Search in sources :

Example 1 with StatisticAggregations

use of io.trino.sql.planner.plan.StatisticAggregations in project trino by trinodb.

the class PruneTableWriterSourceColumns method apply.

@Override
public Result apply(TableWriterNode tableWriterNode, Captures captures, Context context) {
    ImmutableSet.Builder<Symbol> requiredInputs = ImmutableSet.<Symbol>builder().addAll(tableWriterNode.getColumns());
    if (tableWriterNode.getPartitioningScheme().isPresent()) {
        PartitioningScheme partitioningScheme = tableWriterNode.getPartitioningScheme().get();
        partitioningScheme.getPartitioning().getColumns().forEach(requiredInputs::add);
        partitioningScheme.getHashColumn().ifPresent(requiredInputs::add);
    }
    if (tableWriterNode.getStatisticsAggregation().isPresent()) {
        StatisticAggregations aggregations = tableWriterNode.getStatisticsAggregation().get();
        requiredInputs.addAll(aggregations.getGroupingSymbols());
        aggregations.getAggregations().values().stream().map(SymbolsExtractor::extractUnique).forEach(requiredInputs::addAll);
    }
    return restrictChildOutputs(context.getIdAllocator(), tableWriterNode, requiredInputs.build()).map(Result::ofPlanNode).orElse(Result.empty());
}
Also used : ImmutableSet(com.google.common.collect.ImmutableSet) Symbol(io.trino.sql.planner.Symbol) PartitioningScheme(io.trino.sql.planner.PartitioningScheme) StatisticAggregations(io.trino.sql.planner.plan.StatisticAggregations)

Example 2 with StatisticAggregations

use of io.trino.sql.planner.plan.StatisticAggregations in project trino by trinodb.

the class LogicalPlanner method createAnalyzePlan.

private RelationPlan createAnalyzePlan(Analysis analysis, Analyze analyzeStatement) {
    TableHandle targetTable = analysis.getAnalyzeTarget().orElseThrow();
    // Plan table scan
    Map<String, ColumnHandle> columnHandles = metadata.getColumnHandles(session, targetTable);
    ImmutableList.Builder<Symbol> tableScanOutputs = ImmutableList.builder();
    ImmutableMap.Builder<Symbol, ColumnHandle> symbolToColumnHandle = ImmutableMap.builder();
    ImmutableMap.Builder<String, Symbol> columnNameToSymbol = ImmutableMap.builder();
    TableMetadata tableMetadata = metadata.getTableMetadata(session, targetTable);
    for (ColumnMetadata column : tableMetadata.getColumns()) {
        Symbol symbol = symbolAllocator.newSymbol(column.getName(), column.getType());
        tableScanOutputs.add(symbol);
        symbolToColumnHandle.put(symbol, columnHandles.get(column.getName()));
        columnNameToSymbol.put(column.getName(), symbol);
    }
    TableStatisticsMetadata tableStatisticsMetadata = metadata.getStatisticsCollectionMetadata(session, targetTable.getCatalogName().getCatalogName(), tableMetadata.getMetadata());
    TableStatisticAggregation tableStatisticAggregation = statisticsAggregationPlanner.createStatisticsAggregation(tableStatisticsMetadata, columnNameToSymbol.buildOrThrow());
    StatisticAggregations statisticAggregations = tableStatisticAggregation.getAggregations();
    List<Symbol> groupingSymbols = statisticAggregations.getGroupingSymbols();
    PlanNode planNode = new StatisticsWriterNode(idAllocator.getNextId(), new AggregationNode(idAllocator.getNextId(), TableScanNode.newInstance(idAllocator.getNextId(), targetTable, tableScanOutputs.build(), symbolToColumnHandle.buildOrThrow(), false, Optional.empty()), statisticAggregations.getAggregations(), singleGroupingSet(groupingSymbols), ImmutableList.of(), AggregationNode.Step.SINGLE, Optional.empty(), Optional.empty()), new StatisticsWriterNode.WriteStatisticsReference(targetTable), symbolAllocator.newSymbol("rows", BIGINT), tableStatisticsMetadata.getTableStatistics().contains(ROW_COUNT), tableStatisticAggregation.getDescriptor());
    return new RelationPlan(planNode, analysis.getScope(analyzeStatement), planNode.getOutputSymbols(), Optional.empty());
}
Also used : ConnectorTableMetadata(io.trino.spi.connector.ConnectorTableMetadata) TableMetadata(io.trino.metadata.TableMetadata) ColumnHandle(io.trino.spi.connector.ColumnHandle) TableStatisticsMetadata(io.trino.spi.statistics.TableStatisticsMetadata) ColumnMetadata(io.trino.spi.connector.ColumnMetadata) TableStatisticAggregation(io.trino.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation) ImmutableList.toImmutableList(com.google.common.collect.ImmutableList.toImmutableList) ImmutableList(com.google.common.collect.ImmutableList) StatisticsWriterNode(io.trino.sql.planner.plan.StatisticsWriterNode) AggregationNode(io.trino.sql.planner.plan.AggregationNode) ImmutableMap.toImmutableMap(com.google.common.collect.ImmutableMap.toImmutableMap) ImmutableMap(com.google.common.collect.ImmutableMap) StatisticAggregations(io.trino.sql.planner.plan.StatisticAggregations) PlanNode(io.trino.sql.planner.plan.PlanNode) TableHandle(io.trino.metadata.TableHandle)

