Branch data     Line data    Source code

       1                 :            : #pragma once
       2                 :            : 
       3                 :            : #include <mutable/catalog/Catalog.hpp>
       4                 :            : #include <mutable/catalog/DatabaseCommand.hpp>
       5                 :            : #include <mutable/parse/AST.hpp>
       6                 :            : #include <mutable/util/Diagnostic.hpp>
       7                 :            : #include <sstream>
       8                 :            : #include <unordered_map>
       9                 :            : #include <vector>
      10                 :            : 
      11                 :            : 
      12                 :            : namespace m {
      13                 :            : 
      14                 :            : namespace ast {
      15                 :            : 
      16                 :            : struct M_EXPORT Sema : ASTVisitor
      17                 :            : {
      18                 :            :     /** Holds context information used by semantic analysis of a single statement. */
      19   [ +  -  -  + ]:       1401 :     struct SemaContext
      20                 :            :     {
      21                 :            :         ///> if the statement that is being analyzed is a nested query, this is its alias in the outer statement
      22                 :            :         ThreadSafePooledOptionalString alias;
      23                 :            : 
      24                 :            :         ///> the statement that is currently being analyzed and for which this `SemaContext` is used
      25                 :            :         Stmt &stmt;
      26                 :            :         enum stage_t {
      27                 :            :             S_From,
      28                 :            :             S_Where,
      29                 :            :             S_GroupBy,
      30                 :            :             S_Having,
      31                 :            :             S_Select,
      32                 :            :             S_OrderBy,
      33                 :            :             S_Limit,
      34                 :       1401 :         } stage = S_From; ///< current stage
      35                 :            : 
      36                 :       1401 :         bool needs_grouping = false;
      37                 :            : 
      38                 :            :         struct result_t
      39                 :            :         {
      40                 :            :             std::reference_wrapper<Expr> expr_;
      41                 :            :             ///> the order of this result column in the result set
      42                 :            :             unsigned order;
      43                 :            :             ///> alias of the expression; may not have a value
      44                 :            :             ThreadSafePooledOptionalString alias;
      45                 :            : 
      46                 :       1461 :             result_t(Expr &expr, unsigned order) : expr_(expr), order(order) { }
      47                 :         81 :             result_t(Expr &expr, unsigned order, ThreadSafePooledOptionalString alias)
      48                 :         81 :                 : expr_(expr), order(order), alias(std::move(alias))
      49                 :         81 :             { }
      50                 :            : 
      51                 :        194 :             Expr & expr() { return expr_.get(); }
      52                 :            :             const Expr & expr() const { return expr_.get(); }
      53                 :            :         };
      54                 :            : 
      55                 :            :         ///> list of all computed expressions along with their order
      56                 :            :         using named_expr_table = std::unordered_multimap<ThreadSafePooledString,
      57                 :            :                                                          std::pair<std::reference_wrapper<Expr>, unsigned>>;
      58                 :            :         ///> the type of a source of data: either a database table or a nested query with named results
      59                 :            :         using source_type = std::variant<std::monostate, std::reference_wrapper<const Table>, named_expr_table>;
      60                 :            :         ///> associative container mapping source name to data source and its order
      61                 :            :         using source_table = std::unordered_map<ThreadSafePooledString, std::pair<source_type, unsigned>>;
      62                 :            :         ///> list of all sources along with their order
      63                 :            :         source_table sources;
      64                 :            :         ///> list of all results computed by this statement along with their order
      65                 :            :         std::unordered_multimap<ThreadSafePooledString, result_t> results;
      66                 :            :         ///> list of grouping keys
      67                 :            :         std::unordered_multimap<ThreadSafePooledString, std::reference_wrapper<Expr>> grouping_keys;
