ClickHouse/src/Interpreters/TreeRewriter.cpp

#include <Core/Settings.h>
#include <Core/Defines.h>
#include <Core/NamesAndTypes.h>

#include <Interpreters/TreeRewriter.h>
#include <Interpreters/LogicalExpressionsOptimizer.h>
#include <Interpreters/QueryAliasesVisitor.h>
#include <Interpreters/ArrayJoinedColumnsVisitor.h>
#include <Interpreters/TranslateQualifiedNamesVisitor.h>
#include <Interpreters/Context.h>
#include <Interpreters/FunctionNameNormalizer.h>
#include <Interpreters/MarkTableIdentifiersVisitor.h>
#include <Interpreters/QueryNormalizer.h>
#include <Interpreters/ExecuteScalarSubqueriesVisitor.h>
#include <Interpreters/CollectJoinOnKeysVisitor.h>
#include <Interpreters/RequiredSourceColumnsVisitor.h>
#include <Interpreters/GetAggregatesVisitor.h>
#include <Interpreters/TableJoin.h>
#include <Interpreters/ExpressionActions.h> /// getSmallestColumn()
#include <Interpreters/getTableExpressions.h>
#include <Interpreters/TreeOptimizer.h>
#include <Interpreters/replaceAliasColumnsInQuery.h>

#include <Parsers/ASTExpressionList.h>
#include <Parsers/ASTFunction.h>
#include <Parsers/ASTSelectQuery.h>
#include <Parsers/ASTTablesInSelectQuery.h>
#include <Parsers/queryToString.h>
#include <Parsers/ASTLiteral.h>

#include <DataTypes/NestedUtils.h>
#include <DataTypes/DataTypeNullable.h>

#include <IO/WriteHelpers.h>
#include <IO/WriteBufferFromOStream.h>
#include <Storages/IStorage.h>

#include <AggregateFunctions/AggregateFunctionFactory.h>

namespace DB
{

namespace ErrorCodes
{
    extern const int EMPTY_LIST_OF_COLUMNS_QUERIED;
    extern const int EMPTY_NESTED_TABLE;
    extern const int EXPECTED_ALL_OR_ANY;
    extern const int INVALID_JOIN_ON_EXPRESSION;
    extern const int LOGICAL_ERROR;
    extern const int NOT_IMPLEMENTED;
    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
    extern const int UNKNOWN_IDENTIFIER;
}

namespace
{

using LogAST = DebugASTLog<false>; /// set to true to enable logs

/// Select implementation of a function based on settings.
/// Important that it is done as query rewrite. It means rewritten query
///  will be sent to remote servers during distributed query execution,
///  and on all remote servers, function implementation will be same.
template <char const * func_name>
struct CustomizeFunctionsData
{
    using TypeToVisit = ASTFunction;

    const String & customized_func_name;

    void visit(ASTFunction & func, ASTPtr &) const
    {
        if (Poco::toLower(func.name) == func_name)
        {
            func.name = customized_func_name;
        }
    }
};

char countdistinct[] = "countdistinct";
using CustomizeCountDistinctVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsData<countdistinct>>, true>;

char countifdistinct[] = "countifdistinct";
using CustomizeCountIfDistinctVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsData<countifdistinct>>, true>;

char in[] = "in";
using CustomizeInVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsData<in>>, true>;

char notIn[] = "notin";
using CustomizeNotInVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsData<notIn>>, true>;

char globalIn[] = "globalin";
using CustomizeGlobalInVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsData<globalIn>>, true>;

char globalNotIn[] = "globalnotin";
using CustomizeGlobalNotInVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsData<globalNotIn>>, true>;

template <char const * func_suffix>
struct CustomizeFunctionsSuffixData
{
    using TypeToVisit = ASTFunction;

    const String & customized_func_suffix;

    void visit(ASTFunction & func, ASTPtr &) const
    {
        if (endsWith(Poco::toLower(func.name), func_suffix))
        {
            size_t prefix_len = func.name.length() - strlen(func_suffix);
            func.name = func.name.substr(0, prefix_len) + customized_func_suffix;
        }
    }
};

/// Swap 'if' and 'distinct' suffixes to make execution more optimal.
char ifDistinct[] = "ifdistinct";
using CustomizeIfDistinctVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeFunctionsSuffixData<ifDistinct>>, true>;

/// Used to rewrite all aggregate functions to add -OrNull suffix to them if setting `aggregate_functions_null_for_empty` is set.
struct CustomizeAggregateFunctionsSuffixData
{
    using TypeToVisit = ASTFunction;

    const String & customized_func_suffix;

    void visit(ASTFunction & func, ASTPtr &) const
    {
        const auto & instance = AggregateFunctionFactory::instance();
        if (instance.isAggregateFunctionName(func.name) && !endsWith(func.name, customized_func_suffix))
        {
            auto properties = instance.tryGetProperties(func.name);
            if (properties && !properties->returns_default_when_only_null)
            {
                func.name += customized_func_suffix;
            }
        }
    }
};

