ClickHouse/src/Interpreters/InterpreterSelectQuery.cpp

#include <DataStreams/OneBlockInputStream.h>
#include <DataStreams/materializeBlock.h>

#include <DataTypes/DataTypeAggregateFunction.h>

#include <Parsers/ASTFunction.h>
#include <Parsers/ASTIdentifier.h>
#include <Parsers/ASTLiteral.h>
#include <Parsers/ASTOrderByElement.h>
#include <Parsers/ASTSelectWithUnionQuery.h>
#include <Parsers/ASTTablesInSelectQuery.h>
#include <Parsers/ExpressionListParsers.h>
#include <Parsers/parseQuery.h>

#include <Access/AccessFlags.h>
#include <Access/ContextAccess.h>

#include <AggregateFunctions/AggregateFunctionCount.h>

#include <Interpreters/ApplyWithAliasVisitor.h>
#include <Interpreters/ApplyWithSubqueryVisitor.h>
#include <Interpreters/InterpreterSelectQuery.h>
#include <Interpreters/InterpreterSelectWithUnionQuery.h>
#include <Interpreters/InterpreterSetQuery.h>
#include <Interpreters/evaluateConstantExpression.h>
#include <Interpreters/convertFieldToType.h>
#include <Interpreters/addTypeConversionToAST.h>
#include <Interpreters/ExpressionAnalyzer.h>
#include <Interpreters/getTableExpressions.h>
#include <Interpreters/JoinToSubqueryTransformVisitor.h>
#include <Interpreters/CrossToInnerJoinVisitor.h>
#include <Interpreters/TableJoin.h>
#include <Interpreters/JoinSwitcher.h>
#include <Interpreters/JoinedTables.h>
#include <Interpreters/OpenTelemetrySpanLog.h>
#include <Interpreters/QueryAliasesVisitor.h>
#include <Interpreters/replaceAliasColumnsInQuery.h>

#include <Processors/Pipe.h>
#include <Processors/QueryPlan/AddingDelayedSourceStep.h>
#include <Processors/QueryPlan/AggregatingStep.h>
#include <Processors/QueryPlan/ArrayJoinStep.h>
#include <Processors/QueryPlan/CreatingSetsStep.h>
#include <Processors/QueryPlan/CubeStep.h>
#include <Processors/QueryPlan/DistinctStep.h>
#include <Processors/QueryPlan/ExpressionStep.h>
#include <Processors/QueryPlan/ExtremesStep.h>
#include <Processors/QueryPlan/FillingStep.h>
#include <Processors/QueryPlan/FilterStep.h>
#include <Processors/QueryPlan/FinishSortingStep.h>
#include <Processors/QueryPlan/LimitByStep.h>
#include <Processors/QueryPlan/LimitStep.h>
#include <Processors/QueryPlan/MergeSortingStep.h>
#include <Processors/QueryPlan/MergingAggregatedStep.h>
#include <Processors/QueryPlan/MergingSortedStep.h>
#include <Processors/QueryPlan/OffsetStep.h>
#include <Processors/QueryPlan/PartialSortingStep.h>
#include <Processors/QueryPlan/QueryPlan.h>
#include <Processors/QueryPlan/ReadFromPreparedSource.h>
#include <Processors/QueryPlan/ReadNothingStep.h>
#include <Processors/QueryPlan/RollupStep.h>
#include <Processors/QueryPlan/SettingQuotaAndLimitsStep.h>
#include <Processors/QueryPlan/TotalsHavingStep.h>
#include <Processors/QueryPlan/WindowStep.h>
#include <Processors/QueryPlan/Optimizations/QueryPlanOptimizationSettings.h>
#include <Processors/Sources/NullSource.h>
#include <Processors/Sources/SourceFromInputStream.h>
#include <Processors/Transforms/AggregatingTransform.h>
#include <Processors/Transforms/ExpressionTransform.h>
#include <Processors/Transforms/FilterTransform.h>
#include <Processors/Transforms/JoiningTransform.h>

#include <Storages/MergeTree/MergeTreeWhereOptimizer.h>
#include <Storages/IStorage.h>
#include <Storages/StorageView.h>

#include <Functions/IFunction.h>
#include <Core/Field.h>
#include <common/types.h>
#include <Columns/Collator.h>
#include <Common/FieldVisitorsAccurateComparison.h>
#include <Common/typeid_cast.h>
#include <Common/checkStackSize.h>
#include <ext/map.h>
#include <ext/scope_guard_safe.h>
#include <memory>


namespace DB
{

namespace ErrorCodes
{
    extern const int TOO_DEEP_SUBQUERIES;
    extern const int SAMPLING_NOT_SUPPORTED;
    extern const int ILLEGAL_FINAL;
    extern const int ILLEGAL_PREWHERE;
    extern const int TOO_MANY_COLUMNS;
    extern const int LOGICAL_ERROR;
    extern const int NOT_IMPLEMENTED;
    extern const int PARAMETER_OUT_OF_BOUND;
    extern const int INVALID_LIMIT_EXPRESSION;
    extern const int INVALID_WITH_FILL_EXPRESSION;
    extern const int ACCESS_DENIED;
}

/// Assumes `storage` is set and the table filter (row-level security) is not empty.
String InterpreterSelectQuery::generateFilterActions(ActionsDAGPtr & actions, const Names & prerequisite_columns) const
{
    const auto & db_name = table_id.getDatabaseName();
    const auto & table_name = table_id.getTableName();

    /// TODO: implement some AST builders for this kind of stuff
    ASTPtr query_ast = std::make_shared<ASTSelectQuery>();
    auto * select_ast = query_ast->as<ASTSelectQuery>();

    select_ast->setExpression(ASTSelectQuery::Expression::SELECT, std::make_shared<ASTExpressionList>());
    auto expr_list = select_ast->select();

    // The first column is our filter expression.
    expr_list->children.push_back(row_policy_filter);

    /// Keep columns that are required after the filter actions.
    for (const auto & column_str : prerequisite_columns)
    {
        ParserExpression expr_parser;
        expr_list->children.push_back(parseQuery(expr_parser, column_str, 0, context->getSettingsRef().max_parser_depth));
    }

    select_ast->setExpression(ASTSelectQuery::Expression::TABLES, std::make_shared<ASTTablesInSelectQuery>());
    auto tables = select_ast->tables();
    auto tables_elem = std::make_shared<ASTTablesInSelectQueryElement>();
    auto table_expr = std::make_shared<ASTTableExpression>();
    tables->children.push_back(tables_elem);
    tables_elem->table_expression = table_expr;
    tables_elem->children.push_back(table_expr);
    table_expr->database_and_table_name = createTableIdentifier(db_name, table_name);
    table_expr->children.push_back(table_expr->database_and_table_name);

    /// Using separate expression analyzer to prevent any possible alias injection
    auto syntax_result = TreeRewriter(*context).analyzeSelect(query_ast, TreeRewriterResult({}, storage, metadata_snapshot));
    SelectQueryExpressionAnalyzer analyzer(query_ast, syntax_result, *context, metadata_snapshot);
    actions = analyzer.simpleSelectActions();

    auto column_name = expr_list->children.at(0)->getColumnName();
    actions->removeUnusedActions(NameSet{column_name});
    actions->projectInput(false);

    for (const auto * node : actions->getInputs())
        actions->getIndex().push_back(node);

    return column_name;
}

InterpreterSelectQuery::InterpreterSelectQuery(
    const ASTPtr & query_ptr_,
    const Context & context_,
    const SelectQueryOptions & options_,
    const Names & required_result_column_names_)
    : InterpreterSelectQuery(query_ptr_, context_, nullptr, std::nullopt, nullptr, options_, required_result_column_names_)
{
}

InterpreterSelectQuery::InterpreterSelectQuery(
    const ASTPtr & query_ptr_,
    const Context & context_,
    const BlockInputStreamPtr & input_,
    const SelectQueryOptions & options_)
    : InterpreterSelectQuery(query_ptr_, context_, input_, std::nullopt, nullptr, options_.copy().noSubquery())
{}

InterpreterSelectQuery::InterpreterSelectQuery(
        const ASTPtr & query_ptr_,
        const Context & context_,
        Pipe input_pipe_,
        const SelectQueryOptions & options_)
        : InterpreterSelectQuery(query_ptr_, context_, nullptr, std::move(input_pipe_), nullptr, options_.copy().noSubquery())
{}

InterpreterSelectQuery::InterpreterSelectQuery(
    const ASTPtr & query_ptr_,
    const Context & context_,
    const StoragePtr & storage_,
    const StorageMetadataPtr & metadata_snapshot_,
    const SelectQueryOptions & options_)
    : InterpreterSelectQuery(query_ptr_, context_, nullptr, std::nullopt, storage_, options_.copy().noSubquery(), {}, metadata_snapshot_)
{}

InterpreterSelectQuery::~InterpreterSelectQuery() = default;


/** There are no limits on the maximum size of the result for the subquery.
  *  Since the result of the query is not the result of the entire query.
  */
static Context getSubqueryContext(const Context & context)
{
    Context subquery_context = context;
    Settings subquery_settings = context.getSettings();
    subquery_settings.max_result_rows = 0;
    subquery_settings.max_result_bytes = 0;
    /// The calculation of extremes does not make sense and is not necessary (if you do it, then the extremes of the subquery can be taken for whole query).
    subquery_settings.extremes = false;
    subquery_context.setSettings(subquery_settings);
    return subquery_context;
}

static void rewriteMultipleJoins(ASTPtr & query, const TablesWithColumns & tables, const String & database, const Settings & settings)
{
    ASTSelectQuery & select = query->as<ASTSelectQuery &>();

    Aliases aliases;
    if (ASTPtr with = select.with())
        QueryAliasesNoSubqueriesVisitor(aliases).visit(with);
    QueryAliasesNoSubqueriesVisitor(aliases).visit(select.select());

    CrossToInnerJoinVisitor::Data cross_to_inner{tables, aliases, database};
    cross_to_inner.cross_to_inner_join_rewrite = settings.cross_to_inner_join_rewrite;
    CrossToInnerJoinVisitor(cross_to_inner).visit(query);

    JoinToSubqueryTransformVisitor::Data join_to_subs_data{tables, aliases};
    JoinToSubqueryTransformVisitor(join_to_subs_data).visit(query);
}

/// Checks that the current user has the SELECT privilege.
static void checkAccessRightsForSelect(
    const Context & context,
    const StorageID & table_id,
    const StorageMetadataPtr & table_metadata,
    const Strings & required_columns,
    const TreeRewriterResult & syntax_analyzer_result)
{
    if (!syntax_analyzer_result.has_explicit_columns && table_metadata && !table_metadata->getColumns().empty())
    {
        /// For a trivial query like "SELECT count() FROM table" access is granted if at least
        /// one column is accessible.
        /// In this case just checking access for `required_columns` doesn't work correctly
        /// because `required_columns` will contain the name of a column of minimum size (see TreeRewriterResult::collectUsedColumns())
        /// which is probably not the same column as the column the current user has access to.
        auto access = context.getAccess();
        for (const auto & column : table_metadata->getColumns())
        {
            if (access->isGranted(AccessType::SELECT, table_id.database_name, table_id.table_name, column.name))
                return;
        }
        throw Exception(context.getUserName() + ": Not enough privileges. "
                        "To execute this query it's necessary to have grant SELECT for at least one column on " + table_id.getFullTableName(),
                        ErrorCodes::ACCESS_DENIED);
    }

    /// General check.
    context.checkAccess(AccessType::SELECT, table_id, required_columns);
}