      68                 :            : 
      69                 :       2802 :         SemaContext(Stmt &stmt) : stmt(stmt) { }
      70                 :            :     };
      71                 :            : 
      72                 :            :     private:
      73                 :            :     /** Helper class to create a context when one is required but does not yet exist.  Automatically disposes of the
      74                 :            :      * context when the instance goes out of scope. */
      75                 :            :     struct RequireContext
      76                 :            :     {
      77                 :            :         private:
      78                 :            :         Sema &sema_;
      79                 :            :         bool needs_context_ = false;
      80                 :            : 
      81                 :            :         public:
      82                 :       1401 :         RequireContext(Sema *sema, Stmt &stmt)
      83                 :       1401 :             : sema_(*M_notnull(sema))
      84                 :       1401 :             , needs_context_(sema_.contexts_.empty())
      85                 :            :         {
      86         [ +  + ]:       1401 :             if (needs_context_)
      87         [ +  - ]:       1334 :                 sema_.push_context(stmt);
      88                 :       1401 :         }
      89                 :            : 
      90                 :       1401 :         ~RequireContext() {
      91         [ +  + ]:       1401 :             if (needs_context_)
      92         [ +  - ]:       1334 :                 sema_.pop_context();
      93                 :       1401 :         }
      94                 :            :     };
      95                 :            : 
      96                 :            :     public:
      97                 :            :     Diagnostic &diag;
      98                 :            :     private:
      99                 :            :     ///> a stack of sema contexts; one per statement; grows by nesting statements
     100                 :            :     using context_stack_t = std::vector<SemaContext*>;
     101                 :            :     context_stack_t contexts_;
     102                 :            :     ///> used to create textual representation of complex AST objects, e.g. expressions
     103                 :            :     std::ostringstream oss;
     104                 :            :     ///> the command to execute when semantic analysis completes without errors
     105                 :            :     std::unique_ptr<DatabaseCommand> command_;
     106                 :            : 
     107                 :            :     public:
     108         [ +  - ]:       1501 :     Sema(Diagnostic &diag) : diag(diag) { }
     109                 :            : 
     110                 :            :     /** Perform semantic analysis of an `ast::Command`. Returns an `m::DatabaseCommand` to execute when no semantic
     111                 :            :      * errors occurred, `nullptr` otherwise. */
     112                 :            :     std::unique_ptr<DatabaseCommand> analyze(std::unique_ptr<ast::Command> ast);
     113                 :            : 
     114                 :            :     using ASTExprVisitor::operator();
     115                 :            :     using ASTClauseVisitor::operator();
     116                 :            :     using ASTCommandVisitor::operator();
     117                 :            : #define DECLARE(CLASS) void operator()(CLASS&) override;
     118                 :            :     M_AST_EXPR_LIST(DECLARE)
     119                 :            :     M_AST_CLAUSE_LIST(DECLARE)
     120                 :            :     M_AST_COMMAND_LIST(DECLARE)
     121                 :            : #undef DECLARE
     122                 :            : 
     123                 :            :     private:
     124                 :       1401 :     SemaContext & push_context(Stmt &stmt, ThreadSafePooledOptionalString alias = {}) {
     125         [ +  - ]:       1401 :         auto &ref = contexts_.emplace_back(new SemaContext(stmt));
     126         [ -  + ]:       1401 :         ref->alias = std::move(alias);
     127                 :       1401 :         return *ref;
     128                 :          0 :     }
     129                 :       1401 :     SemaContext pop_context() {
     130                 :       1401 :         auto ctx = *contexts_.back();
     131         [ -  + ]:       1401 :         delete contexts_.back();
     132                 :       1401 :         contexts_.pop_back();
     133                 :       1401 :         return ctx;
     134         [ +  - ]:       1401 :     }
     135                 :       4046 :     SemaContext & get_context() {
     136                 :       4046 :         M_insist(not contexts_.empty());
     137                 :       4046 :         return *contexts_.back();
     138                 :            :     }
     139                 :            :     const SemaContext & get_context() const {
     140                 :            :         M_insist(not contexts_.empty());
     141                 :            :         return *contexts_.back();
