Search in sources :

Example 1 with SqlMetastoreAnalyzeTable

use of org.apache.drill.exec.planner.sql.parser.SqlMetastoreAnalyzeTable in project drill by apache.

the class MetastoreAnalyzeTableHandler method convertToDrel.

/**
 * Converts to Drill logical plan
 */
private DrillRel convertToDrel(RelNode relNode, SqlMetastoreAnalyzeTable sqlAnalyzeTable, DrillTableInfo drillTableInfo) throws ForemanSetupException, IOException {
    RelBuilder relBuilder = LOGICAL_BUILDER.create(relNode.getCluster(), null);
    DrillTable table = drillTableInfo.drillTable();
    AnalyzeInfoProvider analyzeInfoProvider = table.getGroupScan().getAnalyzeInfoProvider();
    List<String> schemaPath = drillTableInfo.schemaPath();
    String pluginName = schemaPath.get(0);
    String workspaceName = Strings.join(schemaPath.subList(1, schemaPath.size()), AbstractSchema.SCHEMA_SEPARATOR);
    String tableName = drillTableInfo.tableName();
    TableInfo tableInfo = TableInfo.builder().name(tableName).owner(table.getUserName()).type(analyzeInfoProvider.getTableTypeName()).storagePlugin(pluginName).workspace(workspaceName).build();
    ColumnNamesOptions columnNamesOptions = new ColumnNamesOptions(context.getOptions());
    List<String> segmentColumns = analyzeInfoProvider.getSegmentColumns(table, columnNamesOptions).stream().map(SchemaPath::getRootSegmentPath).collect(Collectors.toList());
    List<NamedExpression> segmentExpressions = segmentColumns.stream().map(partitionName -> new NamedExpression(SchemaPath.getSimplePath(partitionName), FieldReference.getWithQuotedRef(partitionName))).collect(Collectors.toList());
    List<MetadataInfo> rowGroupsInfo = Collections.emptyList();
    List<MetadataInfo> filesInfo = Collections.emptyList();
    Multimap<Integer, MetadataInfo> segments = ArrayListMultimap.create();
    BasicTablesRequests basicRequests;
    try {
        basicRequests = context.getMetastoreRegistry().get().tables().basicRequests();
    } catch (MetastoreException e) {
        logger.error("Error when obtaining Metastore instance for table {}", tableName, e);
        DrillRel convertedRelNode = convertToRawDrel(relBuilder.values(new String[] { MetastoreAnalyzeConstants.OK_FIELD_NAME, MetastoreAnalyzeConstants.SUMMARY_FIELD_NAME }, false, e.getMessage()).build());
        return new DrillScreenRel(convertedRelNode.getCluster(), convertedRelNode.getTraitSet(), convertedRelNode);
    }
    MetadataType metadataLevel = getMetadataType(sqlAnalyzeTable);
    List<SchemaPath> interestingColumns = sqlAnalyzeTable.getFieldNames();
    MetastoreTableInfo metastoreTableInfo = basicRequests.metastoreTableInfo(tableInfo);
    List<MetadataInfo> allMetaToHandle = null;
    List<MetadataInfo> metadataToRemove = new ArrayList<>();
    // whether incremental analyze may be produced
    if (metastoreTableInfo.isExists()) {
        RelNode finalRelNode = relNode;
        CheckedSupplier<TableScan, SqlUnsupportedException> tableScanSupplier = () -> DrillRelOptUtil.findScan(convertToDrel(finalRelNode.getInput(0)));
        MetadataInfoCollector metadataInfoCollector = analyzeInfoProvider.getMetadataInfoCollector(basicRequests, tableInfo, (FormatSelection) table.getSelection(), context.getPlannerSettings(), tableScanSupplier, interestingColumns, metadataLevel, segmentColumns.size());
        if (!metadataInfoCollector.isOutdated()) {
            DrillRel convertedRelNode = convertToRawDrel(relBuilder.values(new String[] { MetastoreAnalyzeConstants.OK_FIELD_NAME, MetastoreAnalyzeConstants.SUMMARY_FIELD_NAME }, false, "Table metadata is up to date, analyze wasn't performed.").build());