/// Returns true if we should ignore quotas and limits for a specified table in the system database.
static bool shouldIgnoreQuotaAndLimits(const StorageID & table_id)
{
    if (table_id.database_name == DatabaseCatalog::SYSTEM_DATABASE)
    {
        static const boost::container::flat_set<String> tables_ignoring_quota{"quotas", "quota_limits", "quota_usage", "quotas_usage", "one"};
        if (tables_ignoring_quota.count(table_id.table_name))
            return true;
    }
    return false;
}

InterpreterSelectQuery::InterpreterSelectQuery(
    const ASTPtr & query_ptr_,
    const Context & context_,
    const BlockInputStreamPtr & input_,
    std::optional<Pipe> input_pipe_,
    const StoragePtr & storage_,
    const SelectQueryOptions & options_,
    const Names & required_result_column_names,
    const StorageMetadataPtr & metadata_snapshot_)
    /// NOTE: the query almost always should be cloned because it will be modified during analysis.
    : IInterpreterUnionOrSelectQuery(options_.modify_inplace ? query_ptr_ : query_ptr_->clone(), context_, options_)
    , storage(storage_)
    , input(input_)
    , input_pipe(std::move(input_pipe_))
    , log(&Poco::Logger::get("InterpreterSelectQuery"))
    , metadata_snapshot(metadata_snapshot_)
{
    checkStackSize();

    initSettings();
    const Settings & settings = context->getSettingsRef();

    if (settings.max_subquery_depth && options.subquery_depth > settings.max_subquery_depth)
        throw Exception("Too deep subqueries. Maximum: " + settings.max_subquery_depth.toString(),
            ErrorCodes::TOO_DEEP_SUBQUERIES);

    bool has_input = input || input_pipe;
    if (input)
    {
        /// Read from prepared input.
        source_header = input->getHeader();
    }
    else if (input_pipe)
    {
        /// Read from prepared input.
        source_header = input_pipe->getHeader();
    }

    // Only propagate WITH elements to subqueries if we're not a subquery
    if (!options.is_subquery)
    {
        if (context->getSettingsRef().enable_global_with_statement)
            ApplyWithAliasVisitor().visit(query_ptr);
        ApplyWithSubqueryVisitor().visit(query_ptr);
    }

    JoinedTables joined_tables(getSubqueryContext(*context), getSelectQuery());

    bool got_storage_from_query = false;
    if (!has_input && !storage)
    {
        storage = joined_tables.getLeftTableStorage();
        got_storage_from_query = true;
    }

    if (storage)
    {
        table_lock = storage->lockForShare(context->getInitialQueryId(), context->getSettingsRef().lock_acquire_timeout);
        table_id = storage->getStorageID();
        if (!metadata_snapshot)
            metadata_snapshot = storage->getInMemoryMetadataPtr();
    }

    if (has_input || !joined_tables.resolveTables())
        joined_tables.makeFakeTable(storage, metadata_snapshot, source_header);

    /// Rewrite JOINs
    if (!has_input && joined_tables.tablesCount() > 1)
    {
        rewriteMultipleJoins(query_ptr, joined_tables.tablesWithColumns(), context->getCurrentDatabase(), context->getSettingsRef());

        joined_tables.reset(getSelectQuery());
        joined_tables.resolveTables();

        if (storage && joined_tables.isLeftTableSubquery())
        {
            /// Rewritten with subquery. Free storage locks here.
            storage = nullptr;
            table_lock.reset();
            table_id = StorageID::createEmpty();
            metadata_snapshot = nullptr;
        }
    }

    if (!has_input)
    {
        interpreter_subquery = joined_tables.makeLeftTableSubquery(options.subquery());
        if (interpreter_subquery)
            source_header = interpreter_subquery->getSampleBlock();
    }

    joined_tables.rewriteDistributedInAndJoins(query_ptr);

    max_streams = settings.max_threads;
    ASTSelectQuery & query = getSelectQuery();
    std::shared_ptr<TableJoin> table_join = joined_tables.makeTableJoin(query);

    if (storage)
        row_policy_filter = context->getRowPolicyCondition(table_id.getDatabaseName(), table_id.getTableName(), RowPolicy::SELECT_FILTER);

    StorageView * view = nullptr;
    if (storage)
        view = dynamic_cast<StorageView *>(storage.get());

    SubqueriesForSets subquery_for_sets;

    auto analyze = [&] (bool try_move_to_prewhere)
    {
        /// Allow push down and other optimizations for VIEW: replace with subquery and rewrite it.
        ASTPtr view_table;
        if (view)
            view->replaceWithSubquery(getSelectQuery(), view_table, metadata_snapshot);

        syntax_analyzer_result = TreeRewriter(*context).analyzeSelect(
            query_ptr,
            TreeRewriterResult(source_header.getNamesAndTypesList(), storage, metadata_snapshot),
            options, joined_tables.tablesWithColumns(), required_result_column_names, table_join);

        /// Save scalar sub queries's results in the query context
        if (!options.only_analyze && context->hasQueryContext())
            for (const auto & it : syntax_analyzer_result->getScalars())
                context->getQueryContext().addScalar(it.first, it.second);

        if (view)
        {
            /// Restore original view name. Save rewritten subquery for future usage in StorageView.
            query_info.view_query = view->restoreViewName(getSelectQuery(), view_table);
            view = nullptr;
        }

        if (try_move_to_prewhere && storage && query.where() && !query.prewhere())
        {
            /// PREWHERE optimization: transfer some condition from WHERE to PREWHERE if enabled and viable
            if (const auto & column_sizes = storage->getColumnSizes(); !column_sizes.empty())
            {
                /// Extract column compressed sizes.
                std::unordered_map<std::string, UInt64> column_compressed_sizes;
                for (const auto & [name, sizes] : column_sizes)
                    column_compressed_sizes[name] = sizes.data_compressed;

                SelectQueryInfo current_info;
                current_info.query = query_ptr;
                current_info.syntax_analyzer_result = syntax_analyzer_result;

                MergeTreeWhereOptimizer{current_info, *context, std::move(column_compressed_sizes), metadata_snapshot, syntax_analyzer_result->requiredSourceColumns(), log};
            }
        }

        query_analyzer = std::make_unique<SelectQueryExpressionAnalyzer>(
                query_ptr, syntax_analyzer_result, *context, metadata_snapshot,
                NameSet(required_result_column_names.begin(), required_result_column_names.end()),
                !options.only_analyze, options, std::move(subquery_for_sets));

        if (!options.only_analyze)
        {
            if (query.sampleSize() && (input || input_pipe || !storage || !storage->supportsSampling()))
                throw Exception("Illegal SAMPLE: table doesn't support sampling", ErrorCodes::SAMPLING_NOT_SUPPORTED);

            if (query.final() && (input || input_pipe || !storage || !storage->supportsFinal()))
                throw Exception((!input && !input_pipe && storage) ? "Storage " + storage->getName() + " doesn't support FINAL" : "Illegal FINAL", ErrorCodes::ILLEGAL_FINAL);

            if (query.prewhere() && (input || input_pipe || !storage || !storage->supportsPrewhere()))
                throw Exception((!input && !input_pipe && storage) ? "Storage " + storage->getName() + " doesn't support PREWHERE" : "Illegal PREWHERE", ErrorCodes::ILLEGAL_PREWHERE);

            /// Save the new temporary tables in the query context
            for (const auto & it : query_analyzer->getExternalTables())
                if (!context->tryResolveStorageID({"", it.first}, Context::ResolveExternal))
                    context->addExternalTable(it.first, std::move(*it.second));
        }

        if (!options.only_analyze || options.modify_inplace)
        {
            if (syntax_analyzer_result->rewrite_subqueries)
            {
                /// remake interpreter_subquery when PredicateOptimizer rewrites subqueries and main table is subquery
                interpreter_subquery = joined_tables.makeLeftTableSubquery(options.subquery());
            }
        }

        if (interpreter_subquery)
        {
            /// If there is an aggregation in the outer query, WITH TOTALS is ignored in the subquery.
            if (query_analyzer->hasAggregation())
                interpreter_subquery->ignoreWithTotals();
        }

        required_columns = syntax_analyzer_result->requiredSourceColumns();

        if (storage)
        {
            /// Fix source_header for filter actions.
            if (row_policy_filter)
            {
                filter_info = std::make_shared<FilterDAGInfo>();
                filter_info->column_name = generateFilterActions(filter_info->actions, required_columns);

                auto required_columns_from_filter = filter_info->actions->getRequiredColumns();

                for (const auto & column : required_columns_from_filter)
                {
                    if (required_columns.end() == std::find(required_columns.begin(), required_columns.end(), column.name))
                        required_columns.push_back(column.name);
                }
            }

            source_header = metadata_snapshot->getSampleBlockForColumns(required_columns, storage->getVirtuals(), storage->getStorageID());
        }

        /// Calculate structure of the result.
        result_header = getSampleBlockImpl();
    };

    analyze(settings.optimize_move_to_prewhere);

    bool need_analyze_again = false;
    if (analysis_result.prewhere_constant_filter_description.always_false || analysis_result.prewhere_constant_filter_description.always_true)
    {
        if (analysis_result.prewhere_constant_filter_description.always_true)
            query.setExpression(ASTSelectQuery::Expression::PREWHERE, {});
        else
            query.setExpression(ASTSelectQuery::Expression::PREWHERE, std::make_shared<ASTLiteral>(0u));
        need_analyze_again = true;
    }
    if (analysis_result.where_constant_filter_description.always_false || analysis_result.where_constant_filter_description.always_true)
    {
        if (analysis_result.where_constant_filter_description.always_true)
            query.setExpression(ASTSelectQuery::Expression::WHERE, {});
        else
            query.setExpression(ASTSelectQuery::Expression::WHERE, std::make_shared<ASTLiteral>(0u));
        need_analyze_again = true;
    }
    if (query.prewhere() && query.where())
    {
        /// Filter block in WHERE instead to get better performance
        query.setExpression(ASTSelectQuery::Expression::WHERE, makeASTFunction("and", query.prewhere()->clone(), query.where()->clone()));
        need_analyze_again = true;
    }

    if (need_analyze_again)
    {
        LOG_TRACE(log, "Running 'analyze' second time");
        query_analyzer->getSubqueriesForSets().clear();
        subquery_for_sets = SubqueriesForSets();

        /// Do not try move conditions to PREWHERE for the second time.
        /// Otherwise, we won't be able to fallback from inefficient PREWHERE to WHERE later.
        analyze(/* try_move_to_prewhere = */ false);
    }

    /// If there is no WHERE, filter blocks as usual
    if (query.prewhere() && !query.where())
        analysis_result.prewhere_info->need_filter = true;

    if (table_id && got_storage_from_query && !joined_tables.isLeftTableFunction())
    {
        /// The current user should have the SELECT privilege.
        /// If this table_id is for a table function we don't check access rights here because in this case they have been already checked in ITableFunction::execute().
        checkAccessRightsForSelect(*context, table_id, metadata_snapshot, required_columns, *syntax_analyzer_result);

        /// Remove limits for some tables in the `system` database.
        if (shouldIgnoreQuotaAndLimits(table_id) && (joined_tables.tablesCount() <= 1))
        {
            options.ignore_quota = true;
            options.ignore_limits = true;
        }
    }

    /// Blocks used in expression analysis contains size 1 const columns for constant folding and
    ///  null non-const columns to avoid useless memory allocations. However, a valid block sample
    ///  requires all columns to be of size 0, thus we need to sanitize the block here.
    sanitizeBlock(result_header, true);
}

void InterpreterSelectQuery::buildQueryPlan(QueryPlan & query_plan)
{
    executeImpl(query_plan, input, std::move(input_pipe));