// Used to rewrite aggregate functions with -OrNull suffix in some cases, such as sumIfOrNull, we should rewrite to sumOrNullIf
struct CustomizeAggregateFunctionsMoveSuffixData
{
    using TypeToVisit = ASTFunction;

    const String & customized_func_suffix;

    String moveSuffixAhead(const String & name) const
    {
        auto prefix = name.substr(0, name.size() - customized_func_suffix.size());

        auto prefix_size = prefix.size();

        if (endsWith(prefix, "MergeState"))
            return prefix.substr(0, prefix_size - 10) + customized_func_suffix + "MergeState";

        if (endsWith(prefix, "Merge"))
            return prefix.substr(0, prefix_size - 5) + customized_func_suffix + "Merge";

        if (endsWith(prefix, "State"))
            return prefix.substr(0, prefix_size - 5) + customized_func_suffix + "State";

        if (endsWith(prefix, "If"))
            return prefix.substr(0, prefix_size - 2) + customized_func_suffix + "If";

        return name;
    }

    void visit(ASTFunction & func, ASTPtr &) const
    {
        const auto & instance = AggregateFunctionFactory::instance();
        if (instance.isAggregateFunctionName(func.name))
        {
            if (endsWith(func.name, customized_func_suffix))
            {
                auto properties = instance.tryGetProperties(func.name);
                if (properties && !properties->returns_default_when_only_null)
                {
                    func.name = moveSuffixAhead(func.name);
                }
            }
        }
    }
};

struct FuseSumCountAggregates
{
    std::vector<ASTFunction *> sums {};
    std::vector<ASTFunction *> counts {};
    std::vector<ASTFunction *> avgs {};

    void addFuncNode(ASTFunction * func)
    {
        if (func->name == "sum")
            sums.push_back(func);
        else if (func->name == "count")
            counts.push_back(func);
        else
        {
            assert(func->name == "avg");
            avgs.push_back(func);
        }
    }

    bool canBeFused() const
    {
        // Need at least two different kinds of functions to fuse.
        if (sums.empty() && counts.empty())
            return false;
        if (sums.empty() && avgs.empty())
            return false;
        if (counts.empty() && avgs.empty())
            return false;
        return true;
    }
};

struct FuseSumCountAggregatesVisitorData
{
    using TypeToVisit = ASTFunction;

    std::unordered_map<String, FuseSumCountAggregates> fuse_map;

    void visit(ASTFunction & func, ASTPtr &)
    {
        if (func.name == "sum" || func.name == "avg" || func.name == "count")
        {
            if (func.arguments->children.empty())
                return;

            // Probably we can extend it to match count() for non-nullable argument
            // to sum/avg with any other argument. Now we require strict match.
            const auto argument = func.arguments->children.at(0)->getColumnName();
            auto it = fuse_map.find(argument);
            if (it != fuse_map.end())
            {
                it->second.addFuncNode(&func);
            }
            else
            {
                FuseSumCountAggregates funcs{};
                funcs.addFuncNode(&func);
                fuse_map[argument] = funcs;
            }
        }
    }
};

using CustomizeAggregateFunctionsOrNullVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeAggregateFunctionsSuffixData>, true>;
using CustomizeAggregateFunctionsMoveOrNullVisitor = InDepthNodeVisitor<OneTypeMatcher<CustomizeAggregateFunctionsMoveSuffixData>, true>;
using FuseSumCountAggregatesVisitor = InDepthNodeVisitor<OneTypeMatcher<FuseSumCountAggregatesVisitorData>, true>;

/// Translate qualified names such as db.table.column, table.column, table_alias.column to names' normal form.
/// Expand asterisks and qualified asterisks with column names.
/// There would be columns in normal form & column aliases after translation. Column & column alias would be normalized in QueryNormalizer.
void translateQualifiedNames(ASTPtr & query, const ASTSelectQuery & select_query, const NameSet & source_columns_set,
                             const TablesWithColumns & tables_with_columns)
{
    LogAST log;
    TranslateQualifiedNamesVisitor::Data visitor_data(source_columns_set, tables_with_columns);
    TranslateQualifiedNamesVisitor visitor(visitor_data, log.stream());
    visitor.visit(query);

    /// This may happen after expansion of COLUMNS('regexp').
    if (select_query.select()->children.empty())
        throw Exception("Empty list of columns in SELECT query", ErrorCodes::EMPTY_LIST_OF_COLUMNS_QUERIED);
}

