).build();
    TableMetadataUnit requiredFieldsExpectedUnit = TableMetadataUnit.builder().storagePlugin(tableInfo.storagePlugin()).workspace(tableInfo.workspace()).tableName(tableInfo.name()).metadataType(metadataInfo.type().name()).metadataKey(metadataInfo.key()).metadataIdentifier(metadataInfo.identifier()).columnsStatistics(data.unitColumnsStatistics).metadataStatistics(data.unitMetadataStatistics).lastModifiedTime(BaseMetadata.UNDEFINED_TIME).path(path.toUri().getPath()).location(unitPath).locations(unitLocations).build();
    TableMetadataUnit requiredFieldsUnit = requiredFieldsMetadata.toMetadataUnit();
    assertEquals(requiredFieldsExpectedUnit, requiredFieldsUnit);
    assertNotNull(SegmentMetadata.builder().metadataUnit(requiredFieldsUnit).build());
    SchemaPath column = SchemaPath.getSimplePath("dir1");
    List<String> partitionValues = Collections.singletonList("part_varchar=g");
    SegmentMetadata allFieldsMetadata = SegmentMetadata.builder().tableInfo(tableInfo).metadataInfo(metadataInfo).schema(data.schema).columnsStatistics(data.columnsStatistics).metadataStatistics(data.metadataStatistics).lastModifiedTime(data.lastModifiedTime).path(path).locations(locations).column(column).partitionValues(partitionValues).build();
    TableMetadataUnit allFieldsExpectedUnit = TableMetadataUnit.builder().storagePlugin(tableInfo.storagePlugin()).workspace(tableInfo.workspace()).tableName(tableInfo.name()).metadataType(metadataInfo.type().name()).metadataKey(metadataInfo.key()).metadataIdentifier(metadataInfo.identifier()).schema(data.unitSchema).columnsStatistics(data.unitColumnsStatistics).metadataStatistics(data.unitMetadataStatistics).lastModifiedTime(data.lastModifiedTime).path(path.toUri().getPath()).location(unitPath).locations(unitLocations).column(column.toString()).partitionValues(partitionValues).build();
    TableMetadataUnit allFieldsUnit = allFieldsMetadata.toMetadataUnit();
    assertEquals(allFieldsExpectedUnit, allFieldsUnit);
    assertNotNull(SegmentMetadata.builder().metadataUnit(allFieldsUnit).build());
}
Also used : Path(org.apache.hadoop.fs.Path) SchemaPath(org.apache.drill.common.expression.SchemaPath) MetadataType(org.apache.drill.metastore.metadata.MetadataType) Arrays(java.util.Arrays) TableInfo(org.apache.drill.metastore.metadata.TableInfo) BeforeClass(org.junit.BeforeClass) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) ColumnStatistics(org.apache.drill.metastore.statistics.ColumnStatistics) HashMap(java.util.HashMap) BaseTableMetadata(org.apache.drill.metastore.metadata.BaseTableMetadata) BaseTest(org.apache.drill.test.BaseTest) HashSet(java.util.HashSet) PartitionMetadata(org.apache.drill.metastore.metadata.PartitionMetadata) Map(java.util.Map) Path(org.apache.hadoop.fs.Path) FileMetadata(org.apache.drill.metastore.metadata.FileMetadata) BaseMetadata(org.apache.drill.metastore.metadata.BaseMetadata) ColumnStatisticsKind(org.apache.drill.metastore.statistics.ColumnStatisticsKind) Assert.assertNotNull(org.junit.Assert.assertNotNull) SegmentMetadata(org.apache.drill.metastore.metadata.SegmentMetadata) Collection(java.util.Collection) SchemaPath(org.apache.drill.common.expression.SchemaPath) RowGroupMetadata(org.apache.drill.metastore.metadata.RowGroupMetadata) Set(java.util.Set) Test(org.junit.Test) Category(org.junit.experimental.categories.Category) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) Collectors(java.util.stream.Collectors) TypeProtos(org.apache.drill.common.types.TypeProtos) List(java.util.List) MetastoreTest(org.apache.drill.categories.MetastoreTest) StatisticsHolder(org.apache.drill.metastore.statistics.StatisticsHolder) Collections(java.util.Collections) Assert.assertEquals(org.junit.Assert.assertEquals) SegmentMetadata(org.apache.drill.metastore.metadata.SegmentMetadata) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) SchemaPath(org.apache.drill.common.expression.SchemaPath) TableInfo(org.apache.drill.metastore.metadata.TableInfo) HashSet(java.util.HashSet) BaseTest(org.apache.drill.test.BaseTest) Test(org.junit.Test) MetastoreTest(org.apache.drill.categories.MetastoreTest)

Example 17 with MetadataInfo

use of org.apache.drill.metastore.metadata.MetadataInfo in project drill by apache.

the class TestMetastoreWithEasyFormatPlugin method testAnalyzeOnJsonTable.

