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

Example 21 with ColumnStatistics

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

the class TestMetastoreCommands method testAnalyzeEmptyRequiredParquetTable.

@Test
public void testAnalyzeEmptyRequiredParquetTable() throws Exception {
    String tableName = "analyze_empty_simple_required";
    run("create table dfs.tmp.%s as select 1 as `date`, 'a' as name from (values(1)) where 1 = 2", tableName);
    File table = new File(dirTestWatcher.getDfsTestTmpDir(), tableName);
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    TupleMetadata schema = new SchemaBuilder().add("date", TypeProtos.MinorType.INT).add("name", TypeProtos.MinorType.VARCHAR).build();
    Map<SchemaPath, ColumnStatistics<?>> columnStatistics = ImmutableMap.<SchemaPath, ColumnStatistics<?>>builder().put(SchemaPath.getSimplePath("name"), getColumnStatistics(null, null, 0L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getSimplePath("date"), getColumnStatistics(null, null, 0L, TypeProtos.MinorType.INT)).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<>(0L, 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(1, filesMetadata.size());
        List<RowGroupMetadata> rowGroupsMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().rowGroupsMetadata(tableInfo, (String) null, null);
        assertEquals(1, rowGroupsMetadata.size());
        testBuilder().sqlQuery("select COLUMN_NAME from INFORMATION_SCHEMA.`COLUMNS` where table_name='%s'", tableName).unOrdered().baselineColumns("COLUMN_NAME").baselineValues("date").baselineValues("name").go();
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
        run("drop table if exists dfs.tmp.`%s`", tableName);
    }
}

Example 22 with ColumnStatistics

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

the class TestMetastoreCommands method testIncrementalAnalyzeUpdatedFile.

@Test
public void testIncrementalAnalyzeUpdatedFile() throws Exception {
    String tableName = "multilevel/parquetUpdatedFile";
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel/parquet"), Paths.get(tableName));
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    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());
        List<FileMetadata> filesMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().filesMetadata(tableInfo, null, null);
        assertEquals(12, filesMetadata.size());
        List<RowGroupMetadata> rowGroupsMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().rowGroupsMetadata(tableInfo, null, (String) null);
        assertEquals(12, rowGroupsMetadata.size());
        File fileToUpdate = new File(new File(new File(table, "1994"), "Q4"), "orders_94_q4.parquet");
        long lastModified = fileToUpdate.lastModified();
        FileUtils.deleteQuietly(fileToUpdate);
        // replaces original file
        dirTestWatcher.copyResourceToTestTmp(Paths.get("multilevel", "parquet", "1994", "Q1", "orders_94_q1.parquet"), Paths.get(tableName, "1994", "Q4", "orders_94_q4.parquet"));
        long newLastModified = lastModified + 1000;
        assertTrue(fileToUpdate.setLastModified(newLastModified));
        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);
        Map<SchemaPath, ColumnStatistics<?>> tableColumnStatistics = new HashMap<>(TABLE_COLUMN_STATISTICS);
        tableColumnStatistics.computeIfPresent(SchemaPath.getSimplePath("o_clerk"), (logicalExpressions, columnStatistics) -> columnStatistics.cloneWith(new ColumnStatistics<>(Collections.singletonList(new StatisticsHolder<>("Clerk#000000006", ColumnStatisticsKind.MIN_VALUE)))));
        tableColumnStatistics.computeIfPresent(SchemaPath.getSimplePath("o_totalprice"), (logicalExpressions, columnStatistics) -> columnStatistics.cloneWith(new ColumnStatistics<>(Collections.singletonList(new StatisticsHolder<>(328207.15, ColumnStatisticsKind.MAX_VALUE)))));
        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(newLastModified).build();
        assertEquals(expectedTableMetadata, actualTableMetadata);
        segmentMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().segmentsMetadataByMetadataKey(tableInfo, null, null);
        assertEquals(15, segmentMetadata.size());
        filesMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().filesMetadata(tableInfo, null, null);
        assertEquals(12, filesMetadata.size());
        rowGroupsMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().rowGroupsMetadata(tableInfo, null, (String) null);
        assertEquals(12, rowGroupsMetadata.size());
    } finally {
        run("analyze table dfs.tmp.`%s` drop metadata if exists", tableName);
        FileUtils.deleteQuietly(table);
    }
}

Example 23 with ColumnStatistics

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

the class TestMetastoreCommands method testAnalyzeEmptyNullableParquetTable.

