Examples with StatisticsHolder org.apache.drill.metastore.statistics.StatisticsHolder used on opensource projects

Example 6 with StatisticsHolder

use of org.apache.drill.metastore.statistics.StatisticsHolder in project drill by apache.

the class MetadataControllerBatch method getFileMetadata.

private FileMetadata getFileMetadata(TupleReader reader, List<StatisticsHolder<?>> metadataStatistics, Map<SchemaPath, ColumnStatistics<?>> columnStatistics, int nestingLevel) {
    List<String> segmentColumns = popConfig.getContext().segmentColumns();
    String segmentKey = segmentColumns.size() > 0 ? reader.column(segmentColumns.iterator().next()).scalar().getString() : MetadataInfo.DEFAULT_SEGMENT_KEY;
    List<String> partitionValues = segmentColumns.stream().limit(nestingLevel - 1).map(columnName -> reader.column(columnName).scalar().getString()).collect(Collectors.toList());
    Path path = new Path(reader.column(MetastoreAnalyzeConstants.LOCATION_FIELD).scalar().getString());
    String metadataIdentifier = MetadataIdentifierUtils.getFileMetadataIdentifier(partitionValues, path);
    MetadataInfo metadataInfo = MetadataInfo.builder().type(MetadataType.FILE).key(segmentKey).identifier(StringUtils.defaultIfEmpty(metadataIdentifier, null)).build();
    return FileMetadata.builder().tableInfo(tableInfo).metadataInfo(metadataInfo).columnsStatistics(columnStatistics).metadataStatistics(metadataStatistics).path(path).lastModifiedTime(Long.parseLong(reader.column(columnNamesOptions.lastModifiedTime()).scalar().getString())).schema(TupleMetadata.of(reader.column(MetastoreAnalyzeConstants.SCHEMA_FIELD).scalar().getString())).build();
}

Example 7 with StatisticsHolder

use of org.apache.drill.metastore.statistics.StatisticsHolder in project drill by apache.

the class DrillStatsTable method getEstimatedColumnStats.

/**
 * Returns list of {@link StatisticsKind} and statistics values obtained from specified {@link DrillStatsTable} for specified column.
 *
 * @param statsProvider the source of statistics
 * @param fieldName     name of the columns whose statistics should be obtained
 * @return list of {@link StatisticsKind} and statistics values
 */
public static List<StatisticsHolder<?>> getEstimatedColumnStats(DrillStatsTable statsProvider, SchemaPath fieldName) {
    if (statsProvider != null && statsProvider.isMaterialized()) {
        List<StatisticsHolder<?>> statisticsValues = new ArrayList<>();
        Double ndv = statsProvider.getNdv(fieldName);
        if (ndv != null) {
            statisticsValues.add(new StatisticsHolder<>(ndv, ColumnStatisticsKind.NDV));
        }
        Double nonNullCount = statsProvider.getNNRowCount(fieldName);
        if (nonNullCount != null) {
            statisticsValues.add(new StatisticsHolder<>(nonNullCount, ColumnStatisticsKind.NON_NULL_COUNT));
        }
        Histogram histogram = statsProvider.getHistogram(fieldName);
        if (histogram != null) {
            statisticsValues.add(new StatisticsHolder<>(histogram, ColumnStatisticsKind.HISTOGRAM));
        }
        Double rowcount = statsProvider.getRowCount();
        if (rowcount != null) {
            statisticsValues.add(new StatisticsHolder<>(rowcount, ColumnStatisticsKind.ROWCOUNT));
        }
        return statisticsValues;
    }
    return Collections.emptyList();
}

Example 8 with StatisticsHolder

use of org.apache.drill.metastore.statistics.StatisticsHolder in project drill by apache.

the class TestMetastoreCommands method testSimpleAnalyze.