Example 3 with StatisticAggregations

use of io.trino.sql.planner.plan.StatisticAggregations in project trino by trinodb.

the class LogicalPlanner method createTableWriterPlan.

private RelationPlan createTableWriterPlan(Analysis analysis, PlanNode source, List<Symbol> symbols, WriterTarget target, List<String> columnNames, List<ColumnMetadata> columnMetadataList, Optional<TableLayout> writeTableLayout, TableStatisticsMetadata statisticsMetadata) {
    Optional<PartitioningScheme> partitioningScheme = Optional.empty();
    Optional<PartitioningScheme> preferredPartitioningScheme = Optional.empty();
    if (writeTableLayout.isPresent()) {
        List<Symbol> partitionFunctionArguments = new ArrayList<>();
        writeTableLayout.get().getPartitionColumns().stream().mapToInt(columnNames::indexOf).mapToObj(symbols::get).forEach(partitionFunctionArguments::add);
        List<Symbol> outputLayout = new ArrayList<>(symbols);
        Optional<PartitioningHandle> partitioningHandle = writeTableLayout.get().getPartitioning();
        if (partitioningHandle.isPresent()) {
            partitioningScheme = Optional.of(new PartitioningScheme(Partitioning.create(partitioningHandle.get(), partitionFunctionArguments), outputLayout));
        } else {
            // empty connector partitioning handle means evenly partitioning on partitioning columns
            preferredPartitioningScheme = Optional.of(new PartitioningScheme(Partitioning.create(FIXED_HASH_DISTRIBUTION, partitionFunctionArguments), outputLayout));
        }
    }
    verify(columnNames.size() == symbols.size(), "columnNames.size() != symbols.size(): %s and %s", columnNames, symbols);
    Map<String, Symbol> columnToSymbolMap = zip(columnNames.stream(), symbols.stream(), SimpleImmutableEntry::new).collect(toImmutableMap(Entry::getKey, Entry::getValue));
    Set<Symbol> notNullColumnSymbols = columnMetadataList.stream().filter(column -> !column.isNullable()).map(ColumnMetadata::getName).map(columnToSymbolMap::get).collect(toImmutableSet());
    if (!statisticsMetadata.isEmpty()) {
        TableStatisticAggregation result = statisticsAggregationPlanner.createStatisticsAggregation(statisticsMetadata, columnToSymbolMap);
        StatisticAggregations.Parts aggregations = result.getAggregations().createPartialAggregations(symbolAllocator, plannerContext);
        // partial aggregation is run within the TableWriteOperator to calculate the statistics for
        // the data consumed by the TableWriteOperator
        // final aggregation is run within the TableFinishOperator to summarize collected statistics
        // by the partial aggregation from all of the writer nodes
        StatisticAggregations partialAggregation = aggregations.getPartialAggregation();