     142                 :            :     }
     143                 :            : 
     144                 :            :     /** Returns true iff the current statement, that is being analyzed, is a nested statement. */
     145                 :            :     bool is_nested() const;
     146                 :            : 
     147                 :            : 
     148                 :            :     /*------------------------------------------------------------------------------------------------------------------
     149                 :            :      * Sema Designator Helpers
     150                 :            :      *----------------------------------------------------------------------------------------------------------------*/
     151                 :            : 
     152                 :            :     /** Creates a fresh `Designator` with the given \p name at location \p tok and with target \p target. */
     153                 :            :     std::unique_ptr<Designator> create_designator(ThreadSafePooledString name, Token tok, const Expr &target);
     154                 :            : 
     155                 :            :     /** Creates a fresh `Designator` with the same syntactical representation as \p name (except the table name if
     156                 :            :      * \p drop_table_name) and with target \p target. */
     157                 :            :     std::unique_ptr<Designator> create_designator(const Expr &name, const Expr &target, bool drop_table_name = false);
     158                 :            : 
     159                 :            :     /** Creates an entirely new `Designator`.  This method is used to introduce artificial `Designator`s to *expand* an
     160                 :            :      * anti-projection (`SELECT` with asterisk `*`). */
     161                 :       1172 :     std::unique_ptr<Designator> create_designator(Position pos, ThreadSafePooledString table_name,
     162                 :            :                                                   ThreadSafePooledString attr_name,
     163                 :            :                                                   typename Designator::target_type target, const Type *type)
     164                 :            :     {
     165                 :       1172 :         auto &C = Catalog::Get();
     166         [ +  - ]:       1172 :         Token dot(pos, C.pool("."), TK_DOT);
     167   [ +  -  +  - ]:       1172 :         Token table(pos, std::move(table_name), TK_IDENTIFIER);
     168   [ +  -  +  - ]:       1172 :         Token attr(pos, std::move(attr_name), TK_IDENTIFIER);
     169         [ +  - ]:       1172 :         auto d = std::make_unique<Designator>(std::move(dot), std::move(table), std::move(attr));
     170                 :       1172 :         d->type_ = type;
     171                 :       1172 :         d->target_ = target;
     172                 :       1172 :         return d;
     173         [ +  - ]:       1172 :     }
     174                 :            : 
     175                 :            :     /** Replaces \p to_replace by a fresh `Designator`, that has the same syntactical representation as \p to_replace
     176                 :            :      * and targets \p target. */
     177                 :            :     void replace_by_fresh_designator_to(std::unique_ptr<Expr> &to_replace, const Expr &target);
     178                 :            : 
     179                 :            : 
     180                 :            :     /*------------------------------------------------------------------------------------------------------------------
     181                 :            :      * Other Sema Helpers
     182                 :            :      *----------------------------------------------------------------------------------------------------------------*/
     183                 :            : 
     184                 :            :     /** Computes whether the bound parts of \p expr are composable of elements in \p components. */
     185                 :            :     bool is_composable_of(const ast::Expr &expr, const std::vector<std::reference_wrapper<ast::Expr>> components);
     186                 :            : 
     187                 :            :     /** Recursively analyzes the `ast::Expr` referenced by \p ptr and replaces subexpressions that can be composed of
     188                 :            :      * the elements in \p components. */
     189                 :            :     void compose_of(std::unique_ptr<ast::Expr> &ptr, const std::vector<std::reference_wrapper<ast::Expr>> components);
     190                 :            : 
     191                 :            :     /** Creates a unique ID from a sequence of `SemaContext`s by concatenating their aliases. */
     192                 :            :     ThreadSafePooledOptionalString make_unique_id_from_binding_path(context_stack_t::reverse_iterator current_ctx,
     193                 :            :                                                                     context_stack_t::reverse_iterator binding_ctx);
     194                 :            : };
     195                 :            : 
     196                 :            : }
     197                 :            : 
     198                 :            : }