@Test
public void testAnalyzeOnJsonTable() throws Exception {
    String tableName = "multilevel/json";
    TableInfo tableInfo = getTableInfo(tableName, "default", "json");
    File table = dirTestWatcher.copyResourceToRoot(Paths.get(tableName));
    Path tablePath = new Path(table.toURI().getPath());
    TupleMetadata schema = new SchemaBuilder().addNullable("dir0", TypeProtos.MinorType.VARCHAR).addNullable("dir1", TypeProtos.MinorType.VARCHAR).addNullable("o_orderkey", TypeProtos.MinorType.BIGINT).addNullable("o_custkey", TypeProtos.MinorType.BIGINT).addNullable("o_orderstatus", TypeProtos.MinorType.VARCHAR).addNullable("o_totalprice", TypeProtos.MinorType.FLOAT8).addNullable("o_orderdate", TypeProtos.MinorType.VARCHAR).addNullable("o_orderpriority", TypeProtos.MinorType.VARCHAR).addNullable("o_clerk", TypeProtos.MinorType.VARCHAR).addNullable("o_shippriority", TypeProtos.MinorType.BIGINT).addNullable("o_comment", TypeProtos.MinorType.VARCHAR).build();
    Map<SchemaPath, ColumnStatistics<?>> tableColumnStatistics = new HashMap<>(TABLE_COLUMN_STATISTICS);
    tableColumnStatistics.put(SchemaPath.getSimplePath("o_custkey"), getColumnStatistics(25L, 1498L, 120L, TypeProtos.MinorType.BIGINT));
    tableColumnStatistics.put(SchemaPath.getSimplePath("o_orderdate"), getColumnStatistics("1994-01-01T00:00:00.000-08:00", "1996-12-19T00:00:00.000-08:00", 120L, TypeProtos.MinorType.VARCHAR));
    tableColumnStatistics.put(SchemaPath.getSimplePath("o_orderkey"), getColumnStatistics(1L, 1319L, 120L, TypeProtos.MinorType.BIGINT));
    tableColumnStatistics.put(SchemaPath.getSimplePath("o_shippriority"), getColumnStatistics(0L, 0L, 120L, TypeProtos.MinorType.BIGINT));
    BaseTableMetadata expectedTableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(TABLE_META_INFO).schema(schema).location(new Path(table.toURI().getPath())).columnsStatistics(tableColumnStatistics).metadataStatistics(Arrays.asList(new StatisticsHolder<>(120L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ALL, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).build();
    TableInfo baseTableInfo = TableInfo.builder().name(tableName).storagePlugin("dfs").workspace("default").build();
    Map<SchemaPath, ColumnStatistics<?>> dir0CSVStats = new HashMap<>(DIR0_1994_SEGMENT_COLUMN_STATISTICS);
    dir0CSVStats.put(SchemaPath.getSimplePath("o_custkey"), getColumnStatistics(25L, 1469L, 40L, TypeProtos.MinorType.BIGINT));
    dir0CSVStats.put(SchemaPath.getSimplePath("o_orderdate"), getColumnStatistics("1994-01-01T00:00:00.000-08:00", "1994-12-23T00:00:00.000-08:00", 40L, TypeProtos.MinorType.VARCHAR));
    dir0CSVStats.put(SchemaPath.getSimplePath("o_orderkey"), getColumnStatistics(5L, 1031L, 40L, TypeProtos.MinorType.BIGINT));
    dir0CSVStats.put(SchemaPath.getSimplePath("o_shippriority"), getColumnStatistics(0L, 0L, 40L, TypeProtos.MinorType.BIGINT));
    SegmentMetadata dir0 = SegmentMetadata.builder().tableInfo(baseTableInfo).metadataInfo(MetadataInfo.builder().type(MetadataType.SEGMENT).identifier("1994").key("1994").build()).path(new Path(tablePath, "1994")).schema(schema).lastModifiedTime(getMaxLastModified(new File(table, "1994"))).column(SchemaPath.getSimplePath("dir0")).columnsStatistics(dir0CSVStats).metadataStatistics(Collections.singletonList(new StatisticsHolder<>(40L, TableStatisticsKind.ROW_COUNT))).locations(ImmutableSet.of(new Path(tablePath, "1994/Q1/orders_94_q1.json"), new Path(tablePath, "1994/Q2/orders_94_q2.json"), new Path(tablePath, "1994/Q3/orders_94_q3.json"), new Path(tablePath, "1994/Q4/orders_94_q4.json"))).partitionValues(Collections.singletonList("1994")).build();
    Set<Path> expectedTopLevelSegmentLocations = ImmutableSet.of(new Path(tablePath, "1994"), new Path(tablePath, "1995"), new Path(tablePath, "1996"));
    Set<Set<Path>> expectedSegmentFilesLocations = new HashSet<>();
    Set<Path> segmentFiles = ImmutableSet.of(new Path(tablePath, "1994/Q2/orders_94_q2.json"), new Path(tablePath, "1994/Q4/orders_94_q4.json"), new Path(tablePath, "1994/Q1/orders_94_q1.json"), new Path(tablePath, "1994/Q3/orders_94_q3.json"));
    expectedSegmentFilesLocations.add(segmentFiles);
    segmentFiles = ImmutableSet.of(new Path(tablePath, "1995/Q2/orders_95_q2.json"), new Path(tablePath, "1995/Q4/orders_95_q4.json"), new Path(tablePath, "1995/Q1/orders_95_q1.json"), new Path(tablePath, "1995/Q3/orders_95_q3.json"));