@Test
public void testSimpleAnalyze() throws Exception {
    String tableName = "multilevel/parquetSimpleAnalyze";
    TableInfo tableInfo = getTableInfo(tableName, "default");
    File table = dirTestWatcher.copyResourceToRoot(Paths.get("multilevel/parquet"), Paths.get(tableName));
    Path tablePath = new Path(table.toURI().getPath());
    BaseTableMetadata expectedTableMetadata = getBaseTableMetadata(tableInfo, table);
    TableInfo baseTableInfo = TableInfo.builder().name(tableName).storagePlugin("dfs").workspace("default").build();
    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(DIR0_1994_SEGMENT_COLUMN_STATISTICS).metadataStatistics(Collections.singletonList(new StatisticsHolder<>(40L, TableStatisticsKind.ROW_COUNT))).locations(ImmutableSet.of(new Path(tablePath, "1994/Q1/orders_94_q1.parquet"), new Path(tablePath, "1994/Q2/orders_94_q2.parquet"), new Path(tablePath, "1994/Q3/orders_94_q3.parquet"), new Path(tablePath, "1994/Q4/orders_94_q4.parquet"))).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.parquet"), new Path(tablePath, "1994/Q4/orders_94_q4.parquet"), new Path(tablePath, "1994/Q1/orders_94_q1.parquet"), new Path(tablePath, "1994/Q3/orders_94_q3.parquet"));
    expectedSegmentFilesLocations.add(segmentFiles);
    segmentFiles = ImmutableSet.of(new Path(tablePath, "1995/Q2/orders_95_q2.parquet"), new Path(tablePath, "1995/Q4/orders_95_q4.parquet"), new Path(tablePath, "1995/Q1/orders_95_q1.parquet"), new Path(tablePath, "1995/Q3/orders_95_q3.parquet"));
    expectedSegmentFilesLocations.add(segmentFiles);
    segmentFiles = ImmutableSet.of(new Path(tablePath, "1996/Q3/orders_96_q3.parquet"), new Path(tablePath, "1996/Q2/orders_96_q2.parquet"), new Path(tablePath, "1996/Q4/orders_96_q4.parquet"), new Path(tablePath, "1996/Q1/orders_96_q1.parquet"));
    expectedSegmentFilesLocations.add(segmentFiles);
    long dir0q1lastModified = new File(new File(new File(table, "1994"), "Q1"), "orders_94_q1.parquet").lastModified();
    FileMetadata dir01994q1File = FileMetadata.builder().tableInfo(baseTableInfo).metadataInfo(MetadataInfo.builder().type(MetadataType.FILE).identifier("1994/Q1/orders_94_q1.parquet").key("1994").build()).schema(SCHEMA).lastModifiedTime(dir0q1lastModified).columnsStatistics(DIR0_1994_Q1_SEGMENT_COLUMN_STATISTICS).metadataStatistics(Collections.singletonList(new StatisticsHolder<>(10L, TableStatisticsKind.ROW_COUNT))).path(new Path(tablePath, "1994/Q1/orders_94_q1.parquet")).build();
    RowGroupMetadata dir01994q1rowGroup = RowGroupMetadata.builder().tableInfo(baseTableInfo).metadataInfo(MetadataInfo.builder().type(MetadataType.ROW_GROUP).identifier("1994/Q1/orders_94_q1.parquet/0").key("1994").build()).schema(SCHEMA).rowGroupIndex(0).hostAffinity(Collections.emptyMap()).lastModifiedTime(dir0q1lastModified).columnsStatistics(DIR0_1994_Q1_SEGMENT_COLUMN_STATISTICS).metadataStatistics(Arrays.asList(new StatisticsHolder<>(10L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(1196L, new BaseStatisticsKind<>(ExactStatisticsConstants.LENGTH, true)), new StatisticsHolder<>(4L, new BaseStatisticsKind<>(ExactStatisticsConstants.START, true)))).path(new Path(tablePath, "1994/Q1/orders_94_q1.parquet")).build();
    try {
        testBuilder().sqlQuery("ANALYZE TABLE dfs.`%s` REFRESH METADATA", tableName).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(DIR0_1994_Q1_SEGMENT_COLUMN_STATISTICS).metadataStatistics(Collections.singletonList(new StatisticsHolder<>(10L, TableStatisticsKind.ROW_COUNT))).locations(ImmutableSet.of(new Path(tablePath, "1994/Q1/orders_94_q1.parquet"))).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.parquet")).findAny().orElse(null));
        // verify row groups metadata
        List<RowGroupMetadata> rowGroupsMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().rowGroupsMetadata(tableInfo, null, (String) null);
        assertEquals(12, rowGroupsMetadata.size());
        // verify first row group dir01994q1rowGroup
        assertEquals(dir01994q1rowGroup, rowGroupsMetadata.stream().filter(unit -> unit.getMetadataInfo().identifier().equals("1994/Q1/orders_94_q1.parquet/0")).findAny().orElse(null));
    } finally {
        run("analyze table dfs.`%s` drop metadata if exists", tableName);
    }
}

Example 9 with StatisticsHolder

use of org.apache.drill.metastore.statistics.StatisticsHolder in project drill by apache.

the class TestMetastoreCommands method testIncrementalAnalyzeWithMoreColumns.