        TableFinishNode commitNode = new TableFinishNode(idAllocator.getNextId(), new TableWriterNode(idAllocator.getNextId(), source, target, symbolAllocator.newSymbol("partialrows", BIGINT), symbolAllocator.newSymbol("fragment", VARBINARY), symbols, columnNames, notNullColumnSymbols, partitioningScheme, preferredPartitioningScheme, Optional.of(partialAggregation), Optional.of(result.getDescriptor().map(aggregations.getMappings()::get))), target, symbolAllocator.newSymbol("rows", BIGINT), Optional.of(aggregations.getFinalAggregation()), Optional.of(result.getDescriptor()));
        return new RelationPlan(commitNode, analysis.getRootScope(), commitNode.getOutputSymbols(), Optional.empty());
    }
    TableFinishNode commitNode = new TableFinishNode(idAllocator.getNextId(), new TableWriterNode(idAllocator.getNextId(), source, target, symbolAllocator.newSymbol("partialrows", BIGINT), symbolAllocator.newSymbol("fragment", VARBINARY), symbols, columnNames, notNullColumnSymbols, partitioningScheme, preferredPartitioningScheme, Optional.empty(), Optional.empty()), target, symbolAllocator.newSymbol("rows", BIGINT), Optional.empty(), Optional.empty());
    return new RelationPlan(commitNode, analysis.getRootScope(), commitNode.getOutputSymbols(), Optional.empty());
}
Also used : FIXED_HASH_DISTRIBUTION(io.trino.sql.planner.SystemPartitioningHandle.FIXED_HASH_DISTRIBUTION) CostProvider(io.trino.cost.CostProvider) TypeSignatureProvider.fromTypes(io.trino.sql.analyzer.TypeSignatureProvider.fromTypes) Delete(io.trino.sql.tree.Delete) InsertReference(io.trino.sql.planner.plan.TableWriterNode.InsertReference) PlanNode(io.trino.sql.planner.plan.PlanNode) NOT_SUPPORTED(io.trino.spi.StandardErrorCode.NOT_SUPPORTED) LongLiteral(io.trino.sql.tree.LongLiteral) Map(java.util.Map) OutputNode(io.trino.sql.planner.plan.OutputNode) TableScanNode(io.trino.sql.planner.plan.TableScanNode) ExplainAnalyzeNode(io.trino.sql.planner.plan.ExplainAnalyzeNode) ExplainAnalyze(io.trino.sql.tree.ExplainAnalyze) Statement(io.trino.sql.tree.Statement) SystemSessionProperties.isCollectPlanStatisticsForAllQueries(io.trino.SystemSessionProperties.isCollectPlanStatisticsForAllQueries) StatisticsWriterNode(io.trino.sql.planner.plan.StatisticsWriterNode) TableExecute(io.trino.sql.tree.TableExecute) ImmutableList.toImmutableList(com.google.common.collect.ImmutableList.toImmutableList) Assignments(io.trino.sql.planner.plan.Assignments) Set(java.util.Set) PlanSanityChecker(io.trino.sql.planner.sanity.PlanSanityChecker) RelationType(io.trino.sql.analyzer.RelationType) ImmutableMap.toImmutableMap(com.google.common.collect.ImmutableMap.toImmutableMap) PlanBuilder.newPlanBuilder(io.trino.sql.planner.PlanBuilder.newPlanBuilder) CachingStatsProvider(io.trino.cost.CachingStatsProvider) ValuesNode(io.trino.sql.planner.plan.ValuesNode) Session(io.trino.Session) AccessDeniedException(io.trino.spi.security.AccessDeniedException) TableExecuteNode(io.trino.sql.planner.plan.TableExecuteNode) PlanPrinter(io.trino.sql.planner.planprinter.PlanPrinter) QueryPlanner.visibleFields(io.trino.sql.planner.QueryPlanner.visibleFields) CreateReference(io.trino.sql.planner.plan.TableWriterNode.CreateReference) ColumnMetadata(io.trino.spi.connector.ColumnMetadata) LimitNode(io.trino.sql.planner.plan.LimitNode) TableLayout(io.trino.metadata.TableLayout) TypeCoercion(io.trino.type.TypeCoercion) BOOLEAN(io.trino.spi.type.BooleanType.BOOLEAN) ConnectorTableMetadata(io.trino.spi.connector.ConnectorTableMetadata) StatsAndCosts(io.trino.cost.StatsAndCosts) ArrayList(java.util.ArrayList) LinkedHashMap(java.util.LinkedHashMap) VARCHAR(io.trino.spi.type.VarcharType.VARCHAR) MetadataUtil.createQualifiedObjectName(io.trino.metadata.MetadataUtil.createQualifiedObjectName) PlanOptimizer(io.trino.sql.planner.optimizations.PlanOptimizer) NodeRef(io.trino.sql.tree.NodeRef) ColumnHandle(io.trino.spi.connector.ColumnHandle) AggregationNode(io.trino.sql.planner.plan.AggregationNode) ImmutableSet.toImmutableSet(com.google.common.collect.ImmutableSet.toImmutableSet) VARBINARY(io.trino.spi.type.VarbinaryType.VARBINARY) Query(io.trino.sql.tree.Query) StringLiteral(io.trino.sql.tree.StringLiteral) TableStatisticAggregation(io.trino.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation) AggregationNode.singleGroupingSet(io.trino.sql.planner.plan.AggregationNode.singleGroupingSet) TableHandle(io.trino.metadata.TableHandle) QualifiedObjectName(io.trino.metadata.QualifiedObjectName) CharType(io.trino.spi.type.CharType) Table(io.trino.sql.tree.Table) TableFinishNode(io.trino.sql.planner.plan.TableFinishNode) TableStatisticsMetadata(io.trino.spi.statistics.TableStatisticsMetadata) UnknownType(io.trino.type.UnknownType) RefreshMaterializedView(io.trino.sql.tree.RefreshMaterializedView) TableWriterNode(io.trino.sql.planner.plan.TableWriterNode) OPTIMIZED_AND_VALIDATED(io.trino.sql.planner.LogicalPlanner.Stage.OPTIMIZED_AND_VALIDATED) Scope(io.trino.sql.analyzer.Scope) FilterNode(io.trino.sql.planner.plan.FilterNode) StatsCalculator(io.trino.cost.StatsCalculator) LambdaArgumentDeclaration(io.trino.sql.tree.LambdaArgumentDeclaration) RelationId(io.trino.sql.analyzer.RelationId) DeleteNode(io.trino.sql.planner.plan.DeleteNode) Update(io.trino.sql.tree.Update) INTEGER(io.trino.spi.type.IntegerType.INTEGER) FunctionCall(io.trino.sql.tree.FunctionCall) ImmutableMap(com.google.common.collect.ImmutableMap) ResolvedFunction(io.trino.metadata.ResolvedFunction) TypeSignatureTranslator.toSqlType(io.trino.sql.analyzer.TypeSignatureTranslator.toSqlType) TrinoException(io.trino.spi.TrinoException) ROW_COUNT(io.trino.spi.statistics.TableStatisticType.ROW_COUNT) TableExecuteHandle(io.trino.