// Replaces one avg/sum/count function with an appropriate expression with
// sumCount().
void replaceWithSumCount(String column_name, ASTFunction & func)
{
    auto func_base = makeASTFunction("sumCount", std::make_shared<ASTIdentifier>(column_name));
    auto exp_list = std::make_shared<ASTExpressionList>();
    if (func.name == "sum" || func.name == "count")
    {
        /// Rewrite "sum" to sumCount().1, rewrite "count" to sumCount().2
        UInt8 idx = (func.name == "sum" ? 1 : 2);
        func.name = "tupleElement";
        exp_list->children.push_back(func_base);
        exp_list->children.push_back(std::make_shared<ASTLiteral>(idx));
    }
    else
    {
        /// Rewrite "avg" to sumCount().1 / sumCount().2
        auto new_arg1 = makeASTFunction("tupleElement", func_base, std::make_shared<ASTLiteral>(UInt8(1)));
        auto new_arg2 = makeASTFunction("tupleElement", func_base, std::make_shared<ASTLiteral>(UInt8(2)));
        func.name = "divide";
        exp_list->children.push_back(new_arg1);
        exp_list->children.push_back(new_arg2);
    }
    func.arguments = exp_list;
    func.children.push_back(func.arguments);
}

void fuseSumCountAggregates(std::unordered_map<String, FuseSumCountAggregates> & fuse_map)
{
    for (auto & it : fuse_map)
    {
        if (it.second.canBeFused())
        {
            for (auto & func: it.second.sums)
                replaceWithSumCount(it.first, *func);
            for (auto & func: it.second.avgs)
                replaceWithSumCount(it.first, *func);
            for (auto & func: it.second.counts)
                replaceWithSumCount(it.first, *func);
        }
    }
}

bool hasArrayJoin(const ASTPtr & ast)
{
    if (const ASTFunction * function = ast->as<ASTFunction>())
        if (function->name == "arrayJoin")
            return true;

    for (const auto & child : ast->children)
        if (!child->as<ASTSelectQuery>() && hasArrayJoin(child))
            return true;

    return false;
}

/// Keep number of columns for 'GLOBAL IN (SELECT 1 AS a, a)'
void renameDuplicatedColumns(const ASTSelectQuery * select_query)
{
    ASTs & elements = select_query->select()->children;

    std::set<String> all_column_names;
    std::set<String> assigned_column_names;

    for (auto & expr : elements)
        all_column_names.insert(expr->getAliasOrColumnName());

    for (auto & expr : elements)
    {
        auto name = expr->getAliasOrColumnName();

        if (!assigned_column_names.insert(name).second)
        {
            size_t i = 1;
            while (all_column_names.end() != all_column_names.find(name + "_" + toString(i)))
                ++i;

            name = name + "_" + toString(i);
            expr = expr->clone();   /// Cancels fuse of the same expressions in the tree.
            expr->setAlias(name);

            all_column_names.insert(name);
            assigned_column_names.insert(name);
        }
    }
}

/// Sometimes we have to calculate more columns in SELECT clause than will be returned from query.
/// This is the case when we have DISTINCT or arrayJoin: we require more columns in SELECT even if we need less columns in result.
/// Also we have to remove duplicates in case of GLOBAL subqueries. Their results are placed into tables so duplicates are impossible.
void removeUnneededColumnsFromSelectClause(const ASTSelectQuery * select_query, const Names & required_result_columns, bool remove_dups)
{
    ASTs & elements = select_query->select()->children;

    std::map<String, size_t> required_columns_with_duplicate_count;

    if (!required_result_columns.empty())
    {
        /// Some columns may be queried multiple times, like SELECT x, y, y FROM table.
        for (const auto & name : required_result_columns)
        {
            if (remove_dups)
                required_columns_with_duplicate_count[name] = 1;
            else
                ++required_columns_with_duplicate_count[name];
        }
    }
    else if (remove_dups)
    {
        /// Even if we have no requirements there could be duplicates cause of asterisks. SELECT *, t.*
        for (const auto & elem : elements)
            required_columns_with_duplicate_count.emplace(elem->getAliasOrColumnName(), 1);
    }
    else
        return;

    ASTs new_elements;
    new_elements.reserve(elements.size());

    for (const auto & elem : elements)
    {
        String name = elem->getAliasOrColumnName();

        auto it = required_columns_with_duplicate_count.find(name);
        if (required_columns_with_duplicate_count.end() != it && it->second)
        {
            new_elements.push_back(elem);
            --it->second;
        }
        else if (select_query->distinct || hasArrayJoin(elem))
        {
            new_elements.push_back(elem);
        }
        else
        {
            ASTFunction * func = elem->as<ASTFunction>();