    expectedSegmentFilesLocations.add(segmentFiles);
    segmentFiles = ImmutableSet.of(new Path(tablePath, "1996/Q3/orders_96_q3.json"), new Path(tablePath, "1996/Q2/orders_96_q2.json"), new Path(tablePath, "1996/Q4/orders_96_q4.json"), new Path(tablePath, "1996/Q1/orders_96_q1.json"));
    expectedSegmentFilesLocations.add(segmentFiles);
    Map<SchemaPath, ColumnStatistics<?>> dir0q1Stats = new HashMap<>(DIR0_1994_Q1_SEGMENT_COLUMN_STATISTICS);
    dir0q1Stats.put(SchemaPath.getSimplePath("o_custkey"), getColumnStatistics(392L, 1411L, 10L, TypeProtos.MinorType.BIGINT));
    dir0q1Stats.put(SchemaPath.getSimplePath("o_orderdate"), getColumnStatistics("1994-01-01T00:00:00.000-08:00", "1994-03-26T00:00:00.000-08:00", 10L, TypeProtos.MinorType.VARCHAR));
    dir0q1Stats.put(SchemaPath.getSimplePath("o_orderkey"), getColumnStatistics(66L, 833L, 10L, TypeProtos.MinorType.BIGINT));
    dir0q1Stats.put(SchemaPath.getSimplePath("o_shippriority"), getColumnStatistics(0L, 0L, 10L, TypeProtos.MinorType.BIGINT));
    long dir0q1lastModified = new File(new File(new File(table, "1994"), "Q1"), "orders_94_q1.json").lastModified();
    FileMetadata dir01994q1File = FileMetadata.builder().tableInfo(baseTableInfo).metadataInfo(MetadataInfo.builder().type(MetadataType.FILE).identifier("1994/Q1/orders_94_q1.json").key("1994").build()).schema(schema).lastModifiedTime(dir0q1lastModified).columnsStatistics(dir0q1Stats).metadataStatistics(Collections.singletonList(new StatisticsHolder<>(10L, TableStatisticsKind.ROW_COUNT))).path(new Path(tablePath, "1994/Q1/orders_94_q1.json")).build();
    try {
        testBuilder().sqlQuery("analyze table table(dfs.`%s`(schema=>%s)) refresh metadata", tableName, SCHEMA_STRING).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.default.%s]", tableName)).go();
        BaseTableMetadata actualTableMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().tableMetadata(tableInfo);
        assertEquals(expectedTableMetadata, actualTableMetadata);
        List<SegmentMetadata> topSegmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByColumn(tableInfo, null, "`dir0`");
        SegmentMetadata actualDir0Metadata = topSegmentMetadata.stream().filter(unit -> unit.getMetadataInfo().identifier().equals("1994")).findAny().orElseThrow(() -> new AssertionError("Segment is absent"));
        Set<Path> locations = actualDir0Metadata.getLocations();
        actualDir0Metadata.toBuilder().locations(locations);
        assertEquals(dir0, actualDir0Metadata);
        Set<Path> topLevelSegmentLocations = topSegmentMetadata.stream().map(SegmentMetadata::getLocation).collect(Collectors.toSet());
        // verify top segments locations
        assertEquals(expectedTopLevelSegmentLocations, topLevelSegmentLocations);
        Set<Set<Path>> segmentFilesLocations = topSegmentMetadata.stream().map(SegmentMetadata::getLocations).collect(Collectors.toSet());
        assertEquals(expectedSegmentFilesLocations, segmentFilesLocations);
        // verify nested segments
        List<SegmentMetadata> nestedSegmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByColumn(tableInfo, null, "`dir1`");
        assertEquals(12, nestedSegmentMetadata.size());
        SegmentMetadata dir01994q1Segment = SegmentMetadata.builder().tableInfo(baseTableInfo).metadataInfo(MetadataInfo.builder().type(MetadataType.SEGMENT).identifier("1994/Q1").key("1994").build()).path(new Path(new Path(tablePath, "1994"), "Q1")).schema(schema).lastModifiedTime(getMaxLastModified(new File(new File(table, "1994"), "Q1"))).column(SchemaPath.getSimplePath("dir1")).columnsStatistics(dir0q1Stats).metadataStatistics(Collections.singletonList(new StatisticsHolder<>(10L, TableStatisticsKind.ROW_COUNT))).locations(ImmutableSet.of(new Path(tablePath, "1994/Q1/orders_94_q1.json"))).partitionValues(Collections.singletonList("Q1")).build();
        // verify segment for 1994
        assertEquals(dir01994q1Segment, nestedSegmentMetadata.stream().filter(unit -> unit.getMetadataInfo().identifier().equals("1994/Q1")).findAny().orElse(null));
        // verify files metadata
        List<FileMetadata> filesMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().filesMetadata(tableInfo, null, null);
        assertEquals(12, filesMetadata.size());
        // verify first file metadata
        assertEquals(dir01994q1File, filesMetadata.stream().filter(unit -> unit.getMetadataInfo().identifier().equals("1994/Q1/orders_94_q1.json")).findAny().orElse(null));