            return new DrillScreenRel(convertedRelNode.getCluster(), convertedRelNode.getTraitSet(), convertedRelNode);
        }
        // updates scan to read updated / new files, pass removed files into metadata handler
        relNode = relNode.copy(relNode.getTraitSet(), Collections.singletonList(metadataInfoCollector.getPrunedScan()));
        filesInfo = metadataInfoCollector.getFilesInfo();
        segments = metadataInfoCollector.getSegmentsInfo();
        rowGroupsInfo = metadataInfoCollector.getRowGroupsInfo();
        allMetaToHandle = metadataInfoCollector.getAllMetaToHandle();
        metadataToRemove = metadataInfoCollector.getMetadataToRemove();
    }
    // Step 2: constructs plan for producing analyze
    DrillRel convertedRelNode = convertToRawDrel(relNode);
    boolean createNewAggregations = true;
    // List of columns for which statistics should be collected: interesting columns + segment columns
    List<SchemaPath> statisticsColumns = interestingColumns == null ? null : new ArrayList<>(interestingColumns);
    if (statisticsColumns != null) {
        segmentColumns.stream().map(SchemaPath::getSimplePath).forEach(statisticsColumns::add);
    }
    SchemaPath locationField = analyzeInfoProvider.getLocationField(columnNamesOptions);
    if (analyzeInfoProvider.supportsMetadataType(MetadataType.ROW_GROUP) && metadataLevel.includes(MetadataType.ROW_GROUP)) {
        MetadataHandlerContext handlerContext = MetadataHandlerContext.builder().tableInfo(tableInfo).metadataToHandle(rowGroupsInfo).metadataType(MetadataType.ROW_GROUP).depthLevel(segmentExpressions.size()).segmentColumns(segmentColumns).build();
        convertedRelNode = getRowGroupAggRelNode(segmentExpressions, convertedRelNode, createNewAggregations, statisticsColumns, handlerContext);
        createNewAggregations = false;
        locationField = SchemaPath.getSimplePath(MetastoreAnalyzeConstants.LOCATION_FIELD);
    }
    if (analyzeInfoProvider.supportsMetadataType(MetadataType.FILE) && metadataLevel.includes(MetadataType.FILE)) {
        MetadataHandlerContext handlerContext = MetadataHandlerContext.builder().tableInfo(tableInfo).metadataToHandle(filesInfo).metadataType(MetadataType.FILE).depthLevel(segmentExpressions.size()).segmentColumns(segmentColumns).build();
        convertedRelNode = getFileAggRelNode(segmentExpressions, convertedRelNode, createNewAggregations, statisticsColumns, locationField, handlerContext);
        locationField = SchemaPath.getSimplePath(MetastoreAnalyzeConstants.LOCATION_FIELD);
        createNewAggregations = false;
    }
    if (analyzeInfoProvider.supportsMetadataType(MetadataType.SEGMENT) && metadataLevel.includes(MetadataType.SEGMENT)) {
        for (int i = segmentExpressions.size(); i > 0; i--) {
            MetadataHandlerContext handlerContext = MetadataHandlerContext.builder().tableInfo(tableInfo).metadataToHandle(new ArrayList<>(segments.get(i - 1))).metadataType(MetadataType.SEGMENT).depthLevel(i).segmentColumns(segmentColumns.subList(0, i)).build();
            convertedRelNode = getSegmentAggRelNode(segmentExpressions, convertedRelNode, createNewAggregations, statisticsColumns, locationField, i, handlerContext);
            locationField = SchemaPath.getSimplePath(MetastoreAnalyzeConstants.LOCATION_FIELD);
            createNewAggregations = false;
        }
    }
    if (analyzeInfoProvider.supportsMetadataType(MetadataType.TABLE) && metadataLevel.includes(MetadataType.TABLE)) {
        MetadataHandlerContext handlerContext = MetadataHandlerContext.builder().tableInfo(tableInfo).metadataToHandle(Collections.emptyList()).metadataType(MetadataType.TABLE).depthLevel(segmentExpressions.size()).segmentColumns(segmentColumns).build();
