Examples with TableStatisticAggregation com.facebook.presto.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation used on opensource projects

Example 1 with TableStatisticAggregation

use of com.facebook.presto.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation in project presto by prestodb.

the class LogicalPlanner method createTableWriterPlan.

private RelationPlan createTableWriterPlan(Analysis analysis, RelationPlan plan, WriterTarget target, List<String> columnNames, List<ColumnMetadata> columnMetadataList, Optional<NewTableLayout> writeTableLayout, Optional<NewTableLayout> preferredShuffleLayout, TableStatisticsMetadata statisticsMetadata) {
    verify(!(writeTableLayout.isPresent() && preferredShuffleLayout.isPresent()), "writeTableLayout and preferredShuffleLayout cannot both exist");
    PlanNode source = plan.getRoot();
    if (!analysis.isCreateTableAsSelectWithData()) {
        source = new LimitNode(source.getSourceLocation(), idAllocator.getNextId(), source, 0L, FINAL);
    }
    List<VariableReferenceExpression> variables = plan.getFieldMappings();
    Optional<PartitioningScheme> tablePartitioningScheme = getPartitioningSchemeForTableWrite(writeTableLayout, columnNames, variables);
    Optional<PartitioningScheme> preferredShufflePartitioningScheme = getPartitioningSchemeForTableWrite(preferredShuffleLayout, columnNames, variables);
    verify(columnNames.size() == variables.size(), "columnNames.size() != variables.size(): %s and %s", columnNames, variables);
    Map<String, VariableReferenceExpression> columnToVariableMap = zip(columnNames.stream(), plan.getFieldMappings().stream(), SimpleImmutableEntry::new).collect(toImmutableMap(Entry::getKey, Entry::getValue));
    Set<VariableReferenceExpression> notNullColumnVariables = columnMetadataList.stream().filter(column -> !column.isNullable()).map(ColumnMetadata::getName).map(columnToVariableMap::get).collect(toImmutableSet());
    if (!statisticsMetadata.isEmpty()) {
        TableStatisticAggregation result = statisticsAggregationPlanner.createStatisticsAggregation(statisticsMetadata, columnToVariableMap, true);
        StatisticAggregations.Parts aggregations = result.getAggregations().splitIntoPartialAndFinal(variableAllocator, metadata.getFunctionAndTypeManager());
        TableFinishNode commitNode = new TableFinishNode(source.getSourceLocation(), idAllocator.getNextId(), new TableWriterNode(source.getSourceLocation(), idAllocator.getNextId(), source, Optional.of(target), variableAllocator.newVariable("rows", BIGINT), variableAllocator.newVariable("fragments", VARBINARY), variableAllocator.newVariable("commitcontext", VARBINARY), plan.getFieldMappings(), columnNames, notNullColumnVariables, tablePartitioningScheme, preferredShufflePartitioningScheme, // the data consumed by the TableWriteOperator
        Optional.of(aggregations.getPartialAggregation())), Optional.of(target), variableAllocator.newVariable("rows", BIGINT), // by the partial aggregation from all of the writer nodes
        Optional.of(aggregations.getFinalAggregation()), Optional.of(result.getDescriptor()));
        return new RelationPlan(commitNode, analysis.getRootScope(), commitNode.getOutputVariables());
    }
    TableFinishNode commitNode = new TableFinishNode(source.getSourceLocation(), idAllocator.getNextId(), new TableWriterNode(source.getSourceLocation(), idAllocator.getNextId(), source, Optional.of(target), variableAllocator.newVariable("rows", BIGINT), variableAllocator.newVariable("fragments", VARBINARY), variableAllocator.newVariable("commitcontext", VARBINARY), plan.getFieldMappings(), columnNames, notNullColumnVariables, tablePartitioningScheme, preferredShufflePartitioningScheme, Optional.empty()), Optional.of(target), variableAllocator.newVariable("rows", BIGINT), Optional.empty(), Optional.empty());
    return new RelationPlan(commitNode, analysis.getRootScope(), commitNode.getOutputVariables());
}

Example 2 with TableStatisticAggregation

use of com.facebook.presto.sql.planner.StatisticsAggregationPlanner.TableStatisticAggregation in project presto by prestodb.

the class LogicalPlanner method createAnalyzePlan.

private RelationPlan createAnalyzePlan(Analysis analysis, Analyze analyzeStatement) {
    TableHandle targetTable = analysis.getAnalyzeTarget().get();
    // Plan table scan
    Map<String, ColumnHandle> columnHandles = metadata.getColumnHandles(session, targetTable);
    ImmutableList.Builder<VariableReferenceExpression> tableScanOutputsBuilder = ImmutableList.builder();
    ImmutableMap.Builder<VariableReferenceExpression, ColumnHandle> variableToColumnHandle = ImmutableMap.builder();
    ImmutableMap.Builder<String, VariableReferenceExpression> columnNameToVariable = ImmutableMap.builder();
    TableMetadata tableMetadata = metadata.getTableMetadata(session, targetTable);
    for (ColumnMetadata column : tableMetadata.getColumns()) {
        VariableReferenceExpression variable = variableAllocator.newVariable(getSourceLocation(analyzeStatement), column.getName(), column.getType());
        tableScanOutputsBuilder.add(variable);
        variableToColumnHandle.put(variable, columnHandles.get(column.getName()));
        columnNameToVariable.put(column.getName(), variable);
    }
    List<VariableReferenceExpression> tableScanOutputs = tableScanOutputsBuilder.build();
    TableStatisticsMetadata tableStatisticsMetadata = metadata.getStatisticsCollectionMetadata(session, targetTable.getConnectorId().getCatalogName(), tableMetadata.getMetadata());
    TableStatisticAggregation tableStatisticAggregation = statisticsAggregationPlanner.createStatisticsAggregation(tableStatisticsMetadata, columnNameToVariable.build(), true);
    StatisticAggregations statisticAggregations = tableStatisticAggregation.getAggregations();
    PlanNode planNode = new StatisticsWriterNode(getSourceLocation(analyzeStatement), idAllocator.getNextId(), new AggregationNode(getSourceLocation(analyzeStatement), idAllocator.getNextId(), new TableScanNode(getSourceLocation(analyzeStatement), idAllocator.getNextId(), targetTable, tableScanOutputs, variableToColumnHandle.build(), TupleDomain.all(), TupleDomain.all()), statisticAggregations.getAggregations(), singleGroupingSet(statisticAggregations.getGroupingVariables()), ImmutableList.of(), AggregationNode.Step.SINGLE, Optional.empty(), Optional.empty()), targetTable, variableAllocator.newVariable(getSourceLocation(analyzeStatement), "rows", BIGINT), tableStatisticsMetadata.getTableStatistics().contains(ROW_COUNT), tableStatisticAggregation.getDescriptor());
    return new RelationPlan(planNode, analysis.getScope(analyzeStatement), planNode.getOutputVariables());
}