            /// Never remove untuple. It's result column may be in required columns.
            /// It is not easy to analyze untuple here, because types were not calculated yes.
            if (func && func->name == "untuple")
                new_elements.push_back(elem);
        }
    }

    elements = std::move(new_elements);
}

/// Replacing scalar subqueries with constant values.
void executeScalarSubqueries(ASTPtr & query, ContextPtr context, size_t subquery_depth, Scalars & scalars, bool only_analyze)
{
    LogAST log;
    ExecuteScalarSubqueriesVisitor::Data visitor_data{WithContext{context}, subquery_depth, scalars, only_analyze};
    ExecuteScalarSubqueriesVisitor(visitor_data, log.stream()).visit(query);
}

void getArrayJoinedColumns(ASTPtr & query, TreeRewriterResult & result, const ASTSelectQuery * select_query,
                           const NamesAndTypesList & source_columns, const NameSet & source_columns_set)
{
    if (ASTPtr array_join_expression_list = select_query->arrayJoinExpressionList())
    {
        ArrayJoinedColumnsVisitor::Data visitor_data{result.aliases,
                                                    result.array_join_name_to_alias,
                                                    result.array_join_alias_to_name,
                                                    result.array_join_result_to_source};
        ArrayJoinedColumnsVisitor(visitor_data).visit(query);

        /// If the result of ARRAY JOIN is not used, it is necessary to ARRAY-JOIN any column,
        /// to get the correct number of rows.
        if (result.array_join_result_to_source.empty())
        {
            if (select_query->arrayJoinExpressionList()->children.empty())
                throw DB::Exception("ARRAY JOIN requires an argument", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);

            ASTPtr expr = select_query->arrayJoinExpressionList()->children.at(0);
            String source_name = expr->getColumnName();
            String result_name = expr->getAliasOrColumnName();

            /// This is an array.
            if (!expr->as<ASTIdentifier>() || source_columns_set.count(source_name))
            {
                result.array_join_result_to_source[result_name] = source_name;
            }
            else /// This is a nested table.
            {
                bool found = false;
                for (const auto & column : source_columns)
                {
                    auto split = Nested::splitName(column.name);
                    if (split.first == source_name && !split.second.empty())
                    {
                        result.array_join_result_to_source[Nested::concatenateName(result_name, split.second)] = column.name;
                        found = true;
                        break;
                    }
                }
                if (!found)
                    throw Exception("No columns in nested table " + source_name, ErrorCodes::EMPTY_NESTED_TABLE);
            }
        }
    }
}

void setJoinStrictness(ASTSelectQuery & select_query, JoinStrictness join_default_strictness, bool old_any, ASTTableJoin & out_table_join)
{
    const ASTTablesInSelectQueryElement * node = select_query.join();
    if (!node)
        return;

    auto & table_join = const_cast<ASTTablesInSelectQueryElement *>(node)->table_join->as<ASTTableJoin &>();

    if (table_join.strictness == ASTTableJoin::Strictness::Unspecified &&
        table_join.kind != ASTTableJoin::Kind::Cross)
    {
        if (join_default_strictness == JoinStrictness::ANY)
            table_join.strictness = ASTTableJoin::Strictness::Any;
        else if (join_default_strictness == JoinStrictness::ALL)
            table_join.strictness = ASTTableJoin::Strictness::All;
        else
            throw Exception("Expected ANY or ALL in JOIN section, because setting (join_default_strictness) is empty",
                            DB::ErrorCodes::EXPECTED_ALL_OR_ANY);
    }

    if (old_any)
    {
        if (table_join.strictness == ASTTableJoin::Strictness::Any &&
            table_join.kind == ASTTableJoin::Kind::Inner)
        {
            table_join.strictness = ASTTableJoin::Strictness::Semi;
            table_join.kind = ASTTableJoin::Kind::Left;
        }

        if (table_join.strictness == ASTTableJoin::Strictness::Any)
            table_join.strictness = ASTTableJoin::Strictness::RightAny;
    }
    else
    {
        if (table_join.strictness == ASTTableJoin::Strictness::Any)
            if (table_join.kind == ASTTableJoin::Kind::Full)
                throw Exception("ANY FULL JOINs are not implemented.", ErrorCodes::NOT_IMPLEMENTED);
    }

    out_table_join = table_join;
}

/// Find the columns that are obtained by JOIN.
void collectJoinedColumns(TableJoin & analyzed_join, const ASTSelectQuery & select_query,
                          const TablesWithColumns & tables, const Aliases & aliases)
{
    const ASTTablesInSelectQueryElement * node = select_query.join();
    if (!node || tables.size() < 2)
        return;

    const auto & table_join = node->table_join->as<ASTTableJoin &>();

    if (table_join.using_expression_list)
    {
        const auto & keys = table_join.using_expression_list->as<ASTExpressionList &>();
        for (const auto & key : keys.children)
            analyzed_join.addUsingKey(key);

        /// `USING` semantic allows to have columns with changed types in result table.
        /// `JOIN ON` should preserve types from original table
        /// We can infer common type on syntax stage for `USING` because join is performed only by columns (not expressions)
        /// We need to know  changed types in result tables because some analysis (e.g. analyzeAggregation) performed before join
        /// For `JOIN ON expr1 == expr2` we will infer common type later in ExpressionAnalyzer, when types of expression will be known
        analyzed_join.inferJoinKeyCommonType(tables[0].columns, tables[1].columns);
    }
    else if (table_join.on_expression)
    {
        bool is_asof = (table_join.strictness == ASTTableJoin::Strictness::Asof);