    /// We must guarantee that result structure is the same as in getSampleBlock()
    if (!blocksHaveEqualStructure(query_plan.getCurrentDataStream().header, result_header))
    {
        auto convert_actions_dag = ActionsDAG::makeConvertingActions(
                query_plan.getCurrentDataStream().header.getColumnsWithTypeAndName(),
                result_header.getColumnsWithTypeAndName(),
                ActionsDAG::MatchColumnsMode::Name,
                true);

        auto converting = std::make_unique<ExpressionStep>(query_plan.getCurrentDataStream(), convert_actions_dag);
        query_plan.addStep(std::move(converting));
    }
}

BlockIO InterpreterSelectQuery::execute()
{
    BlockIO res;
    QueryPlan query_plan;

    buildQueryPlan(query_plan);

    res.pipeline = std::move(*query_plan.buildQueryPipeline(
        QueryPlanOptimizationSettings::fromContext(*context),
        BuildQueryPipelineSettings::fromContext(*context)));
    return res;
}

Block InterpreterSelectQuery::getSampleBlockImpl()
{
    OpenTelemetrySpanHolder span(__PRETTY_FUNCTION__);

    query_info.query = query_ptr;

    if (storage && !options.only_analyze)
        from_stage = storage->getQueryProcessingStage(*context, options.to_stage, query_info);

    /// Do I need to perform the first part of the pipeline?
    /// Running on remote servers during distributed processing or if query is not distributed.
    ///
    /// Also note that with distributed_group_by_no_merge=1 or when there is
    /// only one remote server, it is equal to local query in terms of query
    /// stages (or when due to optimize_distributed_group_by_sharding_key the query was processed up to Complete stage).
    bool first_stage = from_stage < QueryProcessingStage::WithMergeableState
        && options.to_stage >= QueryProcessingStage::WithMergeableState;
    /// Do I need to execute the second part of the pipeline?
    /// Running on the initiating server during distributed processing or if query is not distributed.
    ///
    /// Also note that with distributed_group_by_no_merge=2 (i.e. when optimize_distributed_group_by_sharding_key takes place)
    /// the query on the remote server will be processed up to WithMergeableStateAfterAggregation,
    /// So it will do partial second stage (second_stage=true), and initiator will do the final part.
    bool second_stage = from_stage <= QueryProcessingStage::WithMergeableState
        && options.to_stage > QueryProcessingStage::WithMergeableState;

    analysis_result = ExpressionAnalysisResult(
            *query_analyzer,
            metadata_snapshot,
            first_stage,
            second_stage,
            options.only_analyze,
            filter_info,
            source_header);

    if (options.to_stage == QueryProcessingStage::Enum::FetchColumns)
    {
        auto header = source_header;

        if (analysis_result.prewhere_info)
        {
            ExpressionActions(
                analysis_result.prewhere_info->prewhere_actions,
                ExpressionActionsSettings::fromContext(*context)).execute(header);
            if (analysis_result.prewhere_info->remove_prewhere_column)
                header.erase(analysis_result.prewhere_info->prewhere_column_name);
        }
        return header;
    }

    if (options.to_stage == QueryProcessingStage::Enum::WithMergeableState)
    {
        if (!analysis_result.need_aggregate)
        {
            // What's the difference with selected_columns?
            return analysis_result.before_order_by->getResultColumns();
        }

        Block header = analysis_result.before_aggregation->getResultColumns();

        Block res;

        for (const auto & key : query_analyzer->aggregationKeys())
            res.insert({nullptr, header.getByName(key.name).type, key.name});

        for (const auto & aggregate : query_analyzer->aggregates())
        {
            size_t arguments_size = aggregate.argument_names.size();
            DataTypes argument_types(arguments_size);
            for (size_t j = 0; j < arguments_size; ++j)
                argument_types[j] = header.getByName(aggregate.argument_names[j]).type;

            DataTypePtr type = std::make_shared<DataTypeAggregateFunction>(aggregate.function, argument_types, aggregate.parameters);

            res.insert({nullptr, type, aggregate.column_name});
        }

        return res;
    }

    if (options.to_stage == QueryProcessingStage::Enum::WithMergeableStateAfterAggregation)
    {
        // What's the difference with selected_columns?
        return analysis_result.before_order_by->getResultColumns();
    }

    return analysis_result.final_projection->getResultColumns();
}

static Field getWithFillFieldValue(const ASTPtr & node, const Context & context)
{
    const auto & [field, type] = evaluateConstantExpression(node, context);

    if (!isColumnedAsNumber(type))
        throw Exception("Illegal type " + type->getName() + " of WITH FILL expression, must be numeric type", ErrorCodes::INVALID_WITH_FILL_EXPRESSION);

    return field;
}

static FillColumnDescription getWithFillDescription(const ASTOrderByElement & order_by_elem, const Context & context)
{
    FillColumnDescription descr;
    if (order_by_elem.fill_from)
        descr.fill_from = getWithFillFieldValue(order_by_elem.fill_from, context);
    if (order_by_elem.fill_to)
        descr.fill_to = getWithFillFieldValue(order_by_elem.fill_to, context);
    if (order_by_elem.fill_step)
        descr.fill_step = getWithFillFieldValue(order_by_elem.fill_step, context);
    else
        descr.fill_step = order_by_elem.direction;

    if (applyVisitor(FieldVisitorAccurateEquals(), descr.fill_step, Field{0}))
        throw Exception("WITH FILL STEP value cannot be zero", ErrorCodes::INVALID_WITH_FILL_EXPRESSION);

    if (order_by_elem.direction == 1)
    {
        if (applyVisitor(FieldVisitorAccurateLess(), descr.fill_step, Field{0}))
            throw Exception("WITH FILL STEP value cannot be negative for sorting in ascending direction",
                ErrorCodes::INVALID_WITH_FILL_EXPRESSION);

        if (!descr.fill_from.isNull() && !descr.fill_to.isNull() &&
            applyVisitor(FieldVisitorAccurateLess(), descr.fill_to, descr.fill_from))
        {
            throw Exception("WITH FILL TO value cannot be less than FROM value for sorting in ascending direction",
                ErrorCodes::INVALID_WITH_FILL_EXPRESSION);
        }
    }
    else
    {
        if (applyVisitor(FieldVisitorAccurateLess(), Field{0}, descr.fill_step))
            throw Exception("WITH FILL STEP value cannot be positive for sorting in descending direction",
                ErrorCodes::INVALID_WITH_FILL_EXPRESSION);

        if (!descr.fill_from.isNull() && !descr.fill_to.isNull() &&
            applyVisitor(FieldVisitorAccurateLess(), descr.fill_from, descr.fill_to))
        {
            throw Exception("WITH FILL FROM value cannot be less than TO value for sorting in descending direction",
                ErrorCodes::INVALID_WITH_FILL_EXPRESSION);
        }
    }

    return descr;
}

static SortDescription getSortDescription(const ASTSelectQuery & query, const Context & context)
{
    SortDescription order_descr;
    order_descr.reserve(query.orderBy()->children.size());
    for (const auto & elem : query.orderBy()->children)
    {
        String name = elem->children.front()->getColumnName();
        const auto & order_by_elem = elem->as<ASTOrderByElement &>();

        std::shared_ptr<Collator> collator;
        if (order_by_elem.collation)
            collator = std::make_shared<Collator>(order_by_elem.collation->as<ASTLiteral &>().value.get<String>());

        if (order_by_elem.with_fill)
        {
            FillColumnDescription fill_desc = getWithFillDescription(order_by_elem, context);
            order_descr.emplace_back(name, order_by_elem.direction,
                order_by_elem.nulls_direction, collator, true, fill_desc);
        }
        else
            order_descr.emplace_back(name, order_by_elem.direction, order_by_elem.nulls_direction, collator);
    }

    return order_descr;
}

static SortDescription getSortDescriptionFromGroupBy(const ASTSelectQuery & query)
{
    SortDescription order_descr;
    order_descr.reserve(query.groupBy()->children.size());

    for (const auto & elem : query.groupBy()->children)
    {
        String name = elem->getColumnName();
        order_descr.emplace_back(name, 1, 1);
    }

    return order_descr;
}

static UInt64 getLimitUIntValue(const ASTPtr & node, const Context & context, const std::string & expr)
{
    const auto & [field, type] = evaluateConstantExpression(node, context);

    if (!isNativeNumber(type))
        throw Exception("Illegal type " + type->getName() + " of " + expr + " expression, must be numeric type", ErrorCodes::INVALID_LIMIT_EXPRESSION);

    Field converted = convertFieldToType(field, DataTypeUInt64());
    if (converted.isNull())
        throw Exception("The value " + applyVisitor(FieldVisitorToString(), field) + " of " + expr + " expression is not representable as UInt64", ErrorCodes::INVALID_LIMIT_EXPRESSION);

    return converted.safeGet<UInt64>();
}


static std::pair<UInt64, UInt64> getLimitLengthAndOffset(const ASTSelectQuery & query, const Context & context)
{
    UInt64 length = 0;
    UInt64 offset = 0;

    if (query.limitLength())
    {
        length = getLimitUIntValue(query.limitLength(), context, "LIMIT");
        if (query.limitOffset() && length)
            offset = getLimitUIntValue(query.limitOffset(), context, "OFFSET");
    }
    else if (query.limitOffset())
        offset = getLimitUIntValue(query.limitOffset(), context, "OFFSET");
    return {length, offset};
}


static UInt64 getLimitForSorting(const ASTSelectQuery & query, const Context & context)
{
    /// Partial sort can be done if there is LIMIT but no DISTINCT or LIMIT BY, neither ARRAY JOIN.
    if (!query.distinct && !query.limitBy() && !query.limit_with_ties && !query.arrayJoinExpressionList() && query.limitLength())
    {
        auto [limit_length, limit_offset] = getLimitLengthAndOffset(query, context);
        if (limit_length > std::numeric_limits<UInt64>::max() - limit_offset)
            return 0;

        return limit_length + limit_offset;
    }
    return 0;
}


static bool hasWithTotalsInAnySubqueryInFromClause(const ASTSelectQuery & query)
{
    if (query.group_by_with_totals)
        return true;

    /** NOTE You can also check that the table in the subquery is distributed, and that it only looks at one shard.
      * In other cases, totals will be computed on the initiating server of the query, and it is not necessary to read the data to the end.
      */

    if (auto query_table = extractTableExpression(query, 0))
    {
        if (const auto * ast_union = query_table->as<ASTSelectWithUnionQuery>())
        {
            /// NOTE: Child of subquery can be ASTSelectWithUnionQuery or ASTSelectQuery,
            /// and after normalization, the height of the AST tree is at most 2
            for (const auto & elem : ast_union->list_of_selects->children)
            {
                if (const auto * child_union = elem->as<ASTSelectWithUnionQuery>())
                {
                    for (const auto & child_elem : child_union->list_of_selects->children)
                        if (hasWithTotalsInAnySubqueryInFromClause(child_elem->as<ASTSelectQuery &>()))
                            return true;
                }
                else
                {
                    if (hasWithTotalsInAnySubqueryInFromClause(elem->as<ASTSelectQuery &>()))
                        return true;
                }
            }
        }
    }

    return false;
}


void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, const BlockInputStreamPtr & prepared_input, std::optional<Pipe> prepared_pipe)
{
    /** Streams of data. When the query is executed in parallel, we have several data streams.
     *  If there is no GROUP BY, then perform all operations before ORDER BY and LIMIT in parallel, then
     *  if there is an ORDER BY, then glue the streams using ResizeProcessor, and then MergeSorting transforms,
     *  if not, then glue it using ResizeProcessor,
     *  then apply LIMIT.
     *  If there is GROUP BY, then we will perform all operations up to GROUP BY, inclusive, in parallel;
     *  a parallel GROUP BY will glue streams into one,
     *  then perform the remaining operations with one resulting stream.
     */