    } finally {
        run("analyze table dfs.`%s` drop metadata if exists", tableName);
    }
}
Also used : Path(org.apache.hadoop.fs.Path) SchemaPath(org.apache.drill.common.expression.SchemaPath) ColumnStatistics(org.apache.drill.metastore.statistics.ColumnStatistics) TestMetastoreCommands.getColumnStatistics(org.apache.drill.exec.sql.TestMetastoreCommands.getColumnStatistics) MetadataType(org.apache.drill.metastore.metadata.MetadataType) Arrays(java.util.Arrays) ClusterTest(org.apache.drill.test.ClusterTest) TABLE_META_INFO(org.apache.drill.exec.sql.TestMetastoreCommands.TABLE_META_INFO) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) ColumnStatistics(org.apache.drill.metastore.statistics.ColumnStatistics) TABLE_COLUMN_STATISTICS(org.apache.drill.exec.sql.TestMetastoreCommands.TABLE_COLUMN_STATISTICS) BaseTableMetadata(org.apache.drill.metastore.metadata.BaseTableMetadata) SlowTest(org.apache.drill.categories.SlowTest) Map(java.util.Map) Path(org.apache.hadoop.fs.Path) ColumnStatisticsKind(org.apache.drill.metastore.statistics.ColumnStatisticsKind) SegmentMetadata(org.apache.drill.metastore.metadata.SegmentMetadata) SchemaPath(org.apache.drill.common.expression.SchemaPath) Set(java.util.Set) Category(org.junit.experimental.categories.Category) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) Collectors(java.util.stream.Collectors) TypeProtos(org.apache.drill.common.types.TypeProtos) List(java.util.List) TestMetastoreCommands.getMaxLastModified(org.apache.drill.exec.sql.TestMetastoreCommands.getMaxLastModified) MetastoreTest(org.apache.drill.categories.MetastoreTest) ExecConstants(org.apache.drill.exec.ExecConstants) ClusterFixtureBuilder(org.apache.drill.test.ClusterFixtureBuilder) TestMetastoreCommands.getColumnStatistics(org.apache.drill.exec.sql.TestMetastoreCommands.getColumnStatistics) TableInfo(org.apache.drill.metastore.metadata.TableInfo) BeforeClass(org.junit.BeforeClass) HashMap(java.util.HashMap) HashSet(java.util.HashSet) TableStatisticsKind(org.apache.drill.metastore.statistics.TableStatisticsKind) ExpectedException(org.junit.rules.ExpectedException) Before(org.junit.Before) FileMetadata(org.apache.drill.metastore.metadata.FileMetadata) TestMetastoreCommands.getBaseTableMetadata(org.apache.drill.exec.sql.TestMetastoreCommands.getBaseTableMetadata) ClusterFixture(org.apache.drill.test.ClusterFixture) Assert.assertTrue(org.junit.Assert.assertTrue) FileUtils(org.apache.commons.io.FileUtils) Test(org.junit.Test) File(java.io.File) ImmutableMap(org.apache.drill.shaded.guava.com.google.common.collect.ImmutableMap) Rule(org.junit.Rule) DIR0_1994_SEGMENT_COLUMN_STATISTICS(org.apache.drill.exec.sql.TestMetastoreCommands.DIR0_1994_SEGMENT_COLUMN_STATISTICS) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) Paths(java.nio.file.Paths) DIR0_1994_Q1_SEGMENT_COLUMN_STATISTICS(org.apache.drill.exec.sql.TestMetastoreCommands.DIR0_1994_Q1_SEGMENT_COLUMN_STATISTICS) StatisticsHolder(org.apache.drill.metastore.statistics.StatisticsHolder) ImmutableSet(org.apache.drill.shaded.guava.com.google.common.collect.ImmutableSet) MetastoreTableInfo(org.apache.drill.metastore.components.tables.MetastoreTableInfo) Collections(java.util.Collections) Assert.assertEquals(org.junit.Assert.assertEquals) Set(java.util.Set) HashSet(java.util.HashSet) ImmutableSet(org.apache.drill.shaded.guava.com.google.common.collect.ImmutableSet) HashMap(java.util.HashMap) FileMetadata(org.apache.drill.metastore.metadata.FileMetadata) SegmentMetadata(org.apache.drill.metastore.metadata.SegmentMetadata) StatisticsHolder(org.apache.drill.metastore.statistics.StatisticsHolder) BaseTableMetadata(org.apache.drill.metastore.metadata.BaseTableMetadata) TestMetastoreCommands.getBaseTableMetadata(org.apache.drill.exec.sql.TestMetastoreCommands.getBaseTableMetadata) SchemaPath(org.apache.drill.common.expression.SchemaPath) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) SchemaBuilder(org.apache.drill.exec.record.metadata.SchemaBuilder) TableInfo(org.apache.drill.metastore.metadata.TableInfo) MetastoreTableInfo(org.apache.drill.metastore.components.tables.MetastoreTableInfo) File(java.io.File) HashSet(java.util.HashSet) ClusterTest(org.apache.drill.test.ClusterTest) SlowTest(org.apache.drill.categories.SlowTest) MetastoreTest(org.apache.drill.categories.MetastoreTest) Test(org.junit.Test)