        CollectJoinOnKeysVisitor::Data data{analyzed_join, tables[0], tables[1], aliases, is_asof};
        CollectJoinOnKeysVisitor(data).visit(table_join.on_expression);
        if (!data.has_some)
            throw Exception("Cannot get JOIN keys from JOIN ON section: " + queryToString(table_join.on_expression),
                            ErrorCodes::INVALID_JOIN_ON_EXPRESSION);
        if (is_asof)
            data.asofToJoinKeys();
    }
}


std::vector<const ASTFunction *> getAggregates(ASTPtr & query, const ASTSelectQuery & select_query)
{
    /// There can not be aggregate functions inside the WHERE and PREWHERE.
    if (select_query.where())
        assertNoAggregates(select_query.where(), "in WHERE");
    if (select_query.prewhere())
        assertNoAggregates(select_query.prewhere(), "in PREWHERE");

    GetAggregatesVisitor::Data data;
    GetAggregatesVisitor(data).visit(query);

    /// There can not be other aggregate functions within the aggregate functions.
    for (const ASTFunction * node : data.aggregates)
    {
        if (node->arguments)
        {
            for (auto & arg : node->arguments->children)
            {
                assertNoAggregates(arg, "inside another aggregate function");
                // We also can't have window functions inside aggregate functions,
                // because the window functions are calculated later.
                assertNoWindows(arg, "inside an aggregate function");
            }
        }
    }
    return data.aggregates;
}

std::vector<const ASTFunction *> getWindowFunctions(ASTPtr & query, const ASTSelectQuery & select_query)
{
    /// There can not be window functions inside the WHERE, PREWHERE and HAVING
    if (select_query.having())
        assertNoWindows(select_query.having(), "in HAVING");
    if (select_query.where())
        assertNoWindows(select_query.where(), "in WHERE");
    if (select_query.prewhere())
        assertNoWindows(select_query.prewhere(), "in PREWHERE");
    if (select_query.window())
        assertNoWindows(select_query.window(), "in WINDOW");

    GetAggregatesVisitor::Data data;
    GetAggregatesVisitor(data).visit(query);

    /// Window functions cannot be inside aggregates or other window functions.
    /// Aggregate functions can be inside window functions because they are
    /// calculated earlier.
    for (const ASTFunction * node : data.window_functions)
    {
        if (node->arguments)
        {
            for (auto & arg : node->arguments->children)
            {
                assertNoWindows(arg, "inside another window function");
            }
        }

        if (node->window_definition)
        {
            assertNoWindows(node->window_definition, "inside window definition");
        }
    }

    return data.window_functions;
}

}

TreeRewriterResult::TreeRewriterResult(
    const NamesAndTypesList & source_columns_,
    ConstStoragePtr storage_,
    const StorageMetadataPtr & metadata_snapshot_,
    bool add_special)
    : storage(storage_)
    , metadata_snapshot(metadata_snapshot_)
    , source_columns(source_columns_)
{
    collectSourceColumns(add_special);
    is_remote_storage = storage && storage->isRemote();
}

/// Add columns from storage to source_columns list. Deduplicate resulted list.
/// Special columns are non physical columns, for example ALIAS
void TreeRewriterResult::collectSourceColumns(bool add_special)
{
    if (storage)
    {
        const ColumnsDescription & columns = metadata_snapshot->getColumns();

        NamesAndTypesList columns_from_storage;
        if (storage->supportsSubcolumns())
            columns_from_storage = add_special ? columns.getAllWithSubcolumns() : columns.getAllPhysicalWithSubcolumns();
        else
            columns_from_storage = add_special ? columns.getAll() : columns.getAllPhysical();

        if (source_columns.empty())
            source_columns.swap(columns_from_storage);
        else
            source_columns.insert(source_columns.end(), columns_from_storage.begin(), columns_from_storage.end());
    }

    source_columns_set = removeDuplicateColumns(source_columns);
}


/// Calculate which columns are required to execute the expression.
/// Then, delete all other columns from the list of available columns.
/// After execution, columns will only contain the list of columns needed to read from the table.
void TreeRewriterResult::collectUsedColumns(const ASTPtr & query, bool is_select)
{
    /// We calculate required_source_columns with source_columns modifications and swap them on exit
    required_source_columns = source_columns;

    RequiredSourceColumnsVisitor::Data columns_context;
    RequiredSourceColumnsVisitor(columns_context).visit(query);

    NameSet source_column_names;
    for (const auto & column : source_columns)
        source_column_names.insert(column.name);

    NameSet required = columns_context.requiredColumns();
    if (columns_context.has_table_join)
    {
        NameSet available_columns;
        for (const auto & name : source_columns)
            available_columns.insert(name.name);