    /// Now we will compose block streams that perform the necessary actions.
    auto & query = getSelectQuery();
    const Settings & settings = context->getSettingsRef();
    auto & expressions = analysis_result;
    auto & subqueries_for_sets = query_analyzer->getSubqueriesForSets();
    bool intermediate_stage = false;
    bool to_aggregation_stage = false;
    bool from_aggregation_stage = false;

    if (expressions.filter_info)
    {
        if (!expressions.prewhere_info)
        {
            const bool does_storage_support_prewhere = !input && !input_pipe && storage && storage->supportsPrewhere();
            if (does_storage_support_prewhere && settings.optimize_move_to_prewhere)
            {
                /// Execute row level filter in prewhere as a part of "move to prewhere" optimization.
                expressions.prewhere_info = std::make_shared<PrewhereDAGInfo>(
                    std::move(expressions.filter_info->actions),
                    std::move(expressions.filter_info->column_name));
                expressions.prewhere_info->prewhere_actions->projectInput(false);
                expressions.prewhere_info->remove_prewhere_column = expressions.filter_info->do_remove_column;
                expressions.prewhere_info->need_filter = true;
                expressions.filter_info = nullptr;
            }
        }
        else
        {
            /// Add row level security actions to prewhere.
            expressions.prewhere_info->row_level_filter_actions = std::move(expressions.filter_info->actions);
            expressions.prewhere_info->row_level_column_name = std::move(expressions.filter_info->column_name);
            expressions.prewhere_info->row_level_filter_actions->projectInput(false);
            expressions.filter_info = nullptr;
        }
    }

    if (options.only_analyze)
    {
        auto read_nothing = std::make_unique<ReadNothingStep>(source_header);
        query_plan.addStep(std::move(read_nothing));

        if (expressions.filter_info)
        {
            auto row_level_security_step = std::make_unique<FilterStep>(
                query_plan.getCurrentDataStream(),
                expressions.filter_info->actions,
                expressions.filter_info->column_name,
                expressions.filter_info->do_remove_column);

            row_level_security_step->setStepDescription("Row-level security filter");
            query_plan.addStep(std::move(row_level_security_step));
        }

        if (expressions.prewhere_info)
        {
            if (expressions.prewhere_info->row_level_filter_actions)
            {
                auto row_level_filter_step = std::make_unique<FilterStep>(
                    query_plan.getCurrentDataStream(),
                    expressions.prewhere_info->row_level_filter_actions,
                    expressions.prewhere_info->row_level_column_name,
                    false);

                row_level_filter_step->setStepDescription("Row-level security filter (PREWHERE)");
                query_plan.addStep(std::move(row_level_filter_step));
            }

            auto prewhere_step = std::make_unique<FilterStep>(
                    query_plan.getCurrentDataStream(),
                    expressions.prewhere_info->prewhere_actions,
                    expressions.prewhere_info->prewhere_column_name,
                    expressions.prewhere_info->remove_prewhere_column);

            prewhere_step->setStepDescription("PREWHERE");
            query_plan.addStep(std::move(prewhere_step));

            // To remove additional columns in dry run
            // For example, sample column which can be removed in this stage
            if (expressions.prewhere_info->remove_columns_actions)
            {
                auto remove_columns = std::make_unique<ExpressionStep>(
                        query_plan.getCurrentDataStream(),
                        expressions.prewhere_info->remove_columns_actions);

                remove_columns->setStepDescription("Remove unnecessary columns after PREWHERE");
                query_plan.addStep(std::move(remove_columns));
            }
        }
    }
    else
    {
        if (prepared_input)
        {
            auto prepared_source_step = std::make_unique<ReadFromPreparedSource>(
                    Pipe(std::make_shared<SourceFromInputStream>(prepared_input)), context);
            query_plan.addStep(std::move(prepared_source_step));
        }
        else if (prepared_pipe)
        {
            auto prepared_source_step = std::make_unique<ReadFromPreparedSource>(std::move(*prepared_pipe), context);
            query_plan.addStep(std::move(prepared_source_step));
        }

        if (from_stage == QueryProcessingStage::WithMergeableState &&
            options.to_stage == QueryProcessingStage::WithMergeableState)
            intermediate_stage = true;

        /// Support optimize_distributed_group_by_sharding_key
        /// Is running on the initiating server during distributed processing?
        if (from_stage == QueryProcessingStage::WithMergeableStateAfterAggregation)
            from_aggregation_stage = true;
        /// Is running on remote servers during distributed processing?
        if (options.to_stage == QueryProcessingStage::WithMergeableStateAfterAggregation)
            to_aggregation_stage = true;

        /// Read the data from Storage. from_stage - to what stage the request was completed in Storage.
        executeFetchColumns(from_stage, query_plan);

        LOG_TRACE(log, "{} -> {}", QueryProcessingStage::toString(from_stage), QueryProcessingStage::toString(options.to_stage));
    }

    if (options.to_stage > QueryProcessingStage::FetchColumns)
    {
        /// Do I need to aggregate in a separate row rows that have not passed max_rows_to_group_by.
        bool aggregate_overflow_row =
            expressions.need_aggregate &&
            query.group_by_with_totals &&
            settings.max_rows_to_group_by &&
            settings.group_by_overflow_mode == OverflowMode::ANY &&
            settings.totals_mode != TotalsMode::AFTER_HAVING_EXCLUSIVE;

        /// Do I need to immediately finalize the aggregate functions after the aggregation?
        bool aggregate_final =
            expressions.need_aggregate &&
            options.to_stage > QueryProcessingStage::WithMergeableState &&
            !query.group_by_with_totals && !query.group_by_with_rollup && !query.group_by_with_cube;

        auto preliminary_sort = [&]()
        {
            /** For distributed query processing,
              *  if no GROUP, HAVING set,
              *  but there is an ORDER or LIMIT,
              *  then we will perform the preliminary sorting and LIMIT on the remote server.
              */
            if (!expressions.second_stage && !expressions.need_aggregate && !expressions.hasHaving())
            {
                if (expressions.has_order_by)
                    executeOrder(query_plan, query_info.input_order_info);

                if (expressions.has_order_by && query.limitLength())
                    executeDistinct(query_plan, false, expressions.selected_columns, true);

                if (expressions.hasLimitBy())
                {
                    executeExpression(query_plan, expressions.before_limit_by, "Before LIMIT BY");
                    executeLimitBy(query_plan);
                }

                if (query.limitLength())
                    executePreLimit(query_plan, true);
            }
        };

        if (intermediate_stage)
        {
            if (expressions.first_stage || expressions.second_stage)
                throw Exception("Query with intermediate stage cannot have any other stages", ErrorCodes::LOGICAL_ERROR);

            preliminary_sort();
            if (expressions.need_aggregate)
                executeMergeAggregated(query_plan, aggregate_overflow_row, aggregate_final);
        }
        if (from_aggregation_stage)
        {
            if (intermediate_stage || expressions.first_stage || expressions.second_stage)
                throw Exception("Query with after aggregation stage cannot have any other stages", ErrorCodes::LOGICAL_ERROR);
        }


        if (expressions.first_stage)
        {
            if (expressions.filter_info)
            {
                auto row_level_security_step = std::make_unique<FilterStep>(
                        query_plan.getCurrentDataStream(),
                        expressions.filter_info->actions,
                        expressions.filter_info->column_name,
                        expressions.filter_info->do_remove_column);

                row_level_security_step->setStepDescription("Row-level security filter");
                query_plan.addStep(std::move(row_level_security_step));
            }

            if (expressions.before_array_join)
            {
                QueryPlanStepPtr before_array_join_step = std::make_unique<ExpressionStep>(
                        query_plan.getCurrentDataStream(),
                        expressions.before_array_join);
                before_array_join_step->setStepDescription("Before ARRAY JOIN");
                query_plan.addStep(std::move(before_array_join_step));
            }

            if (expressions.array_join)
            {
                QueryPlanStepPtr array_join_step = std::make_unique<ArrayJoinStep>(
                        query_plan.getCurrentDataStream(),
                        expressions.array_join);

                array_join_step->setStepDescription("ARRAY JOIN");
                query_plan.addStep(std::move(array_join_step));
            }

            if (expressions.before_join)
            {
                QueryPlanStepPtr before_join_step = std::make_unique<ExpressionStep>(
                    query_plan.getCurrentDataStream(),
                    expressions.before_join);
                before_join_step->setStepDescription("Before JOIN");
                query_plan.addStep(std::move(before_join_step));
            }

            /// Optional step to convert key columns to common supertype.
            /// Columns with changed types will be returned to user,
            ///  so its only suitable for `USING` join.
            if (expressions.converting_join_columns)
            {
                QueryPlanStepPtr convert_join_step = std::make_unique<ExpressionStep>(
                    query_plan.getCurrentDataStream(),
                    expressions.converting_join_columns);
                convert_join_step->setStepDescription("Convert JOIN columns");
                query_plan.addStep(std::move(convert_join_step));
            }

            if (expressions.hasJoin())
            {
                JoinPtr join = expressions.join;

                QueryPlanStepPtr join_step = std::make_unique<JoinStep>(
                    query_plan.getCurrentDataStream(),
                    expressions.join);

                join_step->setStepDescription("JOIN");
                query_plan.addStep(std::move(join_step));

                if (expressions.join_has_delayed_stream)
                {
                    const Block & join_result_sample = query_plan.getCurrentDataStream().header;
                    auto stream = std::make_shared<LazyNonJoinedBlockInputStream>(*join, join_result_sample, settings.max_block_size);
                    auto source = std::make_shared<SourceFromInputStream>(std::move(stream));
                    auto add_non_joined_rows_step = std::make_unique<AddingDelayedSourceStep>(
                            query_plan.getCurrentDataStream(), std::move(source));

                    add_non_joined_rows_step->setStepDescription("Add non-joined rows after JOIN");
                    query_plan.addStep(std::move(add_non_joined_rows_step));
                }
            }

            if (expressions.hasWhere())
                executeWhere(query_plan, expressions.before_where, expressions.remove_where_filter);

            if (expressions.need_aggregate)
            {
                executeAggregation(query_plan, expressions.before_aggregation, aggregate_overflow_row, aggregate_final, query_info.input_order_info);
                /// We need to reset input order info, so that executeOrder can't use  it
                query_info.input_order_info.reset();
            }
            else
            {
                executeExpression(query_plan, expressions.before_window, "Before window functions");
                executeWindow(query_plan);
                executeExpression(query_plan, expressions.before_order_by, "Before ORDER BY");
                executeDistinct(query_plan, true, expressions.selected_columns, true);
            }

            preliminary_sort();

            // If there is no global subqueries, we can run subqueries only when receive them on server.
            if (!query_analyzer->hasGlobalSubqueries() && !subqueries_for_sets.empty())
                executeSubqueriesInSetsAndJoins(query_plan, subqueries_for_sets);
        }

        if (expressions.second_stage || from_aggregation_stage)
        {
            if (from_aggregation_stage)
            {
                /// No need to aggregate anything, since this was done on remote shards.
            }
            else if (expressions.need_aggregate)
            {
                /// If you need to combine aggregated results from multiple servers
                if (!expressions.first_stage)
                    executeMergeAggregated(query_plan, aggregate_overflow_row, aggregate_final);

                if (!aggregate_final)
                {
                    if (query.group_by_with_totals)
                    {
                        bool final = !query.group_by_with_rollup && !query.group_by_with_cube;
                        executeTotalsAndHaving(query_plan, expressions.hasHaving(), expressions.before_having, aggregate_overflow_row, final);
                    }

                    if (query.group_by_with_rollup)
                        executeRollupOrCube(query_plan, Modificator::ROLLUP);
                    else if (query.group_by_with_cube)
                        executeRollupOrCube(query_plan, Modificator::CUBE);

                    if ((query.group_by_with_rollup || query.group_by_with_cube) && expressions.hasHaving())
                    {
                        if (query.group_by_with_totals)
                            throw Exception("WITH TOTALS and WITH ROLLUP or CUBE are not supported together in presence of HAVING", ErrorCodes::NOT_IMPLEMENTED);
                        executeHaving(query_plan, expressions.before_having);
                    }
                }
                else if (expressions.hasHaving())
                    executeHaving(query_plan, expressions.before_having);

                executeExpression(query_plan, expressions.before_window,
                    "Before window functions");
                executeWindow(query_plan);
                executeExpression(query_plan, expressions.before_order_by, "Before ORDER BY");
                executeDistinct(query_plan, true, expressions.selected_columns, true);

