#include #include #include #include #include #include #include namespace DB { namespace ErrorCodes { extern const int AMBIGUOUS_COLUMN_NAME; } namespace { template std::optional tryChooseTable(const ASTIdentifier & identifier, const std::vector & tables, bool allow_ambiguous, bool column_match [[maybe_unused]] = false) { using ColumnMatch = IdentifierSemantic::ColumnMatch; size_t best_table_pos = 0; auto best_match = ColumnMatch::NoMatch; size_t same_match = 0; for (size_t i = 0; i < tables.size(); ++i) { auto match = IdentifierSemantic::canReferColumnToTable(identifier, tables[i]); if constexpr (std::is_same_v) { if (column_match && match == ColumnMatch::NoMatch && identifier.isShort() && tables[i].hasColumn(identifier.shortName())) match = ColumnMatch::ColumnName; } if (match != ColumnMatch::NoMatch) { if (match > best_match) { best_match = match; best_table_pos = i; same_match = 0; } else if (match == best_match) ++same_match; } } if ((best_match != ColumnMatch::NoMatch) && same_match) { if (!allow_ambiguous) throw Exception("Ambiguous column '" + identifier.name() + "'", ErrorCodes::AMBIGUOUS_COLUMN_NAME); best_match = ColumnMatch::Ambiguous; return {}; } if (best_match != ColumnMatch::NoMatch) return best_table_pos; return {}; } } std::optional IdentifierSemantic::getColumnName(const ASTIdentifier & node) { if (!node.semantic->special) return node.name(); return {}; } std::optional IdentifierSemantic::getColumnName(const ASTPtr & ast) { if (ast) if (const auto * id = ast->as()) if (!id->semantic->special) return id->name(); return {}; } std::optional IdentifierSemantic::uncover(const ASTIdentifier & identifier) { if (identifier.semantic->covered) { std::vector name_parts = identifier.name_parts; return ASTIdentifier(std::move(name_parts)); } return {}; } void IdentifierSemantic::coverName(ASTIdentifier & identifier, const String & alias) { identifier.setShortName(alias); identifier.semantic->covered = true; } bool IdentifierSemantic::canBeAlias(const ASTIdentifier & identifier) { return identifier.semantic->can_be_alias; } void IdentifierSemantic::setMembership(ASTIdentifier & identifier, size_t table_pos) { identifier.semantic->membership = table_pos; identifier.semantic->can_be_alias = false; } std::optional IdentifierSemantic::getMembership(const ASTIdentifier & identifier) { return identifier.semantic->membership; } std::optional IdentifierSemantic::chooseTable(const ASTIdentifier & identifier, const std::vector & tables, bool ambiguous) { return tryChooseTable(identifier, tables, ambiguous); } std::optional IdentifierSemantic::chooseTable(const ASTIdentifier & identifier, const TablesWithColumns & tables, bool ambiguous) { return tryChooseTable(identifier, tables, ambiguous); } std::optional IdentifierSemantic::chooseTableColumnMatch(const ASTIdentifier & identifier, const TablesWithColumns & tables, bool ambiguous) { return tryChooseTable(identifier, tables, ambiguous, true); } std::optional IdentifierSemantic::extractNestedName(const ASTIdentifier & identifier, const String & table_name) { if (identifier.name_parts.size() == 3 && table_name == identifier.name_parts[0]) return identifier.name_parts[1] + '.' + identifier.name_parts[2]; else if (identifier.name_parts.size() == 2) return identifier.name_parts[0] + '.' + identifier.name_parts[1]; return {}; } bool IdentifierSemantic::doesIdentifierBelongTo(const ASTIdentifier & identifier, const String & database, const String & table) { size_t num_components = identifier.name_parts.size(); if (num_components >= 3) return identifier.name_parts[0] == database && identifier.name_parts[1] == table; return false; } bool IdentifierSemantic::doesIdentifierBelongTo(const ASTIdentifier & identifier, const String & table) { size_t num_components = identifier.name_parts.size(); if (num_components >= 2) return identifier.name_parts[0] == table; return false; } IdentifierSemantic::ColumnMatch IdentifierSemantic::canReferColumnToTable(const ASTIdentifier & identifier, const DatabaseAndTableWithAlias & db_and_table) { /// database.table.column if (doesIdentifierBelongTo(identifier, db_and_table.database, db_and_table.table)) return ColumnMatch::DBAndTable; /// alias.column if (doesIdentifierBelongTo(identifier, db_and_table.alias)) return ColumnMatch::TableAlias; /// table.column if (doesIdentifierBelongTo(identifier, db_and_table.table)) { if (!db_and_table.alias.empty()) return ColumnMatch::AliasedTableName; else return ColumnMatch::TableName; } return ColumnMatch::NoMatch; } IdentifierSemantic::ColumnMatch IdentifierSemantic::canReferColumnToTable(const ASTIdentifier & identifier, const TableWithColumnNamesAndTypes & table_with_columns) { return canReferColumnToTable(identifier, table_with_columns.table); } /// Strip qualifications from left side of column name. /// Example: 'database.table.name' -> 'name'. void IdentifierSemantic::setColumnShortName(ASTIdentifier & identifier, const DatabaseAndTableWithAlias & db_and_table) { auto match = IdentifierSemantic::canReferColumnToTable(identifier, db_and_table); size_t to_strip = 0; switch (match) { case ColumnMatch::TableName: case ColumnMatch::AliasedTableName: case ColumnMatch::TableAlias: to_strip = 1; break; case ColumnMatch::DBAndTable: to_strip = 2; break; default: break; } if (!to_strip) return; identifier.name_parts = std::vector(identifier.name_parts.begin() + to_strip, identifier.name_parts.end()); identifier.resetFullName(); } void IdentifierSemantic::setColumnLongName(ASTIdentifier & identifier, const DatabaseAndTableWithAlias & db_and_table) { String prefix = db_and_table.getQualifiedNamePrefix(); if (!prefix.empty()) { prefix.resize(prefix.size() - 1); /// crop dot identifier.name_parts = {prefix, identifier.shortName()}; identifier.resetFullName(); identifier.semantic->table = prefix; identifier.semantic->legacy_compound = true; } } std::optional IdentifierSemantic::getIdentMembership(const ASTIdentifier & ident, const std::vector & tables) { std::optional table_pos = IdentifierSemantic::getMembership(ident); if (table_pos) return table_pos; return IdentifierSemantic::chooseTableColumnMatch(ident, tables, true); } std::optional IdentifierSemantic::getIdentsMembership(ASTPtr ast, const std::vector & tables, const Aliases & aliases) { auto idents = IdentifiersCollector::collect(ast); std::optional result; for (const auto * ident : idents) { /// short name clashes with alias, ambiguous if (ident->isShort() && aliases.count(ident->shortName())) return {}; const auto pos = getIdentMembership(*ident, tables); if (!pos) return {}; /// identifiers from different tables if (result && *pos != *result) return {}; result = pos; } return result; } IdentifiersCollector::ASTIdentifiers IdentifiersCollector::collect(const ASTPtr & node) { IdentifiersCollector::Data ident_data; ConstInDepthNodeVisitor ident_visitor(ident_data); ident_visitor.visit(node); return ident_data.idents; } bool IdentifiersCollector::needChildVisit(const ASTPtr &, const ASTPtr &) { return true; } void IdentifiersCollector::visit(const ASTPtr & node, IdentifiersCollector::Data & data) { if (const auto * ident = node->as()) data.idents.push_back(ident); } IdentifierMembershipCollector::IdentifierMembershipCollector(const ASTSelectQuery & select, ContextPtr context) { if (ASTPtr with = select.with()) QueryAliasesNoSubqueriesVisitor(aliases).visit(with); QueryAliasesNoSubqueriesVisitor(aliases).visit(select.select()); const auto & settings = context->getSettingsRef(); tables = getDatabaseAndTablesWithColumns(getTableExpressions(select), context, settings.asterisk_include_alias_columns, settings.asterisk_include_materialized_columns); } std::optional IdentifierMembershipCollector::getIdentsMembership(ASTPtr ast) const { return IdentifierSemantic::getIdentsMembership(ast, tables, aliases); } static void collectConjunctions(const ASTPtr & node, std::vector & members) { if (const auto * func = node->as(); func && func->name == "and") { for (const auto & child : func->arguments->children) collectConjunctions(child, members); return; } members.push_back(node); } std::vector collectConjunctions(const ASTPtr & node) { std::vector members; collectConjunctions(node, members); return members; } }