Example 18 with MetadataInfo

use of org.apache.drill.metastore.metadata.MetadataInfo in project drill by apache.

the class MetadataControllerBatch method getTableMetadata.

private BaseTableMetadata getTableMetadata(TupleReader reader, List<StatisticsHolder<?>> metadataStatistics, Map<SchemaPath, ColumnStatistics<?>> columnStatistics) {
    List<StatisticsHolder<?>> updatedMetaStats = new ArrayList<>(metadataStatistics);
    updatedMetaStats.add(new StatisticsHolder<>(popConfig.getContext().analyzeMetadataLevel(), TableStatisticsKind.ANALYZE_METADATA_LEVEL));
    MetadataInfo metadataInfo = MetadataInfo.builder().type(MetadataType.TABLE).key(MetadataInfo.GENERAL_INFO_KEY).build();
    BaseTableMetadata tableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(metadataInfo).columnsStatistics(columnStatistics).metadataStatistics(updatedMetaStats).partitionKeys(Collections.emptyMap()).interestingColumns(popConfig.getContext().interestingColumns()).location(popConfig.getContext().location()).lastModifiedTime(Long.parseLong(reader.column(columnNamesOptions.lastModifiedTime()).scalar().getString())).schema(TupleMetadata.of(reader.column(MetastoreAnalyzeConstants.SCHEMA_FIELD).scalar().getString())).build();
    if (context.getOptions().getOption(PlannerSettings.STATISTICS_USE)) {
        DrillStatsTable statistics = new DrillStatsTable(statisticsCollector.getStatistics());
        Map<SchemaPath, ColumnStatistics<?>> tableColumnStatistics = ParquetTableMetadataUtils.getColumnStatistics(tableMetadata.getSchema(), statistics);
        tableMetadata = tableMetadata.cloneWithStats(tableColumnStatistics, DrillStatsTable.getEstimatedTableStats(statistics));
    }
    return tableMetadata;
}
Also used : ColumnStatistics(org.apache.drill.metastore.statistics.ColumnStatistics) StatisticsHolder(org.apache.drill.metastore.statistics.StatisticsHolder) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) DrillStatsTable(org.apache.drill.exec.planner.common.DrillStatsTable) BaseTableMetadata(org.apache.drill.metastore.metadata.BaseTableMetadata) SchemaPath(org.apache.drill.common.expression.SchemaPath) ArrayList(java.util.ArrayList)

Example 19 with MetadataInfo

use of org.apache.drill.metastore.metadata.MetadataInfo in project drill by apache.

the class MetadataControllerBatch method getMetadataUnits.