            }
            else if (query.group_by_with_totals || query.group_by_with_rollup || query.group_by_with_cube)
                throw Exception("WITH TOTALS, ROLLUP or CUBE are not supported without aggregation", ErrorCodes::NOT_IMPLEMENTED);

            if (expressions.has_order_by)
            {
                /** If there is an ORDER BY for distributed query processing,
                  *  but there is no aggregation, then on the remote servers ORDER BY was made
                  *  - therefore, we merge the sorted streams from remote servers.
                  *
                  * Also in case of remote servers was process the query up to WithMergeableStateAfterAggregation
                  * (distributed_group_by_no_merge=2 or optimize_distributed_group_by_sharding_key=1 takes place),
                  * then merge the sorted streams is enough, since remote servers already did full ORDER BY.
                  */

                if (from_aggregation_stage)
                    executeMergeSorted(query_plan, "for ORDER BY");
                else if (!expressions.first_stage && !expressions.need_aggregate && !(query.group_by_with_totals && !aggregate_final))
                    executeMergeSorted(query_plan, "for ORDER BY");
                else    /// Otherwise, just sort.
                    executeOrder(query_plan, query_info.input_order_info);
            }

            /** Optimization - if there are several sources and there is LIMIT, then first apply the preliminary LIMIT,
              * limiting the number of rows in each up to `offset + limit`.
              */
            bool has_withfill = false;
            if (query.orderBy())
            {
                SortDescription order_descr = getSortDescription(query, *context);
                for (auto & desc : order_descr)
                    if (desc.with_fill)
                    {
                        has_withfill = true;
                        break;
                    }
            }

            bool has_prelimit = false;
            if (!to_aggregation_stage &&
                query.limitLength() && !query.limit_with_ties && !hasWithTotalsInAnySubqueryInFromClause(query) &&
                !query.arrayJoinExpressionList() && !query.distinct && !expressions.hasLimitBy() && !settings.extremes &&
                !has_withfill)
            {
                executePreLimit(query_plan, false);
                has_prelimit = true;
            }

            /** If there was more than one stream,
              * then DISTINCT needs to be performed once again after merging all streams.
              */
            if (query.distinct)
                executeDistinct(query_plan, false, expressions.selected_columns, false);

            if (expressions.hasLimitBy())
            {
                executeExpression(query_plan, expressions.before_limit_by, "Before LIMIT BY");
                executeLimitBy(query_plan);
            }

            executeWithFill(query_plan);

            /// If we have 'WITH TIES', we need execute limit before projection,
            /// because in that case columns from 'ORDER BY' are used.
            if (query.limit_with_ties)
            {
                executeLimit(query_plan);
                has_prelimit = true;
            }

            /// Projection not be done on the shards, since then initiator will not find column in blocks.
            /// (significant only for WithMergeableStateAfterAggregation).
            if (!to_aggregation_stage)
            {
                /// We must do projection after DISTINCT because projection may remove some columns.
                executeProjection(query_plan, expressions.final_projection);
            }

            /// Extremes are calculated before LIMIT, but after LIMIT BY. This is Ok.
            executeExtremes(query_plan);

            /// Limit is no longer needed if there is prelimit.
            if (!to_aggregation_stage && !has_prelimit)
                executeLimit(query_plan);

            if (!to_aggregation_stage)
                executeOffset(query_plan);
        }
    }

    if (!subqueries_for_sets.empty() && (expressions.hasHaving() || query_analyzer->hasGlobalSubqueries()))
        executeSubqueriesInSetsAndJoins(query_plan, subqueries_for_sets);
}

static StreamLocalLimits getLimitsForStorage(const Settings & settings, const SelectQueryOptions & options)
{
    StreamLocalLimits limits;
    limits.mode = LimitsMode::LIMITS_TOTAL;
    limits.size_limits = SizeLimits(settings.max_rows_to_read, settings.max_bytes_to_read,
                                    settings.read_overflow_mode);
    limits.speed_limits.max_execution_time = settings.max_execution_time;
    limits.timeout_overflow_mode = settings.timeout_overflow_mode;

    /** Quota and minimal speed restrictions are checked on the initiating server of the request, and not on remote servers,
      *  because the initiating server has a summary of the execution of the request on all servers.
      *
      * But limits on data size to read and maximum execution time are reasonable to check both on initiator and
      *  additionally on each remote server, because these limits are checked per block of data processed,
      *  and remote servers may process way more blocks of data than are received by initiator.
      *
      * The limits to throttle maximum execution speed is also checked on all servers.
      */
    if (options.to_stage == QueryProcessingStage::Complete)
    {
        limits.speed_limits.min_execution_rps = settings.min_execution_speed;
        limits.speed_limits.min_execution_bps = settings.min_execution_speed_bytes;
    }

    limits.speed_limits.max_execution_rps = settings.max_execution_speed;
    limits.speed_limits.max_execution_bps = settings.max_execution_speed_bytes;
    limits.speed_limits.timeout_before_checking_execution_speed = settings.timeout_before_checking_execution_speed;

    return limits;
}

void InterpreterSelectQuery::addEmptySourceToQueryPlan(QueryPlan & query_plan, const Block & source_header, const SelectQueryInfo & query_info)
{
    Pipe pipe(std::make_shared<NullSource>(source_header));

    if (query_info.prewhere_info)
    {
        auto & prewhere_info = *query_info.prewhere_info;

        if (prewhere_info.alias_actions)
        {
            pipe.addSimpleTransform([&](const Block & header)
            {
                return std::make_shared<ExpressionTransform>(
                    header,
                    prewhere_info.alias_actions);
            });
        }

        if (prewhere_info.row_level_filter)
        {
            pipe.addSimpleTransform([&](const Block & header)
            {
                return std::make_shared<FilterTransform>(
                    header,
                    prewhere_info.row_level_filter,
                    prewhere_info.row_level_column_name,
                    true);
            });
        }

        pipe.addSimpleTransform([&](const Block & header)
        {
            return std::make_shared<FilterTransform>(
                header,
                prewhere_info.prewhere_actions,
                prewhere_info.prewhere_column_name,
                prewhere_info.remove_prewhere_column);
        });

        // To remove additional columns
        // In some cases, we did not read any marks so that the pipeline.streams is empty
        // Thus, some columns in prewhere are not removed as expected
        // This leads to mismatched header in distributed table
        if (prewhere_info.remove_columns_actions)
        {
            pipe.addSimpleTransform([&](const Block & header)
            {
                return std::make_shared<ExpressionTransform>(
                    header,
                    prewhere_info.remove_columns_actions);
            });
        }
    }

    auto read_from_pipe = std::make_unique<ReadFromPreparedSource>(std::move(pipe));
    read_from_pipe->setStepDescription("Read from NullSource");
    query_plan.addStep(std::move(read_from_pipe));
}

void InterpreterSelectQuery::executeFetchColumns(QueryProcessingStage::Enum processing_stage, QueryPlan & query_plan)
{
    auto & query = getSelectQuery();
    const Settings & settings = context->getSettingsRef();
    auto & expressions = analysis_result;
    auto & prewhere_info = expressions.prewhere_info;
    auto & columns_to_remove_after_prewhere = expressions.columns_to_remove_after_prewhere;

    /// Optimization for trivial query like SELECT count() FROM table.
    bool optimize_trivial_count =
        syntax_analyzer_result->optimize_trivial_count
        && (settings.max_parallel_replicas <= 1)
        && storage
        && storage->getName() != "MaterializeMySQL"
        && !row_policy_filter
        && processing_stage == QueryProcessingStage::FetchColumns
        && query_analyzer->hasAggregation()
        && (query_analyzer->aggregates().size() == 1)
        && typeid_cast<AggregateFunctionCount *>(query_analyzer->aggregates()[0].function.get());

    if (optimize_trivial_count)
    {
        const auto & desc = query_analyzer->aggregates()[0];
        const auto & func = desc.function;
        std::optional<UInt64> num_rows{};

        if (!query.prewhere() && !query.where())
        {
            num_rows = storage->totalRows(settings);
        }
        else // It's possible to optimize count() given only partition predicates
        {
            SelectQueryInfo temp_query_info;
            temp_query_info.query = query_ptr;
            temp_query_info.syntax_analyzer_result = syntax_analyzer_result;
            temp_query_info.sets = query_analyzer->getPreparedSets();

            num_rows = storage->totalRowsByPartitionPredicate(temp_query_info, *context);
        }

        if (num_rows)
        {
            AggregateFunctionCount & agg_count = static_cast<AggregateFunctionCount &>(*func);

            /// We will process it up to "WithMergeableState".
            std::vector<char> state(agg_count.sizeOfData());
            AggregateDataPtr place = state.data();

            agg_count.create(place);
            SCOPE_EXIT_MEMORY_SAFE(agg_count.destroy(place));

            agg_count.set(place, *num_rows);

            auto column = ColumnAggregateFunction::create(func);
            column->insertFrom(place);

            Block header = analysis_result.before_aggregation->getResultColumns();
            size_t arguments_size = desc.argument_names.size();
            DataTypes argument_types(arguments_size);
            for (size_t j = 0; j < arguments_size; ++j)
                argument_types[j] = header.getByName(desc.argument_names[j]).type;

            Block block_with_count{
                {std::move(column), std::make_shared<DataTypeAggregateFunction>(func, argument_types, desc.parameters), desc.column_name}};

            auto istream = std::make_shared<OneBlockInputStream>(block_with_count);
            auto prepared_count = std::make_unique<ReadFromPreparedSource>(Pipe(std::make_shared<SourceFromInputStream>(istream)), context);
            prepared_count->setStepDescription("Optimized trivial count");
            query_plan.addStep(std::move(prepared_count));
            from_stage = QueryProcessingStage::WithMergeableState;
            analysis_result.first_stage = false;
            return;
        }
    }

    /// Actions to calculate ALIAS if required.
    ActionsDAGPtr alias_actions;

    if (storage)
    {
        /// Detect, if ALIAS columns are required for query execution
        auto alias_columns_required = false;
        const ColumnsDescription & storage_columns = metadata_snapshot->getColumns();
        for (const auto & column_name : required_columns)
        {
            auto column_default = storage_columns.getDefault(column_name);
            if (column_default && column_default->kind == ColumnDefaultKind::Alias)
            {
                alias_columns_required = true;
                break;
            }
        }

        /// There are multiple sources of required columns:
        ///  - raw required columns,
        ///  - columns deduced from ALIAS columns,
        ///  - raw required columns from PREWHERE,
        ///  - columns deduced from ALIAS columns from PREWHERE.
        /// PREWHERE is a special case, since we need to resolve it and pass directly to `IStorage::read()`
        /// before any other executions.
        if (alias_columns_required)
        {
            NameSet required_columns_from_prewhere; /// Set of all (including ALIAS) required columns for PREWHERE
            NameSet required_aliases_from_prewhere; /// Set of ALIAS required columns for PREWHERE

            if (prewhere_info)
            {
                /// Get some columns directly from PREWHERE expression actions
                auto prewhere_required_columns = prewhere_info->prewhere_actions->getRequiredColumns().getNames();
                required_columns_from_prewhere.insert(prewhere_required_columns.begin(), prewhere_required_columns.end());

                if (prewhere_info->row_level_filter_actions)
                {
                    auto row_level_required_columns = prewhere_info->row_level_filter_actions->getRequiredColumns().getNames();
                    required_columns_from_prewhere.insert(row_level_required_columns.begin(), row_level_required_columns.end());
                }
            }