        /// Add columns obtained by JOIN (if needed).
        for (const auto & joined_column : analyzed_join->columnsFromJoinedTable())
        {
            const auto & name = joined_column.name;
            if (available_columns.count(name))
                continue;

            if (required.count(name))
            {
                /// Optimisation: do not add columns needed only in JOIN ON section.
                if (columns_context.nameInclusion(name) > analyzed_join->rightKeyInclusion(name))
                    analyzed_join->addJoinedColumn(joined_column);

                required.erase(name);
            }
        }
    }

    NameSet array_join_sources;
    if (columns_context.has_array_join)
    {
        /// Insert the columns required for the ARRAY JOIN calculation into the required columns list.
        for (const auto & result_source : array_join_result_to_source)
            array_join_sources.insert(result_source.second);

        for (const auto & column_name_type : source_columns)
            if (array_join_sources.count(column_name_type.name))
                required.insert(column_name_type.name);
    }

    /// Figure out if we're able to use the trivial count optimization.
    has_explicit_columns = !required.empty();
    if (is_select && !has_explicit_columns)
    {
        optimize_trivial_count = true;

        /// You need to read at least one column to find the number of rows.
        /// We will find a column with minimum <compressed_size, type_size, uncompressed_size>.
        /// Because it is the column that is cheapest to read.
        struct ColumnSizeTuple
        {
            size_t compressed_size;
            size_t type_size;
            size_t uncompressed_size;
            String name;

            bool operator<(const ColumnSizeTuple & that) const
            {
                return std::tie(compressed_size, type_size, uncompressed_size)
                    < std::tie(that.compressed_size, that.type_size, that.uncompressed_size);
            }
        };

        std::vector<ColumnSizeTuple> columns;
        if (storage)
        {
            auto column_sizes = storage->getColumnSizes();
            for (auto & source_column : source_columns)
            {
                auto c = column_sizes.find(source_column.name);
                if (c == column_sizes.end())
                    continue;
                size_t type_size = source_column.type->haveMaximumSizeOfValue() ? source_column.type->getMaximumSizeOfValueInMemory() : 100;
                columns.emplace_back(ColumnSizeTuple{c->second.data_compressed, type_size, c->second.data_uncompressed, source_column.name});
            }
        }

        if (!columns.empty())
            required.insert(std::min_element(columns.begin(), columns.end())->name);
        else if (!source_columns.empty())
            /// If we have no information about columns sizes, choose a column of minimum size of its data type.
            required.insert(ExpressionActions::getSmallestColumn(source_columns));
    }
    else if (is_select && metadata_snapshot && !columns_context.has_array_join)
    {
        const auto & partition_desc = metadata_snapshot->getPartitionKey();
        if (partition_desc.expression)
        {
            auto partition_source_columns = partition_desc.expression->getRequiredColumns();
            partition_source_columns.push_back("_part");
            partition_source_columns.push_back("_partition_id");
            partition_source_columns.push_back("_part_uuid");
            partition_source_columns.push_back("_partition_value");
            optimize_trivial_count = true;
            for (const auto & required_column : required)
            {
                if (std::find(partition_source_columns.begin(), partition_source_columns.end(), required_column)
                    == partition_source_columns.end())
                {
                    optimize_trivial_count = false;
                    break;
                }
            }
        }
    }

    NameSet unknown_required_source_columns = required;

    for (NamesAndTypesList::iterator it = source_columns.begin(); it != source_columns.end();)
    {
        const String & column_name = it->name;
        unknown_required_source_columns.erase(column_name);

        if (!required.count(column_name))
            source_columns.erase(it++);
        else
            ++it;
    }

    /// If there are virtual columns among the unknown columns. Remove them from the list of unknown and add
    /// in columns list, so that when further processing they are also considered.
    if (storage)
    {
        const auto storage_virtuals = storage->getVirtuals();
        for (auto it = unknown_required_source_columns.begin(); it != unknown_required_source_columns.end();)
        {
            auto column = storage_virtuals.tryGetByName(*it);
            if (column)
            {
                source_columns.push_back(*column);
                unknown_required_source_columns.erase(it++);
            }
            else
                ++it;
        }
    }

    if (!unknown_required_source_columns.empty())
    {
        WriteBufferFromOwnString ss;
        ss << "Missing columns:";
        for (const auto & name : unknown_required_source_columns)
            ss << " '" << name << "'";
        ss << " while processing query: '" << queryToString(query) << "'";

        ss << ", required columns:";
        for (const auto & name : columns_context.requiredColumns())
            ss << " '" << name << "'";

        if (storage)
        {
            std::vector<String> hint_name{};
            for (const auto & name : columns_context.requiredColumns())
            {
                auto hints = storage->getHints(name);
                hint_name.insert(hint_name.end(), hints.begin(), hints.end());
            }

            if (!hint_name.empty())
            {
                ss << ", maybe you meant: ";
                ss << toString(hint_name);
            }
        }
        else
        {
            if (!source_column_names.empty())
                for (const auto & name : columns_context.requiredColumns())
                    ss << " '" << name << "'";
            else
                ss << ", no source columns";
        }