@Test
public void testAnalyzeEmptyNullableParquetTable() throws Exception {
    File table = dirTestWatcher.copyResourceToRoot(Paths.get("parquet", "empty", "simple", "empty_simple.parquet"));
    String tableName = "parquet/empty/simple/empty_simple.parquet";
    TableInfo tableInfo = getTableInfo(tableName, "default");
    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(null, null, 0L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getSimplePath("id"), getColumnStatistics(null, null, 0L, 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<>(0L, TableStatisticsKind.ROW_COUNT), new StatisticsHolder<>(MetadataType.ALL, TableStatisticsKind.ANALYZE_METADATA_LEVEL))).partitionKeys(Collections.emptyMap()).lastModifiedTime(getMaxLastModified(table)).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();
        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(1, filesMetadata.size());
        List<RowGroupMetadata> rowGroupsMetadata = cluster.drillbit().getContext().getMetastoreRegistry().get().tables().basicRequests().rowGroupsMetadata(tableInfo, (String) null, null);
        assertEquals(1, rowGroupsMetadata.size());
    } finally {
        run("analyze table dfs.`%s` drop metadata if exists", tableName);
    }
}

Example 24 with ColumnStatistics

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

the class TestMetastoreCommands method testAnalyzeWithMapColumns.

@Test
public void testAnalyzeWithMapColumns() throws Exception {
    String tableName = "complex";
    TableInfo tableInfo = getTableInfo(tableName, "tmp");
    File table = dirTestWatcher.copyResourceToTestTmp(Paths.get("store/parquet/complex/complex.parquet"), Paths.get(tableName));
    TupleMetadata schema = new SchemaBuilder().addNullable("trans_id", TypeProtos.MinorType.BIGINT).addNullable("date", TypeProtos.MinorType.VARCHAR).addNullable("time", TypeProtos.MinorType.VARCHAR).addNullable("amount", TypeProtos.MinorType.FLOAT8).addMap("user_info").addNullable("cust_id", TypeProtos.MinorType.BIGINT).addNullable("device", TypeProtos.MinorType.VARCHAR).addNullable("state", TypeProtos.MinorType.VARCHAR).resumeSchema().addMap("marketing_info").addNullable("camp_id", TypeProtos.MinorType.BIGINT).addArray("keywords", TypeProtos.MinorType.VARCHAR).resumeSchema().addMap("trans_info").addArray("prod_id", TypeProtos.MinorType.BIGINT).addNullable("purch_flag", TypeProtos.MinorType.VARCHAR).resumeSchema().build();
    Map<SchemaPath, ColumnStatistics<?>> columnStatistics = ImmutableMap.<SchemaPath, ColumnStatistics<?>>builder().put(SchemaPath.getCompoundPath("user_info", "state"), getColumnStatistics("ct", "nj", 5L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getSimplePath("date"), getColumnStatistics("2013-05-16", "2013-07-26", 5L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getSimplePath("time"), getColumnStatistics("04:56:59", "15:31:45", 5L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getCompoundPath("user_info", "cust_id"), getColumnStatistics(11L, 86623L, 5L, TypeProtos.MinorType.BIGINT)).put(SchemaPath.getSimplePath("amount"), getColumnStatistics(20.25, 500.75, 5L, TypeProtos.MinorType.FLOAT8)).put(SchemaPath.getCompoundPath("user_info", "device"), getColumnStatistics("AOS4.2", "IOS7", 5L, TypeProtos.MinorType.VARCHAR)).put(SchemaPath.getCompoundPath("marketing_info", "camp_id"), getColumnStatistics(4L, 17L, 5L, TypeProtos.MinorType.BIGINT)).put(SchemaPath.getSimplePath("trans_id"), getColumnStatistics(0L, 4L, 5L, TypeProtos.MinorType.BIGINT)).put(SchemaPath.getCompoundPath("trans_info", "purch_flag"), getColumnStatistics("false", "true", 5L, TypeProtos.MinorType.VARCHAR)).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<>(5L, 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();
        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 25 with ColumnStatistics

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

the class TestMetastoreCommands method testIncrementalAnalyzeWithFewerColumns.

@Test
public void testIncrementalAnalyzeWithFewerColumns() throws Exception {
    String tableName = "multilevel/parquetFewerColumns";
    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(o_orderstatus) 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);
    }
}