            /// Expression, that contains all raw required columns
            ASTPtr required_columns_all_expr = std::make_shared<ASTExpressionList>();

            /// Expression, that contains raw required columns for PREWHERE
            ASTPtr required_columns_from_prewhere_expr = std::make_shared<ASTExpressionList>();

            /// Sort out already known required columns between expressions,
            /// also populate `required_aliases_from_prewhere`.
            for (const auto & column : required_columns)
            {
                ASTPtr column_expr;
                const auto column_default = storage_columns.getDefault(column);
                bool is_alias = column_default && column_default->kind == ColumnDefaultKind::Alias;
                if (is_alias)
                {
                    auto column_decl = storage_columns.get(column);
                    column_expr = column_default->expression->clone();
                    // recursive visit for alias to alias
                    replaceAliasColumnsInQuery(column_expr, metadata_snapshot->getColumns(), syntax_analyzer_result->getArrayJoinSourceNameSet(), *context);

                    column_expr = addTypeConversionToAST(std::move(column_expr), column_decl.type->getName(), metadata_snapshot->getColumns().getAll(), *context);
                    column_expr = setAlias(column_expr, column);
                }
                else
                    column_expr = std::make_shared<ASTIdentifier>(column);

                if (required_columns_from_prewhere.count(column))
                {
                    required_columns_from_prewhere_expr->children.emplace_back(std::move(column_expr));

                    if (is_alias)
                        required_aliases_from_prewhere.insert(column);
                }
                else
                    required_columns_all_expr->children.emplace_back(std::move(column_expr));
            }

            /// Columns, which we will get after prewhere and filter executions.
            NamesAndTypesList required_columns_after_prewhere;
            NameSet required_columns_after_prewhere_set;

            /// Collect required columns from prewhere expression actions.
            if (prewhere_info)
            {
                NameSet columns_to_remove(columns_to_remove_after_prewhere.begin(), columns_to_remove_after_prewhere.end());
                Block prewhere_actions_result = prewhere_info->prewhere_actions->getResultColumns();

                /// Populate required columns with the columns, added by PREWHERE actions and not removed afterwards.
                /// XXX: looks hacky that we already know which columns after PREWHERE we won't need for sure.
                for (const auto & column : prewhere_actions_result)
                {
                    if (prewhere_info->remove_prewhere_column && column.name == prewhere_info->prewhere_column_name)
                        continue;

                    if (columns_to_remove.count(column.name))
                        continue;

                    required_columns_all_expr->children.emplace_back(std::make_shared<ASTIdentifier>(column.name));
                    required_columns_after_prewhere.emplace_back(column.name, column.type);
                }

                required_columns_after_prewhere_set
                    = ext::map<NameSet>(required_columns_after_prewhere, [](const auto & it) { return it.name; });
            }

            auto syntax_result = TreeRewriter(*context).analyze(required_columns_all_expr, required_columns_after_prewhere, storage, metadata_snapshot);
            alias_actions = ExpressionAnalyzer(required_columns_all_expr, syntax_result, *context).getActionsDAG(true);

            /// The set of required columns could be added as a result of adding an action to calculate ALIAS.
            required_columns = alias_actions->getRequiredColumns().getNames();

            /// Do not remove prewhere filter if it is a column which is used as alias.
            if (prewhere_info && prewhere_info->remove_prewhere_column)
                if (required_columns.end()
                    != std::find(required_columns.begin(), required_columns.end(), prewhere_info->prewhere_column_name))
                    prewhere_info->remove_prewhere_column = false;

            /// Remove columns which will be added by prewhere.
            required_columns.erase(std::remove_if(required_columns.begin(), required_columns.end(), [&](const String & name)
            {
                return required_columns_after_prewhere_set.count(name) != 0;
            }), required_columns.end());

            if (prewhere_info)
            {
                /// Don't remove columns which are needed to be aliased.
                for (const auto & name : required_columns)
                    prewhere_info->prewhere_actions->tryRestoreColumn(name);

                auto analyzed_result
                    = TreeRewriter(*context).analyze(required_columns_from_prewhere_expr, metadata_snapshot->getColumns().getAllPhysical());
                prewhere_info->alias_actions
                    = ExpressionAnalyzer(required_columns_from_prewhere_expr, analyzed_result, *context).getActionsDAG(true, false);

                /// Add (physical?) columns required by alias actions.
                auto required_columns_from_alias = prewhere_info->alias_actions->getRequiredColumns();
                Block prewhere_actions_result = prewhere_info->prewhere_actions->getResultColumns();
                for (auto & column : required_columns_from_alias)
                    if (!prewhere_actions_result.has(column.name))
                        if (required_columns.end() == std::find(required_columns.begin(), required_columns.end(), column.name))
                            required_columns.push_back(column.name);

                /// Add physical columns required by prewhere actions.
                for (const auto & column : required_columns_from_prewhere)
                    if (required_aliases_from_prewhere.count(column) == 0)
                        if (required_columns.end() == std::find(required_columns.begin(), required_columns.end(), column))
                            required_columns.push_back(column);
            }
        }
    }

    /// Limitation on the number of columns to read.
    /// It's not applied in 'only_analyze' mode, because the query could be analyzed without removal of unnecessary columns.
    if (!options.only_analyze && settings.max_columns_to_read && required_columns.size() > settings.max_columns_to_read)
        throw Exception("Limit for number of columns to read exceeded. "
            "Requested: " + toString(required_columns.size())
            + ", maximum: " + settings.max_columns_to_read.toString(),
            ErrorCodes::TOO_MANY_COLUMNS);

    /// General limit for the number of threads.
    size_t max_threads_execute_query = settings.max_threads;

    /** With distributed query processing, almost no computations are done in the threads,
     *  but wait and receive data from remote servers.
     *  If we have 20 remote servers, and max_threads = 8, then it would not be very good
     *  connect and ask only 8 servers at a time.
     *  To simultaneously query more remote servers,
     *  instead of max_threads, max_distributed_connections is used.
     */
    bool is_remote = false;
    if (storage && storage->isRemote())
    {
        is_remote = true;
        max_threads_execute_query = max_streams = settings.max_distributed_connections;
    }

    UInt64 max_block_size = settings.max_block_size;

    auto [limit_length, limit_offset] = getLimitLengthAndOffset(query, *context);

    /** Optimization - if not specified DISTINCT, WHERE, GROUP, HAVING, ORDER, LIMIT BY, WITH TIES but LIMIT is specified, and limit + offset < max_block_size,
     *  then as the block size we will use limit + offset (not to read more from the table than requested),
     *  and also set the number of threads to 1.
     */
    if (!query.distinct
        && !query.limit_with_ties
        && !query.prewhere()
        && !query.where()
        && !query.groupBy()
        && !query.having()
        && !query.orderBy()
        && !query.limitBy()
        && query.limitLength()
        && !query_analyzer->hasAggregation()
        && limit_length <= std::numeric_limits<UInt64>::max() - limit_offset
        && limit_length + limit_offset < max_block_size)
    {
        max_block_size = std::max(UInt64(1), limit_length + limit_offset);
        max_threads_execute_query = max_streams = 1;
    }

    if (!max_block_size)
        throw Exception("Setting 'max_block_size' cannot be zero", ErrorCodes::PARAMETER_OUT_OF_BOUND);

    /// Initialize the initial data streams to which the query transforms are superimposed. Table or subquery or prepared input?
    if (query_plan.isInitialized())
    {
        /// Prepared input.
    }
    else if (interpreter_subquery)
    {
        /// Subquery.
        /// If we need less number of columns that subquery have - update the interpreter.
        if (required_columns.size() < source_header.columns())
        {
            ASTPtr subquery = extractTableExpression(query, 0);
            if (!subquery)
                throw Exception("Subquery expected", ErrorCodes::LOGICAL_ERROR);

            interpreter_subquery = std::make_unique<InterpreterSelectWithUnionQuery>(
                subquery, getSubqueryContext(*context),
                options.copy().subquery().noModify(), required_columns);

            if (query_analyzer->hasAggregation())
                interpreter_subquery->ignoreWithTotals();
        }

        interpreter_subquery->buildQueryPlan(query_plan);
        query_plan.addInterpreterContext(context);
    }
    else if (storage)
    {
        /// Table.
        if (max_streams == 0)
            throw Exception("Logical error: zero number of streams requested", ErrorCodes::LOGICAL_ERROR);

        /// If necessary, we request more sources than the number of threads - to distribute the work evenly over the threads.
        if (max_streams > 1 && !is_remote)
            max_streams *= settings.max_streams_to_max_threads_ratio;

        query_info.syntax_analyzer_result = syntax_analyzer_result;
        query_info.sets = query_analyzer->getPreparedSets();
        auto actions_settings = ExpressionActionsSettings::fromContext(*context);

        if (prewhere_info)
        {
            query_info.prewhere_info = std::make_shared<PrewhereInfo>();
            query_info.prewhere_info->prewhere_actions = std::make_shared<ExpressionActions>(prewhere_info->prewhere_actions, actions_settings);

            if (prewhere_info->row_level_filter_actions)
                query_info.prewhere_info->row_level_filter = std::make_shared<ExpressionActions>(prewhere_info->row_level_filter_actions, actions_settings);
            if (prewhere_info->alias_actions)
                query_info.prewhere_info->alias_actions = std::make_shared<ExpressionActions>(prewhere_info->alias_actions, actions_settings);
            if (prewhere_info->remove_columns_actions)
                query_info.prewhere_info->remove_columns_actions = std::make_shared<ExpressionActions>(prewhere_info->remove_columns_actions, actions_settings);

            query_info.prewhere_info->prewhere_column_name = prewhere_info->prewhere_column_name;
            query_info.prewhere_info->remove_prewhere_column = prewhere_info->remove_prewhere_column;
            query_info.prewhere_info->row_level_column_name = prewhere_info->row_level_column_name;
            query_info.prewhere_info->need_filter = prewhere_info->need_filter;
        }

        /// Create optimizer with prepared actions.
        /// Maybe we will need to calc input_order_info later, e.g. while reading from StorageMerge.
        if (analysis_result.optimize_read_in_order || analysis_result.optimize_aggregation_in_order)
        {
            if (analysis_result.optimize_read_in_order)
                query_info.order_optimizer = std::make_shared<ReadInOrderOptimizer>(
                    analysis_result.order_by_elements_actions,
                    getSortDescription(query, *context),
                    query_info.syntax_analyzer_result);
            else
                query_info.order_optimizer = std::make_shared<ReadInOrderOptimizer>(
                    analysis_result.group_by_elements_actions,
                    getSortDescriptionFromGroupBy(query),
                    query_info.syntax_analyzer_result);

            query_info.input_order_info = query_info.order_optimizer->getInputOrder(metadata_snapshot, *context);
        }

        StreamLocalLimits limits;
        SizeLimits leaf_limits;
        std::shared_ptr<const EnabledQuota> quota;


        /// Set the limits and quota for reading data, the speed and time of the query.
        if (!options.ignore_limits)
        {
            limits = getLimitsForStorage(settings, options);
            leaf_limits = SizeLimits(settings.max_rows_to_read_leaf, settings.max_bytes_to_read_leaf,
                                          settings.read_overflow_mode_leaf);
        }

        if (!options.ignore_quota && (options.to_stage == QueryProcessingStage::Complete))
            quota = context->getQuota();

        storage->read(query_plan, required_columns, metadata_snapshot,
                      query_info, *context, processing_stage, max_block_size, max_streams);

        if (context->hasQueryContext() && !options.is_internal)
        {
            auto local_storage_id = storage->getStorageID();
            context->getQueryContext().addQueryAccessInfo(
                backQuoteIfNeed(local_storage_id.getDatabaseName()), local_storage_id.getFullTableName(), required_columns);
        }