        if (columns_context.has_table_join)
        {
            ss << ", joined columns:";
            for (const auto & column : analyzed_join->columnsFromJoinedTable())
                ss << " '" << column.name << "'";
        }

        if (!array_join_sources.empty())
        {
            ss << ", arrayJoin columns:";
            for (const auto & name : array_join_sources)
                ss << " '" << name << "'";
        }

        throw Exception(ss.str(), ErrorCodes::UNKNOWN_IDENTIFIER);
    }

    required_source_columns.swap(source_columns);
}

NameSet TreeRewriterResult::getArrayJoinSourceNameSet() const
{
    NameSet forbidden_columns;
    for (const auto & elem : array_join_result_to_source)
        forbidden_columns.insert(elem.first);
    return forbidden_columns;
}

TreeRewriterResultPtr TreeRewriter::analyzeSelect(
    ASTPtr & query,
    TreeRewriterResult && result,
    const SelectQueryOptions & select_options,
    const std::vector<TableWithColumnNamesAndTypes> & tables_with_columns,
    const Names & required_result_columns,
    std::shared_ptr<TableJoin> table_join) const
{
    auto * select_query = query->as<ASTSelectQuery>();
    if (!select_query)
        throw Exception("Select analyze for not select asts.", ErrorCodes::LOGICAL_ERROR);

    size_t subquery_depth = select_options.subquery_depth;
    bool remove_duplicates = select_options.remove_duplicates;

    const auto & settings = getContext()->getSettingsRef();

    const NameSet & source_columns_set = result.source_columns_set;

    if (table_join)
    {
        result.analyzed_join = table_join;
        result.analyzed_join->resetCollected();
    }
    else /// TODO: remove. For now ExpressionAnalyzer expects some not empty object here
        result.analyzed_join = std::make_shared<TableJoin>();

    if (remove_duplicates)
        renameDuplicatedColumns(select_query);

    if (tables_with_columns.size() > 1)
    {
        result.analyzed_join->columns_from_joined_table = tables_with_columns[1].columns;
        result.analyzed_join->deduplicateAndQualifyColumnNames(
            source_columns_set, tables_with_columns[1].table.getQualifiedNamePrefix());
    }

    translateQualifiedNames(query, *select_query, source_columns_set, tables_with_columns);

    /// Optimizes logical expressions.
    LogicalExpressionsOptimizer(select_query, settings.optimize_min_equality_disjunction_chain_length.value).perform();

    NameSet all_source_columns_set = source_columns_set;
    if (table_join)
    {
        for (const auto & [name, _] : table_join->columns_from_joined_table)
            all_source_columns_set.insert(name);
    }

    normalize(query, result.aliases, all_source_columns_set, select_options.ignore_alias, settings, /* allow_self_aliases = */ true);

    /// Remove unneeded columns according to 'required_result_columns'.
    /// Leave all selected columns in case of DISTINCT; columns that contain arrayJoin function inside.
    /// Must be after 'normalizeTree' (after expanding aliases, for aliases not get lost)
    ///  and before 'executeScalarSubqueries', 'analyzeAggregation', etc. to avoid excessive calculations.
    removeUnneededColumnsFromSelectClause(select_query, required_result_columns, remove_duplicates);

    /// Executing scalar subqueries - replacing them with constant values.
    executeScalarSubqueries(query, getContext(), subquery_depth, result.scalars, select_options.only_analyze);

    TreeOptimizer::apply(query, result, tables_with_columns, getContext());

    /// array_join_alias_to_name, array_join_result_to_source.
    getArrayJoinedColumns(query, result, select_query, result.source_columns, source_columns_set);

    setJoinStrictness(
        *select_query, settings.join_default_strictness, settings.any_join_distinct_right_table_keys, result.analyzed_join->table_join);

    collectJoinedColumns(*result.analyzed_join, *select_query, tables_with_columns, result.aliases);

    result.aggregates = getAggregates(query, *select_query);
    result.window_function_asts = getWindowFunctions(query, *select_query);
    result.collectUsedColumns(query, true);
    result.required_source_columns_before_expanding_alias_columns = result.required_source_columns.getNames();

    /// rewrite filters for select query, must go after getArrayJoinedColumns
    if (settings.optimize_respect_aliases && result.metadata_snapshot)
    {
        replaceAliasColumnsInQuery(query, result.metadata_snapshot->getColumns(), result.array_join_result_to_source, getContext());
        result.collectUsedColumns(query, true);
    }

    result.ast_join = select_query->join();

    if (result.optimize_trivial_count)
        result.optimize_trivial_count = settings.optimize_trivial_count_query &&
            !select_query->groupBy() && !select_query->having() &&
            !select_query->sampleSize() && !select_query->sampleOffset() && !select_query->final() &&
            (tables_with_columns.size() < 2 || isLeft(result.analyzed_join->kind()));

    return std::make_shared<const TreeRewriterResult>(result);
}

TreeRewriterResultPtr TreeRewriter::analyze(
    ASTPtr & query,
    const NamesAndTypesList & source_columns,
    ConstStoragePtr storage,
    const StorageMetadataPtr & metadata_snapshot,
    bool allow_aggregations,
    bool allow_self_aliases) const
{
    if (query->as<ASTSelectQuery>())
        throw Exception("Not select analyze for select asts.", ErrorCodes::LOGICAL_ERROR);

    const auto & settings = getContext()->getSettingsRef();