        convertedRelNode = getTableAggRelNode(convertedRelNode, createNewAggregations, statisticsColumns, locationField, handlerContext);
    } else {
        throw new IllegalStateException("Analyze table with NONE level");
    }
    boolean useStatistics = context.getOptions().getOption(PlannerSettings.STATISTICS_USE);
    SqlNumericLiteral samplePercentLiteral = sqlAnalyzeTable.getSamplePercent();
    double samplePercent = samplePercentLiteral == null ? 100.0 : samplePercentLiteral.intValue(true);
    // Step 3: adds rel nodes for producing statistics analyze if required
    RelNode analyzeRel = useStatistics ? new DrillAnalyzeRel(convertedRelNode.getCluster(), convertedRelNode.getTraitSet(), convertToRawDrel(relNode), samplePercent) : convertToRawDrel(relBuilder.values(new String[] { "" }, "").build());
    MetadataControllerContext metadataControllerContext = MetadataControllerContext.builder().tableInfo(tableInfo).metastoreTableInfo(metastoreTableInfo).location(((FormatSelection) table.getSelection()).getSelection().getSelectionRoot()).interestingColumns(interestingColumns).segmentColumns(segmentColumns).metadataToHandle(allMetaToHandle).metadataToRemove(metadataToRemove).analyzeMetadataLevel(metadataLevel).build();
    convertedRelNode = new MetadataControllerRel(convertedRelNode.getCluster(), convertedRelNode.getTraitSet(), convertedRelNode, analyzeRel, metadataControllerContext);
    return new DrillScreenRel(convertedRelNode.getCluster(), convertedRelNode.getTraitSet(), convertedRelNode);
}
Also used : MetadataType(org.apache.drill.metastore.metadata.MetadataType) Arrays(java.util.Arrays) UserException(org.apache.drill.common.exceptions.UserException) DrillRel(org.apache.drill.exec.planner.logical.DrillRel) LoggerFactory(org.slf4j.LoggerFactory) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) ArrayListMultimap(org.apache.drill.shaded.guava.com.google.common.collect.ArrayListMultimap) MetadataHandlerContext(org.apache.drill.exec.metastore.analyze.MetadataHandlerContext) PhysicalOperator(org.apache.drill.exec.physical.base.PhysicalOperator) SqlNode(org.apache.calcite.sql.SqlNode) RelBuilder(org.apache.calcite.tools.RelBuilder) FieldReference(org.apache.drill.common.expression.FieldReference) Multimap(org.apache.drill.shaded.guava.com.google.common.collect.Multimap) SqlSelect(org.apache.calcite.sql.SqlSelect) BasicTablesRequests(org.apache.drill.metastore.components.tables.BasicTablesRequests) SchemaPath(org.apache.drill.common.expression.SchemaPath) MetastoreAnalyzeConstants(org.apache.drill.exec.metastore.analyze.MetastoreAnalyzeConstants) Collectors(java.util.stream.Collectors) List(java.util.List) ValidationException(org.apache.calcite.tools.ValidationException) ForemanSetupException(org.apache.drill.exec.work.foreman.ForemanSetupException) SqlNumericLiteral(org.apache.calcite.sql.SqlNumericLiteral) MetadataInfoCollector(org.apache.drill.exec.metastore.analyze.MetadataInfoCollector) ExecConstants(org.apache.drill.exec.ExecConstants) SqlMetastoreAnalyzeTable(org.apache.drill.exec.planner.sql.parser.SqlMetastoreAnalyzeTable) MetadataAggregateContext(org.apache.drill.exec.metastore.analyze.MetadataAggregateContext) TableScan(org.apache.calcite.rel.core.TableScan) TableInfo(org.apache.drill.metastore.metadata.TableInfo) MetastoreException(org.apache.drill.metastore.exceptions.MetastoreException) MetadataControllerContext(org.apache.drill.exec.metastore.analyze.MetadataControllerContext) Pointer(org.apache.drill.exec.util.Pointer) DrillTable(org.apache.drill.exec.