        /// Create step which reads from empty source if storage has no data.
        if (!query_plan.isInitialized())
        {
            auto header = metadata_snapshot->getSampleBlockForColumns(
                    required_columns, storage->getVirtuals(), storage->getStorageID());
            addEmptySourceToQueryPlan(query_plan, header, query_info);
        }

        /// Extend lifetime of context, table lock, storage. Set limits and quota.
        auto adding_limits_and_quota = std::make_unique<SettingQuotaAndLimitsStep>(
                query_plan.getCurrentDataStream(),
                storage,
                std::move(table_lock),
                limits,
                leaf_limits,
                std::move(quota),
                context);
        adding_limits_and_quota->setStepDescription("Set limits and quota after reading from storage");
        query_plan.addStep(std::move(adding_limits_and_quota));
    }
    else
        throw Exception("Logical error in InterpreterSelectQuery: nowhere to read", ErrorCodes::LOGICAL_ERROR);

    /// Specify the number of threads only if it wasn't specified in storage.
    ///
    /// But in case of remote query and prefer_localhost_replica=1 (default)
    /// The inner local query (that is done in the same process, without
    /// network interaction), it will setMaxThreads earlier and distributed
    /// query will not update it.
    if (!query_plan.getMaxThreads() || is_remote)
        query_plan.setMaxThreads(max_threads_execute_query);

    /// Aliases in table declaration.
    if (processing_stage == QueryProcessingStage::FetchColumns && alias_actions)
    {
        auto table_aliases = std::make_unique<ExpressionStep>(query_plan.getCurrentDataStream(), alias_actions);
        table_aliases->setStepDescription("Add table aliases");
        query_plan.addStep(std::move(table_aliases));
    }
}


void InterpreterSelectQuery::executeWhere(QueryPlan & query_plan, const ActionsDAGPtr & expression, bool remove_filter)
{
    auto where_step = std::make_unique<FilterStep>(
            query_plan.getCurrentDataStream(),
            expression,
            getSelectQuery().where()->getColumnName(),
            remove_filter);

    where_step->setStepDescription("WHERE");
    query_plan.addStep(std::move(where_step));
}


void InterpreterSelectQuery::executeAggregation(QueryPlan & query_plan, const ActionsDAGPtr & expression, bool overflow_row, bool final, InputOrderInfoPtr group_by_info)
{
    auto expression_before_aggregation = std::make_unique<ExpressionStep>(query_plan.getCurrentDataStream(), expression);
    expression_before_aggregation->setStepDescription("Before GROUP BY");
    query_plan.addStep(std::move(expression_before_aggregation));

    const auto & header_before_aggregation = query_plan.getCurrentDataStream().header;
    ColumnNumbers keys;
    for (const auto & key : query_analyzer->aggregationKeys())
        keys.push_back(header_before_aggregation.getPositionByName(key.name));

    AggregateDescriptions aggregates = query_analyzer->aggregates();
    for (auto & descr : aggregates)
        if (descr.arguments.empty())
            for (const auto & name : descr.argument_names)
                descr.arguments.push_back(header_before_aggregation.getPositionByName(name));

    const Settings & settings = context->getSettingsRef();

    Aggregator::Params params(header_before_aggregation, keys, aggregates,
                              overflow_row, settings.max_rows_to_group_by, settings.group_by_overflow_mode,
                              settings.group_by_two_level_threshold,
                              settings.group_by_two_level_threshold_bytes,
                              settings.max_bytes_before_external_group_by,
                              settings.empty_result_for_aggregation_by_empty_set,
                              context->getTemporaryVolume(),
                              settings.max_threads,
                              settings.min_free_disk_space_for_temporary_data);

    SortDescription group_by_sort_description;

    if (group_by_info && settings.optimize_aggregation_in_order)
        group_by_sort_description = getSortDescriptionFromGroupBy(getSelectQuery());
    else
        group_by_info = nullptr;

    auto merge_threads = max_streams;
    auto temporary_data_merge_threads = settings.aggregation_memory_efficient_merge_threads
                                        ? static_cast<size_t>(settings.aggregation_memory_efficient_merge_threads)
                                        : static_cast<size_t>(settings.max_threads);

    bool storage_has_evenly_distributed_read = storage && storage->hasEvenlyDistributedRead();

    auto aggregating_step = std::make_unique<AggregatingStep>(
            query_plan.getCurrentDataStream(),
            params, final,
            settings.max_block_size,
            merge_threads,
            temporary_data_merge_threads,
            storage_has_evenly_distributed_read,
            std::move(group_by_info),
            std::move(group_by_sort_description));

    query_plan.addStep(std::move(aggregating_step));
}


void InterpreterSelectQuery::executeMergeAggregated(QueryPlan & query_plan, bool overflow_row, bool final)
{
    const auto & header_before_merge = query_plan.getCurrentDataStream().header;

    ColumnNumbers keys;
    for (const auto & key : query_analyzer->aggregationKeys())
        keys.push_back(header_before_merge.getPositionByName(key.name));

    /** There are two modes of distributed aggregation.
      *
      * 1. In different threads read from the remote servers blocks.
      * Save all the blocks in the RAM. Merge blocks.
      * If the aggregation is two-level - parallelize to the number of buckets.
      *
      * 2. In one thread, read blocks from different servers in order.
      * RAM stores only one block from each server.
      * If the aggregation is a two-level aggregation, we consistently merge the blocks of each next level.
      *
      * The second option consumes less memory (up to 256 times less)
      *  in the case of two-level aggregation, which is used for large results after GROUP BY,
      *  but it can work more slowly.
      */

    const Settings & settings = context->getSettingsRef();

    Aggregator::Params params(header_before_merge, keys, query_analyzer->aggregates(), overflow_row, settings.max_threads);

    auto transform_params = std::make_shared<AggregatingTransformParams>(params, final);

    auto merging_aggregated = std::make_unique<MergingAggregatedStep>(
            query_plan.getCurrentDataStream(),
            std::move(transform_params),
            settings.distributed_aggregation_memory_efficient && storage && storage->isRemote(),
            settings.max_threads,
            settings.aggregation_memory_efficient_merge_threads);

    query_plan.addStep(std::move(merging_aggregated));
}


void InterpreterSelectQuery::executeHaving(QueryPlan & query_plan, const ActionsDAGPtr & expression)
{
    auto having_step = std::make_unique<FilterStep>(
            query_plan.getCurrentDataStream(),
            expression, getSelectQuery().having()->getColumnName(), false);

    having_step->setStepDescription("HAVING");
    query_plan.addStep(std::move(having_step));
}


void InterpreterSelectQuery::executeTotalsAndHaving(QueryPlan & query_plan, bool has_having, const ActionsDAGPtr & expression, bool overflow_row, bool final)
{
    const Settings & settings = context->getSettingsRef();

    auto totals_having_step = std::make_unique<TotalsHavingStep>(
            query_plan.getCurrentDataStream(),
            overflow_row, expression,
            has_having ? getSelectQuery().having()->getColumnName() : "",
            settings.totals_mode, settings.totals_auto_threshold, final);

    query_plan.addStep(std::move(totals_having_step));
}


void InterpreterSelectQuery::executeRollupOrCube(QueryPlan & query_plan, Modificator modificator)
{
    const auto & header_before_transform = query_plan.getCurrentDataStream().header;

    ColumnNumbers keys;

    for (const auto & key : query_analyzer->aggregationKeys())
        keys.push_back(header_before_transform.getPositionByName(key.name));

    const Settings & settings = context->getSettingsRef();

    Aggregator::Params params(header_before_transform, keys, query_analyzer->aggregates(),
                              false, settings.max_rows_to_group_by, settings.group_by_overflow_mode, 0, 0,
                              settings.max_bytes_before_external_group_by, settings.empty_result_for_aggregation_by_empty_set,
                              context->getTemporaryVolume(), settings.max_threads, settings.min_free_disk_space_for_temporary_data);

    auto transform_params = std::make_shared<AggregatingTransformParams>(params, true);

    QueryPlanStepPtr step;
    if (modificator == Modificator::ROLLUP)
        step = std::make_unique<RollupStep>(query_plan.getCurrentDataStream(), std::move(transform_params));
    else
        step = std::make_unique<CubeStep>(query_plan.getCurrentDataStream(), std::move(transform_params));

    query_plan.addStep(std::move(step));
}


void InterpreterSelectQuery::executeExpression(QueryPlan & query_plan, const ActionsDAGPtr & expression, const std::string & description)
{
    if (!expression)
    {
        return;
    }

    auto expression_step = std::make_unique<ExpressionStep>(query_plan.getCurrentDataStream(), expression);

    expression_step->setStepDescription(description);
    query_plan.addStep(std::move(expression_step));
}

static bool windowDescriptionComparator(const WindowDescription * _left,
    const WindowDescription * _right)
{
    const auto & left = _left->full_sort_description;
    const auto & right = _right->full_sort_description;

    for (size_t i = 0; i < std::min(left.size(), right.size()); ++i)
    {
        if (left[i].column_name < right[i].column_name)
        {
            return true;
        }
        else if (left[i].column_name > right[i].column_name)
        {
            return false;
        }
        else if (left[i].column_number < right[i].column_number)
        {
            return true;
        }
        else if (left[i].column_number > right[i].column_number)
        {
            return false;
        }
        else if (left[i].direction < right[i].direction)
        {
            return true;
        }
        else if (left[i].direction > right[i].direction)
        {
            return false;
        }
        else if (left[i].nulls_direction < right[i].nulls_direction)
        {
            return true;
        }
        else if (left[i].nulls_direction > right[i].nulls_direction)
        {
            return false;
        }

        assert(left[i] == right[i]);
    }

    // Note that we check the length last, because we want to put together the
    // sort orders that have common prefix but different length.
    return left.size() > right.size();
}

static bool sortIsPrefix(const WindowDescription & _prefix,
    const WindowDescription & _full)
{
    const auto & prefix = _prefix.full_sort_description;
    const auto & full = _full.full_sort_description;

    if (prefix.size() > full.size())
    {
        return false;
    }

    for (size_t i = 0; i < prefix.size(); ++i)
    {
        if (full[i] != prefix[i])
        {
            return false;
        }
    }

    return true;
}

void InterpreterSelectQuery::executeWindow(QueryPlan & query_plan)
{
    // Try to sort windows in such an order that the window with the longest
    // sort description goes first, and all window that use its prefixes follow.
    std::vector<const WindowDescription *> windows_sorted;
    for (const auto & [_, w] : query_analyzer->windowDescriptions())
    {
        windows_sorted.push_back(&w);
    }

    std::sort(windows_sorted.begin(), windows_sorted.end(),
        windowDescriptionComparator);

    const Settings & settings = context->getSettingsRef();
    for (size_t i = 0; i < windows_sorted.size(); ++i)
    {
        const auto & w = *windows_sorted[i];