    TreeRewriterResult result(source_columns, storage, metadata_snapshot, false);

    normalize(query, result.aliases, result.source_columns_set, false, settings, allow_self_aliases);

    /// Executing scalar subqueries. Column defaults could be a scalar subquery.
    executeScalarSubqueries(query, getContext(), 0, result.scalars, false);

    TreeOptimizer::optimizeIf(query, result.aliases, settings.optimize_if_chain_to_multiif);

    if (allow_aggregations)
    {
        GetAggregatesVisitor::Data data;
        GetAggregatesVisitor(data).visit(query);

        /// There can not be other aggregate functions within the aggregate functions.
        for (const ASTFunction * node : data.aggregates)
            for (auto & arg : node->arguments->children)
                assertNoAggregates(arg, "inside another aggregate function");
        result.aggregates = data.aggregates;
    }
    else
        assertNoAggregates(query, "in wrong place");

    result.collectUsedColumns(query, false);
    return std::make_shared<const TreeRewriterResult>(result);
}

void TreeRewriter::normalize(
    ASTPtr & query, Aliases & aliases, const NameSet & source_columns_set, bool ignore_alias, const Settings & settings, bool allow_self_aliases)
{
    CustomizeCountDistinctVisitor::Data data_count_distinct{settings.count_distinct_implementation};
    CustomizeCountDistinctVisitor(data_count_distinct).visit(query);

    CustomizeCountIfDistinctVisitor::Data data_count_if_distinct{settings.count_distinct_implementation.toString() + "If"};
    CustomizeCountIfDistinctVisitor(data_count_if_distinct).visit(query);

    CustomizeIfDistinctVisitor::Data data_distinct_if{"DistinctIf"};
    CustomizeIfDistinctVisitor(data_distinct_if).visit(query);

    if (settings.transform_null_in)
    {
        CustomizeInVisitor::Data data_null_in{"nullIn"};
        CustomizeInVisitor(data_null_in).visit(query);

        CustomizeNotInVisitor::Data data_not_null_in{"notNullIn"};
        CustomizeNotInVisitor(data_not_null_in).visit(query);

        CustomizeGlobalInVisitor::Data data_global_null_in{"globalNullIn"};
        CustomizeGlobalInVisitor(data_global_null_in).visit(query);

        CustomizeGlobalNotInVisitor::Data data_global_not_null_in{"globalNotNullIn"};
        CustomizeGlobalNotInVisitor(data_global_not_null_in).visit(query);
    }

    // Try to fuse sum/avg/count with identical arguments to one sumCount call,
    // if we have at least two different functions. E.g. we will replace sum(x)
    // and count(x) with sumCount(x).1 and sumCount(x).2, and sumCount() will
    // be calculated only once because of CSE.
    if (settings.optimize_fuse_sum_count_avg)
    {
        FuseSumCountAggregatesVisitor::Data data;
        FuseSumCountAggregatesVisitor(data).visit(query);
        fuseSumCountAggregates(data.fuse_map);
    }

    /// Rewrite all aggregate functions to add -OrNull suffix to them
    if (settings.aggregate_functions_null_for_empty)
    {
        CustomizeAggregateFunctionsOrNullVisitor::Data data_or_null{"OrNull"};
        CustomizeAggregateFunctionsOrNullVisitor(data_or_null).visit(query);
    }

    /// Move -OrNull suffix ahead, this should execute after add -OrNull suffix
    CustomizeAggregateFunctionsMoveOrNullVisitor::Data data_or_null{"OrNull"};
    CustomizeAggregateFunctionsMoveOrNullVisitor(data_or_null).visit(query);

    /// Creates a dictionary `aliases`: alias -> ASTPtr
    QueryAliasesVisitor(aliases).visit(query);

    /// Mark table ASTIdentifiers with not a column marker
    MarkTableIdentifiersVisitor::Data identifiers_data{aliases};
    MarkTableIdentifiersVisitor(identifiers_data).visit(query);

    /// Rewrite function names to their canonical ones.
    if (settings.normalize_function_names)
        FunctionNameNormalizer().visit(query.get());

    /// Common subexpression elimination. Rewrite rules.
    QueryNormalizer::Data normalizer_data(aliases, source_columns_set, ignore_alias, settings, allow_self_aliases);
    QueryNormalizer(normalizer_data).visit(query);
}

}