metadata.TableExecuteHandle) StatsProvider(io.trino.cost.StatsProvider) ComparisonExpression(io.trino.sql.tree.ComparisonExpression) String.format(java.lang.String.format) Preconditions.checkState(com.google.common.base.Preconditions.checkState) CoalesceExpression(io.trino.sql.tree.CoalesceExpression) GenericLiteral(io.trino.sql.tree.GenericLiteral) Objects(java.util.Objects) TableMetadata(io.trino.metadata.TableMetadata) List(java.util.List) IfExpression(io.trino.sql.tree.IfExpression) BIGINT(io.trino.spi.type.BigintType.BIGINT) StandardErrorCode(io.trino.spi.StandardErrorCode) Analyze(io.trino.sql.tree.Analyze) Entry(java.util.Map.Entry) Optional(java.util.Optional) Expression(io.trino.sql.tree.Expression) PlannerContext(io.trino.sql.PlannerContext) Analysis(io.trino.sql.analyzer.Analysis) PERMISSION_DENIED(io.trino.spi.StandardErrorCode.PERMISSION_DENIED) Logger(io.airlift.log.Logger) Type(io.trino.spi.type.Type) CreateTableAsSelect(io.trino.sql.tree.CreateTableAsSelect) HashMap(java.util.HashMap) StatisticAggregations(io.trino.sql.planner.plan.StatisticAggregations) SimpleImmutableEntry(java.util.AbstractMap.SimpleImmutableEntry) OPTIMIZED(io.trino.sql.planner.LogicalPlanner.Stage.OPTIMIZED) Function(java.util.function.Function) Cast(io.trino.sql.tree.Cast) VarcharType(io.trino.spi.type.VarcharType) CachingCostProvider(io.trino.cost.CachingCostProvider) ImmutableList(com.google.common.collect.ImmutableList) Verify.verify(com.google.common.base.Verify.verify) Objects.requireNonNull(java.util.Objects.requireNonNull) RefreshMaterializedViewNode(io.trino.sql.planner.plan.RefreshMaterializedViewNode) NullLiteral(io.trino.sql.tree.NullLiteral) Field(io.trino.sql.analyzer.Field) GREATER_THAN_OR_EQUAL(io.trino.sql.tree.ComparisonExpression.Operator.GREATER_THAN_OR_EQUAL) ProjectNode(io.trino.sql.planner.plan.ProjectNode) DISTRIBUTED_PLAN_SANITY_CHECKER(io.trino.sql.planner.sanity.PlanSanityChecker.DISTRIBUTED_PLAN_SANITY_CHECKER) WriterTarget(io.trino.sql.planner.plan.TableWriterNode.WriterTarget) TRUE_LITERAL(io.trino.sql.tree.BooleanLiteral.TRUE_LITERAL) UpdateNode(io.trino.sql.planner.plan.UpdateNode) QualifiedName(io.trino.sql.tree.QualifiedName) CostCalculator(io.trino.cost.CostCalculator) WarningCollector(io.trino.execution.warnings.WarningCollector) Row(io.trino.sql.tree.Row) Metadata(io.trino.metadata.Metadata) Insert(io.trino.sql.tree.Insert) Streams.zip(com.google.common.collect.Streams.zip) ColumnMetadata(io.trino.spi.connector.ColumnMetadata) TableStatisticAggregation(io.trino.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation) ArrayList(java.util.ArrayList) TableFinishNode(io.trino.sql.planner.plan.TableFinishNode) StatisticAggregations(io.trino.sql.planner.plan.StatisticAggregations) TableWriterNode(io.trino.sql.planner.plan.TableWriterNode)