private List<TableMetadataUnit> getMetadataUnits(VectorContainer container) {
    List<TableMetadataUnit> metadataUnits = new ArrayList<>();
    RowSetReader reader = DirectRowSet.fromContainer(container).reader();
    while (reader.next()) {
        metadataUnits.addAll(getMetadataUnits(reader, 0));
    }
    if (metadataToHandle != null) {
        // leaves only table metadata and metadata which belongs to segments to be overridden
        metadataUnits = metadataUnits.stream().filter(tableMetadataUnit -> metadataToHandle.values().stream().map(MetadataInfo::key).anyMatch(s -> s.equals(tableMetadataUnit.metadataKey())) || MetadataType.TABLE.name().equals(tableMetadataUnit.metadataType())).collect(Collectors.toList());
        // leaves only metadata which should be fetched from the Metastore
        metadataUnits.stream().map(TableMetadataUnit::metadataIdentifier).forEach(metadataToHandle::remove);
        List<TableMetadataUnit> metadata = metadataToHandle.isEmpty() ? Collections.emptyList() : tables.basicRequests().metadata(popConfig.getContext().tableInfo(), metadataToHandle.values());
        metadataUnits.addAll(metadata);
    }
    // checks whether metadataUnits contains not only table metadata before adding default segment
    // to avoid case when only table metadata should be updated and / or root segments removed
    boolean insertDefaultSegment = metadataUnits.size() > 1 && metadataUnits.stream().noneMatch(metadataUnit -> metadataUnit.metadataType().equals(MetadataType.SEGMENT.name()));
    if (insertDefaultSegment) {
        TableMetadataUnit defaultSegmentMetadata = getDefaultSegment(metadataUnits);
        metadataUnits.add(defaultSegmentMetadata);
    }
    return metadataUnits;
}
Also used : MetadataType(org.apache.drill.metastore.metadata.MetadataType) BaseStatisticsKind(org.apache.drill.metastore.statistics.BaseStatisticsKind) MetastoreColumn(org.apache.drill.metastore.MetastoreColumn) UserException(org.apache.drill.common.exceptions.UserException) LoggerFactory(org.slf4j.LoggerFactory) Types(org.apache.drill.common.types.Types) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) RowSetReader(org.apache.drill.exec.physical.rowSet.RowSetReader) VectorContainer(org.apache.drill.exec.record.VectorContainer) ColumnStatistics(org.apache.drill.metastore.statistics.ColumnStatistics) ArrayListMultimap(org.apache.drill.shaded.guava.com.google.common.collect.ArrayListMultimap) StringUtils(org.apache.commons.lang3.StringUtils) ArrayReader(org.apache.drill.exec.vector.accessor.ArrayReader) BaseTableMetadata(org.apache.drill.metastore.metadata.BaseTableMetadata) StatisticsRecordWriterImpl(org.apache.drill.exec.store.StatisticsRecordWriterImpl) PartitionMetadata(org.apache.drill.metastore.metadata.PartitionMetadata) Map(java.util.Map) FieldConverter(org.apache.drill.exec.store.EventBasedRecordWriter.FieldConverter) Path(org.apache.hadoop.fs.Path) BatchSchema(org.apache.drill.exec.record.BatchSchema) ColumnMetadata(org.apache.drill.exec.record.metadata.ColumnMetadata) Multimap(org.apache.drill.shaded.guava.com.google.common.collect.Multimap) ColumnStatisticsKind(org.apache.drill.metastore.statistics.ColumnStatisticsKind) Delete(org.apache.drill.metastore.operate.Delete) TableMetadataUnit(org.apache.drill.metastore.components.tables.TableMetadataUnit) SegmentMetadata(org.apache.drill.metastore.metadata.SegmentMetadata) SchemaPath(org.apache.drill.common.expression.SchemaPath) RecordBatch(org.apache.drill.exec.record.RecordBatch) Set(java.util.Set) MetastoreAnalyzeConstants(org.apache.drill.exec.metastore.analyze.MetastoreAnalyzeConstants) Collectors(java.util.stream.Collectors) TupleMetadata(org.apache.drill.exec.record.metadata.TupleMetadata) FieldReader(org.apache.drill.exec.vector.complex.reader.FieldReader) TypeProtos(org.apache.drill.common.types.TypeProtos) List(java.util.List) AbstractBinaryRecordBatch(org.apache.drill.exec.record.AbstractBinaryRecordBatch) Preconditions(org.apache.drill.shaded.guava.com.google.common.base.Preconditions) ObjectReader(org.apache.drill.exec.vector.accessor.ObjectReader) TableInfo(org.apache.drill.metastore.metadata.TableInfo) MetadataIdentifierUtils(org.apache.drill.exec.metastore.analyze.MetadataIdentifierUtils) TupleReader(org.apache.drill.exec.vector.accessor.TupleReader) Modify(org.apache.drill.metastore.operate.Modify) MetadataControllerContext(org.apache.drill.exec.metastore.analyze.MetadataControllerContext) HashMap(java.util.HashMap) BitVector(org.apache.drill.exec.vector.BitVector) Function(java.util.function.Function) VectorWrapper(org.apache.drill.exec.record.VectorWrapper) ArrayList(java.util.ArrayList) ColumnNamesOptions(org.apache.drill.exec.metastore.ColumnNamesOptions) HashSet(java.util.HashSet) OutOfMemoryException(org.apache.drill.exec.exception.OutOfMemoryException) DirectRowSet(org.apache.drill.exec.physical.rowSet.DirectRowSet) DrillStatsTable(org.apache.drill.exec.planner.common.DrillStatsTable) WriterPrel(org.apache.drill.exec.planner.physical.WriterPrel) TableStatisticsKind(org.apache.drill.metastore.statistics.TableStatisticsKind) FragmentContext(org.apache.drill.exec.ops.FragmentContext) FileMetadata(org.apache.drill.metastore.metadata.FileMetadata) StatisticsRecordCollector(org.apache.drill.exec.store.StatisticsRecordCollector) BaseMetadata(org.apache.drill.metastore.metadata.BaseMetadata) Logger(org.slf4j.Logger) ExactStatisticsConstants(org.apache.drill.metastore.statistics.ExactStatisticsConstants) RowGroupMetadata(org.apache.drill.metastore.metadata.RowGroupMetadata) StatisticsKind(org.apache.drill.metastore.statistics.StatisticsKind) IOException(java.io.IOException) FilterExpression(org.apache.drill.metastore.expressions.FilterExpression) StatisticsCollectorImpl(org.apache.drill.exec.store.easy.json.StatisticsCollectorImpl) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) ParquetTableMetadataUtils(org.apache.drill.exec.store.parquet.ParquetTableMetadataUtils) VarCharVector(org.apache.drill.exec.vector.VarCharVector) StatisticsHolder(org.apache.drill.metastore.statistics.StatisticsHolder) MetadataControllerPOP(org.apache.drill.exec.physical.config.MetadataControllerPOP) Tables(org.apache.drill.metastore.components.tables.Tables) Collections(java.util.Collections) AnalyzeColumnUtils(org.apache.drill.exec.metastore.analyze.AnalyzeColumnUtils) MetastoreTableInfo(org.apache.drill.metastore.components.tables.MetastoreTableInfo) ObjectType(org.apache.drill.exec.vector.accessor.ObjectType) TableMetadataUnit(org.apache.drill.metastore.components.tables.TableMetadataUnit) ArrayList(java.util.ArrayList) RowSetReader(org.apache.drill.exec.physical.rowSet.RowSetReader)