planner.logical.DrillTable) ColumnNamesOptions(org.apache.drill.exec.metastore.ColumnNamesOptions) ArrayList(java.util.ArrayList) SqlLiteral(org.apache.calcite.sql.SqlLiteral) SqlUnsupportedException(org.apache.drill.exec.work.foreman.SqlUnsupportedException) NamedExpression(org.apache.drill.common.logical.data.NamedExpression) MetadataAggRel(org.apache.drill.exec.planner.logical.MetadataAggRel) DrillRelOptUtil(org.apache.drill.exec.planner.common.DrillRelOptUtil) FormatSelection(org.apache.drill.exec.store.dfs.FormatSelection) SqlIdentifier(org.apache.calcite.sql.SqlIdentifier) MetadataHandlerRel(org.apache.drill.exec.planner.logical.MetadataHandlerRel) CheckedSupplier(org.apache.drill.common.util.function.CheckedSupplier) RelDataType(org.apache.calcite.rel.type.RelDataType) SqlParserPos(org.apache.calcite.sql.parser.SqlParserPos) Logger(org.slf4j.Logger) DrillAnalyzeRel(org.apache.drill.exec.planner.logical.DrillAnalyzeRel) IOException(java.io.IOException) RelNode(org.apache.calcite.rel.RelNode) Prel(org.apache.drill.exec.planner.physical.Prel) AbstractSchema(org.apache.drill.exec.store.AbstractSchema) PlannerSettings(org.apache.drill.exec.planner.physical.PlannerSettings) RelConversionException(org.apache.calcite.tools.RelConversionException) Strings(org.apache.parquet.Strings) DrillScreenRel(org.apache.drill.exec.planner.logical.DrillScreenRel) PhysicalPlan(org.apache.drill.exec.physical.PhysicalPlan) LOGICAL_BUILDER(org.apache.drill.exec.planner.logical.DrillRelFactories.LOGICAL_BUILDER) SqlNodeList(org.apache.calcite.sql.SqlNodeList) MetadataControllerRel(org.apache.drill.exec.planner.logical.MetadataControllerRel) MetastoreTableInfo(org.apache.drill.metastore.components.tables.MetastoreTableInfo) Collections(java.util.Collections) AnalyzeInfoProvider(org.apache.drill.exec.metastore.analyze.AnalyzeInfoProvider) MetastoreTableInfo(org.apache.drill.metastore.components.tables.MetastoreTableInfo) DrillTable(org.apache.drill.exec.planner.logical.DrillTable) AnalyzeInfoProvider(org.apache.drill.exec.metastore.analyze.AnalyzeInfoProvider) ArrayList(java.util.ArrayList) FormatSelection(org.apache.drill.exec.store.dfs.FormatSelection) BasicTablesRequests(org.apache.drill.metastore.components.tables.BasicTablesRequests) DrillAnalyzeRel(org.apache.drill.exec.planner.logical.DrillAnalyzeRel) SchemaPath(org.apache.drill.common.expression.SchemaPath) SqlUnsupportedException(org.apache.drill.exec.work.foreman.SqlUnsupportedException) ColumnNamesOptions(org.apache.drill.exec.metastore.ColumnNamesOptions) MetadataHandlerContext(org.apache.drill.exec.metastore.analyze.MetadataHandlerContext) TableInfo(org.apache.drill.metastore.metadata.TableInfo) MetastoreTableInfo(org.apache.drill.metastore.components.tables.MetastoreTableInfo) MetadataControllerRel(org.apache.drill.exec.planner.logical.MetadataControllerRel) MetadataInfo(org.apache.drill.metastore.metadata.MetadataInfo) TableScan(org.apache.calcite.rel.core.TableScan) RelBuilder(org.apache.calcite.tools.RelBuilder) MetastoreException(org.apache.drill.metastore.exceptions.MetastoreException) MetadataType(org.apache.drill.metastore.metadata.MetadataType) DrillScreenRel(org.apache.drill.exec.planner.logical.DrillScreenRel) RelNode(org.apache.calcite.rel.RelNode) NamedExpression(org.apache.drill.common.logical.data.NamedExpression) DrillRel(org.apache.drill.exec.planner.logical.DrillRel) MetadataInfoCollector(org.apache.drill.exec.metastore.analyze.MetadataInfoCollector) MetadataControllerContext(org.apache.drill.exec.metastore.analyze.MetadataControllerContext) SqlNumericLiteral(org.apache.calcite.sql.SqlNumericLiteral)