Example 4 with StatisticAggregations

use of io.trino.sql.planner.plan.StatisticAggregations in project trino by trinodb.

the class StatisticsAggregationPlanner method createStatisticsAggregation.

public TableStatisticAggregation createStatisticsAggregation(TableStatisticsMetadata statisticsMetadata, Map<String, Symbol> columnToSymbolMap) {
    StatisticAggregationsDescriptor.Builder<Symbol> descriptor = StatisticAggregationsDescriptor.builder();
    List<String> groupingColumns = statisticsMetadata.getGroupingColumns();
    List<Symbol> groupingSymbols = groupingColumns.stream().map(columnToSymbolMap::get).collect(toImmutableList());
    for (int i = 0; i < groupingSymbols.size(); i++) {
        descriptor.addGrouping(groupingColumns.get(i), groupingSymbols.get(i));
    }
    ImmutableMap.Builder<Symbol, AggregationNode.Aggregation> aggregations = ImmutableMap.builder();
    for (TableStatisticType type : statisticsMetadata.getTableStatistics()) {
        if (type != ROW_COUNT) {
            throw new TrinoException(NOT_SUPPORTED, "Table-wide statistic type not supported: " + type);
        }
        AggregationNode.Aggregation aggregation = new AggregationNode.Aggregation(metadata.resolveFunction(session, QualifiedName.of("count"), ImmutableList.of()), ImmutableList.of(), false, Optional.empty(), Optional.empty(), Optional.empty());
        Symbol symbol = symbolAllocator.newSymbol("rowCount", BIGINT);
        aggregations.put(symbol, aggregation);
        descriptor.addTableStatistic(ROW_COUNT, symbol);
    }
    for (ColumnStatisticMetadata columnStatisticMetadata : statisticsMetadata.getColumnStatistics()) {
        String columnName = columnStatisticMetadata.getColumnName();
        ColumnStatisticType statisticType = columnStatisticMetadata.getStatisticType();
        Symbol inputSymbol = columnToSymbolMap.get(columnName);
        verifyNotNull(inputSymbol, "inputSymbol is null");
        Type inputType = symbolAllocator.getTypes().get(inputSymbol);
        verifyNotNull(inputType, "inputType is null for symbol: %s", inputSymbol);
        ColumnStatisticsAggregation aggregation = createColumnAggregation(statisticType, inputSymbol, inputType);
        Symbol symbol = symbolAllocator.newSymbol(statisticType + ":" + columnName, aggregation.getOutputType());
        aggregations.put(symbol, aggregation.getAggregation());
        descriptor.addColumnStatistic(columnStatisticMetadata, symbol);
    }
    StatisticAggregations aggregation = new StatisticAggregations(aggregations.buildOrThrow(), groupingSymbols);
    return new TableStatisticAggregation(aggregation, descriptor.build());
}
Also used : ColumnStatisticMetadata(io.trino.spi.statistics.ColumnStatisticMetadata) AggregationNode(io.trino.sql.planner.plan.AggregationNode) ImmutableMap(com.google.common.collect.ImmutableMap) StatisticAggregations(io.trino.sql.planner.plan.StatisticAggregations) Type(io.trino.spi.type.Type) ColumnStatisticType(io.trino.spi.statistics.ColumnStatisticType) TableStatisticType(io.trino.spi.statistics.TableStatisticType) ColumnStatisticType(io.trino.spi.statistics.ColumnStatisticType) TrinoException(io.trino.spi.TrinoException) StatisticAggregationsDescriptor(io.trino.sql.planner.plan.StatisticAggregationsDescriptor) TableStatisticType(io.trino.spi.statistics.TableStatisticType)

Aggregations

StatisticAggregations (io.trino.sql.planner.plan.StatisticAggregations)4 ImmutableMap (com.google.common.collect.ImmutableMap)3 AggregationNode (io.trino.sql.planner.plan.AggregationNode)3 ImmutableList (com.google.common.collect.ImmutableList)2 ImmutableList.toImmutableList (com.google.common.collect.ImmutableList.toImmutableList)2 ImmutableMap.toImmutableMap (com.google.common.collect.ImmutableMap.toImmutableMap)2 TableHandle (io.trino.metadata.TableHandle)2 TableMetadata (io.trino.metadata.TableMetadata)2 ColumnHandle (io.trino.spi.connector.ColumnHandle)2 ColumnMetadata (io.trino.spi.connector.ColumnMetadata)2 ConnectorTableMetadata (io.trino.spi.connector.ConnectorTableMetadata)2 TableStatisticsMetadata (io.trino.spi.statistics.TableStatisticsMetadata)2 TableStatisticAggregation (io.trino.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation)2 PlanNode (io.trino.sql.planner.plan.PlanNode)2 StatisticsWriterNode (io.trino.sql.planner.plan.StatisticsWriterNode)2 Preconditions.checkState (com.google.common.base.Preconditions.checkState)1 Verify.verify (com.google.common.base.Verify.verify)1 ImmutableSet (com.google.common.collect.ImmutableSet)1 ImmutableSet.toImmutableSet (com.google.common.collect.ImmutableSet.toImmutableSet)1 Streams.zip (com.google.common.collect.Streams.zip)1