Example 20 with MetadataInfo

use of org.apache.drill.metastore.metadata.MetadataInfo in project drill by apache.

the class FileMetadataInfoCollector method init.

private void init(FormatSelection selection, PlannerSettings settings, Supplier<TableScan> tableScanSupplier, List<SchemaPath> interestingColumns, int segmentColumnsCount) throws IOException {
    List<SchemaPath> metastoreInterestingColumns = Optional.ofNullable(basicRequests.interestingColumnsAndPartitionKeys(tableInfo).interestingColumns()).map(metastoreInterestingColumnNames -> metastoreInterestingColumnNames.stream().map(SchemaPath::parseFromString).collect(Collectors.toList())).orElse(null);
    Map<String, Long> filesNamesLastModifiedTime = basicRequests.filesLastModifiedTime(tableInfo, null, null);
    List<String> newFiles = new ArrayList<>();
    List<String> updatedFiles = new ArrayList<>();
    List<String> removedFiles = new ArrayList<>(filesNamesLastModifiedTime.keySet());
    List<String> allFiles = new ArrayList<>();
    for (FileStatus fileStatus : getFileStatuses(selection)) {
        String path = Path.getPathWithoutSchemeAndAuthority(fileStatus.getPath()).toUri().getPath();
        Long lastModificationTime = filesNamesLastModifiedTime.get(path);
        if (lastModificationTime == null) {
            newFiles.add(path);
        } else if (lastModificationTime < fileStatus.getModificationTime()) {
            updatedFiles.add(path);
        }
        removedFiles.remove(path);
        allFiles.add(path);
    }
    String selectionRoot = selection.getSelection().getSelectionRoot().toUri().getPath();
    if (!Objects.equals(metastoreInterestingColumns, interestingColumns) && metastoreInterestingColumns != null && (interestingColumns == null || !metastoreInterestingColumns.containsAll(interestingColumns)) || TableStatisticsKind.ANALYZE_METADATA_LEVEL.getValue(basicRequests.tableMetadata(tableInfo)).compareTo(metadataLevel) != 0) {
        // do not update table scan and lists of segments / files / row groups,
        // metadata should be recalculated
        tableScan = tableScanSupplier.get();
        metadataToRemove.addAll(getMetadataInfoList(selectionRoot, removedFiles, MetadataType.SEGMENT, 0));
        return;
    }
    // checks whether there are no new, updated and removed files
    if (!newFiles.isEmpty() || !updatedFiles.isEmpty() || !removedFiles.isEmpty()) {
        List<String> scanFiles = new ArrayList<>(newFiles);
        scanFiles.addAll(updatedFiles);
        // updates scan to read updated / new files
        tableScan = getTableScan(settings, tableScanSupplier.get(), scanFiles);
        // iterates from the end;
        // takes deepest updated segments;
        // finds their parents:
        // - fetches all segments for parent level;
        // - filters segments to leave parents only;
        // obtains all child segments;
        // filters child segments for filtered parent segments
        int lastSegmentIndex = segmentColumnsCount - 1;
        List<String> scanAndRemovedFiles = new ArrayList<>(scanFiles);
        scanAndRemovedFiles.addAll(removedFiles);
        // 1. Obtain files info for files from the same folder without removed files
        // 2. Get segments for obtained files + segments for removed files
        // 3. Get parent segments
        // 4. Get other segments for the same parent segment
        // 5. Remove segments which have only removed files (matched for removedFileInfo and don't match to filesInfo)
        // 6. Do the same for parent segments
        List<MetadataInfo> allFilesInfo = getMetadataInfoList(selectionRoot, allFiles, MetadataType.FILE, 0);
        // first pass: collect updated segments even without files, they will be removed later
        List<MetadataInfo> leafSegments = getMetadataInfoList(selectionRoot, scanAndRemovedFiles, MetadataType.SEGMENT, lastSegmentIndex);
        List<MetadataInfo> removedFilesMetadata = getMetadataInfoList(selectionRoot, removedFiles, MetadataType.FILE, 0);
        List<MetadataInfo> scanFilesInfo = getMetadataInfoList(selectionRoot, scanAndRemovedFiles, MetadataType.FILE, 0);
        // files from scan + files from the same folder without removed files
        filesInfo = leafSegments.stream().filter(parent -> scanFilesInfo.stream().anyMatch(child -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier()))).flatMap(parent -> allFilesInfo.stream().filter(child -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier()))).collect(Collectors.toList());
        Multimap<Integer, MetadataInfo> allSegments = populateSegments(removedFiles, allFiles, selectionRoot, lastSegmentIndex, leafSegments, removedFilesMetadata);
        List<MetadataInfo> allRowGroupsInfo = getAllRowGroupsMetadataInfos(allFiles);
        rowGroupsInfo = allRowGroupsInfo.stream().filter(child -> filesInfo.stream().map(MetadataInfo::identifier).anyMatch(parent -> MetadataIdentifierUtils.isMetadataKeyParent(parent, child.identifier()))).collect(Collectors.toList());