        // We don't need to sort again if the input from previous window already
        // has suitable sorting. Also don't create sort steps when there are no
        // columns to sort by, because the sort nodes are confused by this. It
        // happens in case of `over ()`.
        if (!w.full_sort_description.empty()
            && (i == 0 || !sortIsPrefix(w, *windows_sorted[i - 1])))
        {
            auto partial_sorting = std::make_unique<PartialSortingStep>(
                query_plan.getCurrentDataStream(),
                w.full_sort_description,
                0 /* LIMIT */,
                SizeLimits(settings.max_rows_to_sort, settings.max_bytes_to_sort,
                    settings.sort_overflow_mode));
            partial_sorting->setStepDescription("Sort each block for window '"
                + w.window_name + "'");
            query_plan.addStep(std::move(partial_sorting));

            auto merge_sorting_step = std::make_unique<MergeSortingStep>(
                query_plan.getCurrentDataStream(),
                w.full_sort_description,
                settings.max_block_size,
                0 /* LIMIT */,
                settings.max_bytes_before_remerge_sort,
                settings.remerge_sort_lowered_memory_bytes_ratio,
                settings.max_bytes_before_external_sort,
                context->getTemporaryVolume(),
                settings.min_free_disk_space_for_temporary_data);
            merge_sorting_step->setStepDescription(
                "Merge sorted blocks for window '" + w.window_name + "'");
            query_plan.addStep(std::move(merge_sorting_step));

            // First MergeSorted, now MergingSorted.
            auto merging_sorted = std::make_unique<MergingSortedStep>(
                query_plan.getCurrentDataStream(),
                w.full_sort_description,
                settings.max_block_size,
                0 /* LIMIT */);
            merging_sorted->setStepDescription(
                "Merge sorted streams for window '" + w.window_name + "'");
            query_plan.addStep(std::move(merging_sorted));
        }

        auto window_step = std::make_unique<WindowStep>(
            query_plan.getCurrentDataStream(),
            w,
            w.window_functions);
        window_step->setStepDescription("Window step for window '"
            + w.window_name + "'");

        query_plan.addStep(std::move(window_step));
    }
}


void InterpreterSelectQuery::executeOrderOptimized(QueryPlan & query_plan, InputOrderInfoPtr input_sorting_info, UInt64 limit, SortDescription & output_order_descr)
{
    const Settings & settings = context->getSettingsRef();

    auto finish_sorting_step = std::make_unique<FinishSortingStep>(
            query_plan.getCurrentDataStream(),
            input_sorting_info->order_key_prefix_descr,
            output_order_descr,
            settings.max_block_size,
            limit);

    query_plan.addStep(std::move(finish_sorting_step));
}

void InterpreterSelectQuery::executeOrder(QueryPlan & query_plan, InputOrderInfoPtr input_sorting_info)
{
    auto & query = getSelectQuery();
    SortDescription output_order_descr = getSortDescription(query, *context);
    UInt64 limit = getLimitForSorting(query, *context);

    if (input_sorting_info)
    {
        /* Case of sorting with optimization using sorting key.
         * We have several threads, each of them reads batch of parts in direct
         *  or reverse order of sorting key using one input stream per part
         *  and then merge them into one sorted stream.
         * At this stage we merge per-thread streams into one.
         */
        executeOrderOptimized(query_plan, input_sorting_info, limit, output_order_descr);
        return;
    }

    const Settings & settings = context->getSettingsRef();

    auto partial_sorting = std::make_unique<PartialSortingStep>(
            query_plan.getCurrentDataStream(),
            output_order_descr,
            limit,
            SizeLimits(settings.max_rows_to_sort, settings.max_bytes_to_sort, settings.sort_overflow_mode));

    partial_sorting->setStepDescription("Sort each block for ORDER BY");
    query_plan.addStep(std::move(partial_sorting));

    /// Merge the sorted blocks.
    auto merge_sorting_step = std::make_unique<MergeSortingStep>(
            query_plan.getCurrentDataStream(),
            output_order_descr,
            settings.max_block_size,
            limit,
            settings.max_bytes_before_remerge_sort,
            settings.remerge_sort_lowered_memory_bytes_ratio,
            settings.max_bytes_before_external_sort,
            context->getTemporaryVolume(),
            settings.min_free_disk_space_for_temporary_data);

    merge_sorting_step->setStepDescription("Merge sorted blocks for ORDER BY");
    query_plan.addStep(std::move(merge_sorting_step));

    /// If there are several streams, we merge them into one
    executeMergeSorted(query_plan, output_order_descr, limit, "for ORDER BY");
}


void InterpreterSelectQuery::executeMergeSorted(QueryPlan & query_plan, const std::string & description)
{
    auto & query = getSelectQuery();
    SortDescription order_descr = getSortDescription(query, *context);
    UInt64 limit = getLimitForSorting(query, *context);

    executeMergeSorted(query_plan, order_descr, limit, description);
}

void InterpreterSelectQuery::executeMergeSorted(QueryPlan & query_plan, const SortDescription & sort_description, UInt64 limit, const std::string & description)
{
    const Settings & settings = context->getSettingsRef();

    auto merging_sorted = std::make_unique<MergingSortedStep>(
            query_plan.getCurrentDataStream(),
            sort_description,
            settings.max_block_size, limit);

    merging_sorted->setStepDescription("Merge sorted streams " + description);
    query_plan.addStep(std::move(merging_sorted));
}


void InterpreterSelectQuery::executeProjection(QueryPlan & query_plan, const ActionsDAGPtr & expression)
{
    auto projection_step = std::make_unique<ExpressionStep>(query_plan.getCurrentDataStream(), expression);
    projection_step->setStepDescription("Projection");
    query_plan.addStep(std::move(projection_step));
}


void InterpreterSelectQuery::executeDistinct(QueryPlan & query_plan, bool before_order, Names columns, bool pre_distinct)
{
    auto & query = getSelectQuery();
    if (query.distinct)
    {
        const Settings & settings = context->getSettingsRef();

        auto [limit_length, limit_offset] = getLimitLengthAndOffset(query, *context);
        UInt64 limit_for_distinct = 0;

        /// If after this stage of DISTINCT ORDER BY is not executed,
        /// then you can get no more than limit_length + limit_offset of different rows.
        if ((!query.orderBy() || !before_order) && limit_length <= std::numeric_limits<UInt64>::max() - limit_offset)
            limit_for_distinct = limit_length + limit_offset;

        SizeLimits limits(settings.max_rows_in_distinct, settings.max_bytes_in_distinct, settings.distinct_overflow_mode);

        auto distinct_step = std::make_unique<DistinctStep>(
                query_plan.getCurrentDataStream(),
                limits, limit_for_distinct, columns, pre_distinct);

        if (pre_distinct)
            distinct_step->setStepDescription("Preliminary DISTINCT");

        query_plan.addStep(std::move(distinct_step));
    }
}


/// Preliminary LIMIT - is used in every source, if there are several sources, before they are combined.
void InterpreterSelectQuery::executePreLimit(QueryPlan & query_plan, bool do_not_skip_offset)
{
    auto & query = getSelectQuery();
    /// If there is LIMIT
    if (query.limitLength())
    {
        auto [limit_length, limit_offset] = getLimitLengthAndOffset(query, *context);

        if (do_not_skip_offset)
        {
            if (limit_length > std::numeric_limits<UInt64>::max() - limit_offset)
                return;

            limit_length += limit_offset;
            limit_offset = 0;
        }

        auto limit = std::make_unique<LimitStep>(query_plan.getCurrentDataStream(), limit_length, limit_offset);
        limit->setStepDescription("preliminary LIMIT");
        query_plan.addStep(std::move(limit));
    }
}


void InterpreterSelectQuery::executeLimitBy(QueryPlan & query_plan)
{
    auto & query = getSelectQuery();
    if (!query.limitByLength() || !query.limitBy())
        return;

    Names columns;
    for (const auto & elem : query.limitBy()->children)
        columns.emplace_back(elem->getColumnName());

    UInt64 length = getLimitUIntValue(query.limitByLength(), *context, "LIMIT");
    UInt64 offset = (query.limitByOffset() ? getLimitUIntValue(query.limitByOffset(), *context, "OFFSET") : 0);

    auto limit_by = std::make_unique<LimitByStep>(query_plan.getCurrentDataStream(), length, offset, columns);
    query_plan.addStep(std::move(limit_by));
}

void InterpreterSelectQuery::executeWithFill(QueryPlan & query_plan)
{
    auto & query = getSelectQuery();
    if (query.orderBy())
    {
        SortDescription order_descr = getSortDescription(query, *context);
        SortDescription fill_descr;
        for (auto & desc : order_descr)
        {
            if (desc.with_fill)
                fill_descr.push_back(desc);
        }

        if (fill_descr.empty())
            return;

        auto filling_step = std::make_unique<FillingStep>(query_plan.getCurrentDataStream(), std::move(fill_descr));
        query_plan.addStep(std::move(filling_step));
    }
}


void InterpreterSelectQuery::executeLimit(QueryPlan & query_plan)
{
    auto & query = getSelectQuery();
    /// If there is LIMIT
    if (query.limitLength())
    {
        /** Rare case:
          *  if there is no WITH TOTALS and there is a subquery in FROM, and there is WITH TOTALS on one of the levels,
          *  then when using LIMIT, you should read the data to the end, rather than cancel the query earlier,
          *  because if you cancel the query, we will not get `totals` data from the remote server.
          *
          * Another case:
          *  if there is WITH TOTALS and there is no ORDER BY, then read the data to the end,
          *  otherwise TOTALS is counted according to incomplete data.
          */
        bool always_read_till_end = false;

        if (query.group_by_with_totals && !query.orderBy())
            always_read_till_end = true;

        if (!query.group_by_with_totals && hasWithTotalsInAnySubqueryInFromClause(query))
            always_read_till_end = true;

        UInt64 limit_length;
        UInt64 limit_offset;
        std::tie(limit_length, limit_offset) = getLimitLengthAndOffset(query, *context);

        SortDescription order_descr;
        if (query.limit_with_ties)
        {
            if (!query.orderBy())
                throw Exception("LIMIT WITH TIES without ORDER BY", ErrorCodes::LOGICAL_ERROR);
            order_descr = getSortDescription(query, *context);
        }

        auto limit = std::make_unique<LimitStep>(
                query_plan.getCurrentDataStream(),
                limit_length, limit_offset, always_read_till_end, query.limit_with_ties, order_descr);

        if (query.limit_with_ties)
            limit->setStepDescription("LIMIT WITH TIES");

        query_plan.addStep(std::move(limit));
    }
}


void InterpreterSelectQuery::executeOffset(QueryPlan & query_plan)
{
    auto & query = getSelectQuery();
    /// If there is not a LIMIT but an offset
    if (!query.limitLength() && query.limitOffset())
    {
        UInt64 limit_length;
        UInt64 limit_offset;
        std::tie(limit_length, limit_offset) = getLimitLengthAndOffset(query, *context);

        auto offsets_step = std::make_unique<OffsetStep>(query_plan.getCurrentDataStream(), limit_offset);
        query_plan.addStep(std::move(offsets_step));
    }
}

void InterpreterSelectQuery::executeExtremes(QueryPlan & query_plan)
{
    if (!context->getSettingsRef().extremes)
        return;

    auto extremes_step = std::make_unique<ExtremesStep>(query_plan.getCurrentDataStream());
    query_plan.addStep(std::move(extremes_step));
}

void InterpreterSelectQuery::executeSubqueriesInSetsAndJoins(QueryPlan & query_plan, SubqueriesForSets & subqueries_for_sets)
{
    if (query_info.input_order_info)
        executeMergeSorted(query_plan, query_info.input_order_info->order_key_prefix_descr, 0, "before creating sets for subqueries and joins");

    const Settings & settings = context->getSettingsRef();

    SizeLimits limits(settings.max_rows_to_transfer, settings.max_bytes_to_transfer, settings.transfer_overflow_mode);
    addCreatingSetsStep(query_plan, std::move(subqueries_for_sets), limits, *context);
}


void InterpreterSelectQuery::ignoreWithTotals()
{
    getSelectQuery().group_by_with_totals = false;
}


void InterpreterSelectQuery::initSettings()
{
    auto & query = getSelectQuery();
    if (query.settings())
        InterpreterSetQuery(query.settings(), *context).executeForCurrentContext();
}

}