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

Example 26 with ColumnStatistics

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

the class TestMetastoreCommands method testAnalyzeNonEmptyTableWithEmptyFile.

@Test
public void testAnalyzeNonEmptyTableWithEmptyFile() throws Exception {
    String tableName = "parquet_with_empty_file";
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("parquet", "empty", "simple"), Paths.get(tableName));
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    TupleMetadata schema = new SchemaBuilder().addNullable("id", TypeProtos.MinorType.BIGINT).addNullable("name", TypeProtos.MinorType.VARCHAR).build();
    Map<SchemaPath, ColumnStatistics<?>> columnStatistics = ImmutableMap.<SchemaPath, ColumnStatistics<?>>builder().put(SchemaPath.getSimplePath("name"), getColumnStatistics("Tom", "Tom", 1L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getSimplePath("id"), getColumnStatistics(2L, 2L, 1L, TypeProtos.MinorType.BIGINT)).build();
    BaseTableMetadata expectedTableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(TABLE_META_INFO).schema(schema).location(new Path(table.toURI().getPath())).columnsStatistics(columnStatistics).metadataStatistics(Arrays.asList(new StatisticsHolder<>(1L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ALL, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).build();
    try {
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` REFRESH METADATA", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.tmp.%s]", tableName)).go();
        MetastoreTableInfo metastoreTableInfo = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().metastoreTableInfo(tableInfo);
        assertTrue("table metadata wasn't found", metastoreTableInfo.isExists());
        BaseTableMetadata tableMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().tableMetadata(tableInfo);
        assertEquals(expectedTableMetadata, tableMetadata);
        List<FileMetadata> filesMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().filesMetadata(tableInfo, null, null);
        assertEquals(2, filesMetadata.size());
        List<RowGroupMetadata> rowGroupsMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().rowGroupsMetadata(tableInfo, (String) null, null);
        assertEquals(2, rowGroupsMetadata.size());
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
    }
}

Example 27 with ColumnStatistics

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

the class TestMetastoreCommands method testIncrementalAnalyzeNewChildSegment.

@Test
public void testIncrementalAnalyzeNewChildSegment() throws Exception {
    String tableName = "multilevel/parquetNewChildSegment";
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel/parquet"), Paths.get(tableName));
    Path tablePath = new Path(table.toURI().getPath());
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    // updates statistics values due to new segment
    Map<SchemaPath, ColumnStatistics<?>> updatedStatistics = new HashMap<>(TABLE_COLUMN_STATISTICS);
    updatedStatistics.replaceAll((logicalExpressions, columnStatistics) -> columnStatistics.cloneWith(new ColumnStatistics<>(Arrays.asList(new StatisticsHolder<>(130L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(130L, ColumnStatisticsKind.NON_NULL_VALUES_COUNT)))));
    updatedStatistics.computeIfPresent(SchemaPath.getSimplePath("dir1"), (logicalExpressions, columnStatistics) -> columnStatistics.cloneWith(new ColumnStatistics<>(Collections.singletonList(new StatisticsHolder<>("Q5", ColumnStatisticsKind.MAX_VALUE)))));
    BaseTableMetadata expectedTableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(TABLE_META_INFO).schema(SCHEMA).location(tablePath).columnsStatistics(updatedStatistics).metadataStatistics(Arrays.asList(new StatisticsHolder<>(130L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ALL, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).build();
    try {
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` REFRESH METADATA", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.tmp.%s]", tableName)).go();
        List<SegmentMetadata> segmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByMetadataKey(tableInfo, null, null);
        assertEquals(15, segmentMetadata.size());
        dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel", "parquet", "1994", "Q4"), Paths.get(tableName, "1994", "Q5"));
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` REFRESH METADATA", 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);
        segmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByMetadataKey(tableInfo, null, null);
        assertEquals(16, segmentMetadata.size());
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
        FileUtils.deleteQuietly(table);
    }
}

Example 28 with ColumnStatistics

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

the class TestMetastoreCommands method testIncrementalAnalyzeNewParentSegment.