        List<MetadataInfo> segmentsToUpdate = getMetadataInfoList(selectionRoot, scanAndRemovedFiles, MetadataType.SEGMENT, 0);
        allMetaToHandle = Streams.concat(allSegments.values().stream(), allFilesInfo.stream(), allRowGroupsInfo.stream()).filter(child -> segmentsToUpdate.stream().anyMatch(parent -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier()))).filter(parent -> removedFilesMetadata.stream().noneMatch(child -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier())) || filesInfo.stream().anyMatch(child -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier()))).collect(Collectors.toList());
        // removed top-level segments are handled separately since their metadata is not overridden when producing writing to the Metastore
        List<MetadataInfo> removedTopSegments = getMetadataInfoList(selectionRoot, removedFiles, MetadataType.SEGMENT, 0).stream().filter(parent -> removedFilesMetadata.stream().anyMatch(child -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier())) && allFilesInfo.stream().noneMatch(child -> MetadataIdentifierUtils.isMetadataKeyParent(parent.identifier(), child.identifier()))).collect(Collectors.toList());
        metadataToRemove.addAll(removedTopSegments);
        segmentsToUpdate.stream().filter(segment -> !removedTopSegments.contains(segment)).forEach(allMetaToHandle::add);
    } else {
        // table metadata may still be actual
        outdated = false;
    }
}
Also used : SchemalessScan(org.apache.drill.exec.physical.base.SchemalessScan) MetadataType(org.apache.drill.metastore.metadata.MetadataType) TableScan(org.apache.calcite.rel.core.TableScan) Arrays(java.util.Arrays) TableInfo(org.apache.drill.metastore.metadata.TableInfo) MetastoreColumn(org.apache.drill.metastore.MetastoreColumn) FileSystem(org.apache.hadoop.fs.FileSystem) DrillRel(org.apache.drill.exec.planner.logical.DrillRel) Streams(org.apache.drill.shaded.guava.com.google.common.collect.Streams) DrillScanRel(org.apache.drill.exec.planner.logical.DrillScanRel) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) DrillFileSystem(org.apache.drill.exec.store.dfs.DrillFileSystem) ArrayListMultimap(org.apache.drill.shaded.guava.com.google.common.collect.ArrayListMultimap) FileStatus(org.apache.hadoop.fs.FileStatus) DrillTable(org.apache.drill.exec.planner.logical.DrillTable) ColumnExplorer(org.apache.drill.exec.store.ColumnExplorer) Supplier(java.util.function.Supplier) ArrayList(java.util.ArrayList) Configuration(org.apache.hadoop.conf.Configuration) Map(java.util.Map) FormatSelection(org.apache.drill.exec.store.dfs.FormatSelection) ImpersonationUtil(org.apache.drill.exec.util.ImpersonationUtil) Path(org.apache.hadoop.fs.Path) FileSelection(org.apache.drill.exec.store.dfs.FileSelection) Multimap(org.apache.drill.shaded.guava.com.google.common.collect.Multimap) TableStatisticsKind(org.apache.drill.metastore.statistics.TableStatisticsKind) PartitionLocation(org.apache.drill.exec.planner.PartitionLocation) BasicTablesRequests(org.apache.drill.metastore.components.tables.BasicTablesRequests) Collection(java.util.Collection) SchemaPath(org.apache.drill.common.expression.SchemaPath) IOException(java.io.IOException) Collectors(java.util.stream.Collectors) Objects(java.util.Objects) DrillFileSystemUtil(org.apache.drill.exec.util.DrillFileSystemUtil) List(java.util.List) Lists(org.apache.drill.shaded.guava.com.google.common.collect.Lists) FileSystemPartitionDescriptor(org.apache.drill.exec.planner.FileSystemPartitionDescriptor) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) Optional(java.util.Optional) Collections(java.util.Collections) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) FileStatus(org.apache.hadoop.fs.FileStatus) ArrayList(java.util.ArrayList) SchemaPath(org.apache.drill.common.expression.SchemaPath)

Aggregations

MetadataInfo (org.apache.drill.metastore.metadata.MetadataInfo)23 SchemaPath (org.apache.drill.common.expression.SchemaPath)21 StatisticsHolder (org.apache.drill.metastore.statistics.StatisticsHolder)16 Path (org.apache.hadoop.fs.Path)16 TableInfo (org.apache.drill.metastore.metadata.TableInfo)15 ColumnStatistics (org.apache.drill.metastore.statistics.ColumnStatistics)14 HashMap (java.util.HashMap)13 HashSet (java.util.HashSet)13 List (java.util.List)13 BaseTableMetadata (org.apache.drill.metastore.metadata.BaseTableMetadata)13 FileMetadata (org.apache.drill.metastore.metadata.FileMetadata)13 ArrayList (java.util.ArrayList)12 Collections (java.util.Collections)12 Collectors (java.util.stream.Collectors)12 TupleMetadata (org.apache.drill.exec.record.metadata.TupleMetadata)12 MetadataType (org.apache.drill.metastore.metadata.MetadataType)12 Map (java.util.Map)11 Set (java.util.Set)10 TypeProtos (org.apache.drill.common.types.TypeProtos)10 PlannerSettings (org.apache.drill.exec.planner.physical.PlannerSettings)10