@Test
public void testIncrementalAnalyzeWithMoreColumns() throws Exception {
    String tableName = "multilevel/parquetMoreColumns";
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel/parquet"), Paths.get(tableName));
    Path tablePath = new Path(table.toURI().getPath());
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    Map<SchemaPath, ColumnStatistics<?>> updatedTableColumnStatistics = new HashMap<>();
    SchemaPath orderStatusPath = SchemaPath.getSimplePath("o_orderstatus");
    SchemaPath orderDatePath = SchemaPath.getSimplePath("o_orderdate");
    SchemaPath dir0Path = SchemaPath.getSimplePath("dir0");
    SchemaPath dir1Path = SchemaPath.getSimplePath("dir1");
    updatedTableColumnStatistics.put(orderStatusPath, TABLE_COLUMN_STATISTICS.get(orderStatusPath));
    updatedTableColumnStatistics.put(dir0Path, TABLE_COLUMN_STATISTICS.get(dir0Path));
    updatedTableColumnStatistics.put(dir1Path, TABLE_COLUMN_STATISTICS.get(dir1Path));
    BaseTableMetadata expectedTableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(TABLE_META_INFO).schema(SCHEMA).location(tablePath).columnsStatistics(updatedTableColumnStatistics).metadataStatistics(Arrays.asList(new StatisticsHolder<>(120L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ROW_GROUP, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).interestingColumns(Collections.singletonList(orderStatusPath)).build();
    try {
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` columns(o_orderstatus) REFRESH METADATA 'row_group' LEVEL", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.tmp.%s]", tableName)).go();
        BaseTableMetadata actualTableMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().tableMetadata(tableInfo);
        assertEquals(expectedTableMetadata, actualTableMetadata);
        // checks that analyze was produced since interesting columns list differs, and second columns list isn't a sublist of previously analyzed table
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` columns(o_orderstatus, o_orderdate) REFRESH METADATA 'row_group' LEVEL", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.tmp.%s]", tableName)).go();
        actualTableMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().tableMetadata(tableInfo);
        updatedTableColumnStatistics.put(orderDatePath, TABLE_COLUMN_STATISTICS.get(orderDatePath));
        assertEquals(expectedTableMetadata.toBuilder().columnsStatistics(updatedTableColumnStatistics).interestingColumns(Arrays.asList(orderStatusPath, orderDatePath)).build(), actualTableMetadata);
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
    }
}

Example 10 with StatisticsHolder

use of org.apache.drill.metastore.statistics.StatisticsHolder in project drill by apache.

the class TestMetastoreCommands method testIncrementalAnalyzeWithEmptyColumns.

@Test
public void testIncrementalAnalyzeWithEmptyColumns() throws Exception {
    String tableName = "multilevel/parquetEmptyColumns";
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel/parquet"), Paths.get(tableName));
    Path tablePath = new Path(table.toURI().getPath());
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    Map<SchemaPath, ColumnStatistics<?>> updatedTableColumnStatistics = new HashMap<>();
    SchemaPath orderStatusPath = SchemaPath.getSimplePath("o_orderstatus");
    SchemaPath orderDatePath = SchemaPath.getSimplePath("o_orderdate");
    SchemaPath dir0Path = SchemaPath.getSimplePath("dir0");
    SchemaPath dir1Path = SchemaPath.getSimplePath("dir1");
    updatedTableColumnStatistics.put(orderStatusPath, TABLE_COLUMN_STATISTICS.get(orderStatusPath));
    updatedTableColumnStatistics.put(orderDatePath, TABLE_COLUMN_STATISTICS.get(orderDatePath));
    updatedTableColumnStatistics.put(dir0Path, TABLE_COLUMN_STATISTICS.get(dir0Path));
    updatedTableColumnStatistics.put(dir1Path, TABLE_COLUMN_STATISTICS.get(dir1Path));
    BaseTableMetadata expectedTableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(TABLE_META_INFO).schema(SCHEMA).location(tablePath).columnsStatistics(updatedTableColumnStatistics).metadataStatistics(Arrays.asList(new StatisticsHolder<>(120L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ROW_GROUP, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).interestingColumns(Arrays.asList(orderStatusPath, orderDatePath)).build();
    try {
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` columns(o_orderstatus, o_orderdate) REFRESH METADATA 'row_group' LEVEL", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.tmp.%s]", tableName)).go();
        BaseTableMetadata actualTableMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().tableMetadata(tableInfo);
        assertEquals(expectedTableMetadata, actualTableMetadata);
        // checks that analyze wasn't produced though interesting columns list differs, but it is a sublist of previously analyzed table
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` columns NONE REFRESH METADATA 'row_group' LEVEL", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(false, "Table metadata is up to date, analyze wasn't performed.").go();
        actualTableMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().tableMetadata(tableInfo);
        assertEquals(expectedTableMetadata, actualTableMetadata);
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
    }
}