@Test
public void testIncrementalAnalyzeNewParentSegment() throws Exception {
    String tableName = "multilevel/parquetNewParentSegment";
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel/parquet"), Paths.get(tableName));
    Path tablePath = new Path(table.toURI().getPath());
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    // updates statistics values due to new segment
    Map<SchemaPath, ColumnStatistics<?>> updatedStatistics = new HashMap<>(TABLE_COLUMN_STATISTICS);
    updatedStatistics.replaceAll((logicalExpressions, columnStatistics) -> columnStatistics.cloneWith(new ColumnStatistics<>(Arrays.asList(new StatisticsHolder<>(160L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(160L, ColumnStatisticsKind.NON_NULL_VALUES_COUNT)))));
    updatedStatistics.computeIfPresent(SchemaPath.getSimplePath("dir0"), (logicalExpressions, columnStatistics) -> columnStatistics.cloneWith(new ColumnStatistics<>(Collections.singletonList(new StatisticsHolder<>("1993", ColumnStatisticsKind.MIN_VALUE)))));
    BaseTableMetadata expectedTableMetadata = BaseTableMetadata.builder().tableInfo(tableInfo).metadataInfo(TABLE_META_INFO).schema(SCHEMA).location(tablePath).columnsStatistics(updatedStatistics).metadataStatistics(Arrays.asList(new StatisticsHolder<>(160L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ALL, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).build();
    try {
        assertEquals(0, cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByMetadataKey(tableInfo, null, null).size());
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` REFRESH METADATA", tableName).unOrdered().baselineColumns("ok", "summary").baselineValues(true, String.format("Collected / refreshed metadata for table [dfs.tmp.%s]", tableName)).go();
        List<SegmentMetadata> segmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByMetadataKey(tableInfo, null, null);
        assertEquals(15, segmentMetadata.size());
        dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel/parquet", "1994"), Paths.get(tableName, "1993"));
        testBuilder().sqlQuery("ANALYZE TABLE dfs.tmp.`%s` REFRESH METADATA", 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);
        segmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByMetadataKey(tableInfo, null, null);
        assertEquals(20, segmentMetadata.size());
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
        FileUtils.deleteQuietly(table);
    }
}

Example 29 with ColumnStatistics

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

the class TestMetastoreCommands method testAnalyzeWithColumns.

@Test
public void testAnalyzeWithColumns() throws Exception {
    String tableName = "multilevel/parquetColumns";
    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 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);
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
    }
}

Example 30 with ColumnStatistics

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

the class MetastoreFileTableMetadataProvider method getNonInterestingColumnsMetadata.

@Override
public NonInterestingColumnsMetadata getNonInterestingColumnsMetadata() {
    throwIfChanged();
    if (nonInterestingColumnsMetadata == null) {
        TupleMetadata schema = getTableMetadata().getSchema();
        List<StatisticsHolder<?>> statistics = Collections.singletonList(new StatisticsHolder<>(Statistic.NO_COLUMN_STATS, ColumnStatisticsKind.NULLS_COUNT));
        List<SchemaPath> columnPaths = SchemaUtil.getSchemaPaths(schema);
        List<SchemaPath> interestingColumns = getInterestingColumns(columnPaths);
        // populates statistics for non-interesting columns and columns for which statistics wasn't collected
        Map<SchemaPath, ColumnStatistics<?>> columnsStatistics = columnPaths.stream().filter(schemaPath -> !interestingColumns.contains(schemaPath) || SchemaPathUtils.getColumnMetadata(schemaPath, schema).isArray()).collect(Collectors.toMap(Function.identity(), schemaPath -> new ColumnStatistics<>(statistics, SchemaPathUtils.getColumnMetadata(schemaPath, schema).type())));
        nonInterestingColumnsMetadata = new NonInterestingColumnsMetadata(columnsStatistics);
    }
    return nonInterestingColumnsMetadata;
}