Example 2 with SqlMetastoreAnalyzeTable

use of org.apache.drill.exec.planner.sql.parser.SqlMetastoreAnalyzeTable in project drill by apache.

the class MetastoreAnalyzeTableHandler method getPlan.

@Override
public PhysicalPlan getPlan(SqlNode sqlNode) throws ValidationException, RelConversionException, IOException, ForemanSetupException {
    if (!context.getOptions().getOption(ExecConstants.METASTORE_ENABLED_VALIDATOR)) {
        throw UserException.validationError().message("Running ANALYZE TABLE REFRESH METADATA command when Metastore is disabled (`metastore.enabled` is set to false)").build(logger);
    }
    // disables during analyze to prevent using data about locations from the Metastore
    context.getOptions().setLocalOption(ExecConstants.METASTORE_ENABLED, false);
    SqlMetastoreAnalyzeTable sqlAnalyzeTable = unwrap(sqlNode, SqlMetastoreAnalyzeTable.class);
    SqlNode tableRef = sqlAnalyzeTable.getTableRef();
    DrillTableInfo drillTableInfo = DrillTableInfo.getTableInfoHolder(tableRef, config);
    AnalyzeInfoProvider analyzeInfoProvider = drillTableInfo.drillTable().getGroupScan().getAnalyzeInfoProvider();
    if (analyzeInfoProvider == null) {
        throw UserException.validationError().message("ANALYZE is not supported for group scan [%s]", drillTableInfo.drillTable().getGroupScan()).build(logger);
    }
    ColumnNamesOptions columnNamesOptions = new ColumnNamesOptions(context.getOptions());
    // creates select with DYNAMIC_STAR column and analyze specific columns to obtain corresponding table scan
    SqlSelect scanSql = new SqlSelect(SqlParserPos.ZERO, SqlNodeList.EMPTY, getColumnList(analyzeInfoProvider.getProjectionFields(drillTableInfo.drillTable(), getMetadataType(sqlAnalyzeTable), columnNamesOptions)), tableRef, null, null, null, null, null, null, null);
    ConvertedRelNode convertedRelNode = validateAndConvert(rewrite(scanSql));
    RelDataType validatedRowType = convertedRelNode.getValidatedRowType();
    RelNode relScan = convertedRelNode.getConvertedNode();
    DrillRel drel = convertToDrel(relScan, sqlAnalyzeTable, drillTableInfo);
    Prel prel = convertToPrel(drel, validatedRowType);
    logAndSetTextPlan("Drill Physical", prel, logger);
    PhysicalOperator pop = convertToPop(prel);
    PhysicalPlan plan = convertToPlan(pop);
    log("Drill Plan", plan, logger);
    return plan;
}
Also used : PhysicalPlan(org.apache.drill.exec.physical.PhysicalPlan) AnalyzeInfoProvider(org.apache.drill.exec.metastore.analyze.AnalyzeInfoProvider) RelDataType(org.apache.calcite.rel.type.RelDataType) Prel(org.apache.drill.exec.planner.physical.Prel) SqlSelect(org.apache.calcite.sql.SqlSelect) RelNode(org.apache.calcite.rel.RelNode) SqlMetastoreAnalyzeTable(org.apache.drill.exec.planner.sql.parser.SqlMetastoreAnalyzeTable) PhysicalOperator(org.apache.drill.exec.physical.base.PhysicalOperator) ColumnNamesOptions(org.apache.drill.exec.metastore.ColumnNamesOptions) DrillRel(org.apache.drill.exec.planner.logical.DrillRel) SqlNode(org.apache.calcite.sql.SqlNode)

Aggregations

RelNode (org.apache.calcite.rel.RelNode)2 RelDataType (org.apache.calcite.rel.type.RelDataType)2 SqlNode (org.apache.calcite.sql.SqlNode)2 SqlSelect (org.apache.calcite.sql.SqlSelect)2 ColumnNamesOptions (org.apache.drill.exec.metastore.ColumnNamesOptions)2 AnalyzeInfoProvider (org.apache.drill.exec.metastore.analyze.AnalyzeInfoProvider)2 PhysicalPlan (org.apache.drill.exec.physical.PhysicalPlan)2 PhysicalOperator (org.apache.drill.exec.physical.base.PhysicalOperator)2 DrillRel (org.apache.drill.exec.planner.logical.DrillRel)2 IOException (java.io.IOException)1 ArrayList (java.util.ArrayList)1 Arrays (java.util.Arrays)1 Collections (java.util.Collections)1 List (java.util.List)1 Collectors (java.util.stream.Collectors)1 TableScan (org.apache.calcite.rel.core.TableScan)1 SqlIdentifier (org.apache.calcite.sql.SqlIdentifier)1 SqlLiteral (org.apache.calcite.sql.SqlLiteral)1 SqlNodeList (org.apache.calcite.sql.SqlNodeList)1 SqlNumericLiteral (org.apache.calcite.sql.SqlNumericLiteral)1