source: src/Parser/ExpressionNode.cc @ ba54f7d

//
// Cforall Version 1.0.0 Copyright (C) 2015 University of Waterloo
//
// The contents of this file are covered under the licence agreement in the
// file "LICENCE" distributed with Cforall.
//
// ExpressionNode.cc --
//
// Author           : Peter A. Buhr
// Created On       : Sat May 16 13:17:07 2015
// Last Modified By : Peter A. Buhr
// Last Modified On : Wed Sep 13 14:54:19 2017
// Update Count     : 683
//

#include <cassert>                 // for assert
#include <stdio.h>                 // for sscanf, size_t
#include <climits>                 // for LLONG_MAX, LONG_MAX, INT_MAX, UINT...
#include <list>                    // for list
#include <sstream>                 // for basic_istream::operator>>, basic_i...
#include <string>                  // for string, operator+, operator==

#include "Common/SemanticError.h"  // for SemanticError
#include "Common/utility.h"        // for maybeMoveBuild, maybeBuild, CodeLo...
#include "ParseNode.h"             // for ExpressionNode, maybeMoveBuildType
#include "SynTree/Constant.h"      // for Constant
#include "SynTree/Declaration.h"   // for EnumDecl, StructDecl, UnionDecl
#include "SynTree/Expression.h"    // for Expression, ConstantExpr, NameExpr
#include "SynTree/Statement.h"     // for CompoundStmt, Statement
#include "SynTree/Type.h"          // for BasicType, Type, Type::Qualifiers
#include "parserutility.h"         // for notZeroExpr

class Initializer;

using namespace std;

//##############################################################################

// Difficult to separate extra parts of constants during lexing because actions are not allow in the middle of patterns:
//
//              prefix action constant action suffix
//
// Alternatively, breaking a pattern using BEGIN does not work if the following pattern can be empty:
//
//              constant BEGIN CONT ...
//              <CONT>(...)? BEGIN 0 ... // possible empty suffix
//
// because the CONT rule is NOT triggered if the pattern is empty. Hence, constants are reparsed here to determine their
// type.

extern const Type::Qualifiers noQualifiers;                             // no qualifiers on constants

static inline bool checkH( char c ) { return c == 'h' || c == 'H'; }
static inline bool checkL( char c ) { return c == 'l' || c == 'L'; }
static inline bool checkZ( char c ) { return c == 'z' || c == 'Z'; }
static inline bool checkU( char c ) { return c == 'u' || c == 'U'; }
static inline bool checkF( char c ) { return c == 'f' || c == 'F'; }
static inline bool checkD( char c ) { return c == 'd' || c == 'D'; }
static inline bool checkI( char c ) { return c == 'i' || c == 'I'; }
static inline bool checkX( char c ) { return c == 'x' || c == 'X'; }

static void sepNumeric( string & str, string & units ) {
        string::size_type posn = str.find_first_of( "`" );
        if ( posn != string::npos ) {
                units = "?" + str.substr( posn );                               // extract units
                str.erase( posn );                                                              // remove units
        } // if
} // sepNumeric

Expression * build_constantInteger( string & str ) {
        static const BasicType::Kind kind[2][5] = {
                // short (h) must be before char (hh)
                { BasicType::ShortSignedInt, BasicType::SignedChar, BasicType::SignedInt, BasicType::LongSignedInt, BasicType::LongLongSignedInt },
                { BasicType::ShortUnsignedInt, BasicType::UnsignedChar, BasicType::UnsignedInt, BasicType::LongUnsignedInt, BasicType::LongLongUnsignedInt },
        };

        string units;                                                                           // units
        sepNumeric( str, units );                                                       // separate constant from units

        bool dec = true, Unsigned = false;                                      // decimal, unsigned constant
        int size;                                                                                       // 0 => short, 1 => char, 2 => int, 3 => long int, 4 => long long int, 5 => size_t
        unsigned long long int v;                                                       // converted integral value
        size_t last = str.length() - 1;                                         // last character of constant
        Expression * ret;

        // special constants
        if ( str == "0" ) {
                ret = new ConstantExpr( Constant( (Type *)new ZeroType( noQualifiers ), str, (unsigned long long int)0 ) );
                goto CLEANUP;
        } // if
        if ( str == "1" ) {
                ret = new ConstantExpr( Constant( (Type *)new OneType( noQualifiers ), str, (unsigned long long int)1 ) );
                goto CLEANUP;
        } // if

        if ( str[0] == '0' ) {                                                          // octal/hex constant ?
                dec = false;
                if ( last != 0 && checkX( str[1] ) ) {                  // hex constant ?
                        sscanf( (char *)str.c_str(), "%llx", &v );
                        //printf( "%llx %llu\n", v, v );
                } else {                                                                                // octal constant
                        sscanf( (char *)str.c_str(), "%llo", &v );
                        //printf( "%llo %llu\n", v, v );
                } // if
        } else {                                                                                        // decimal constant ?
                sscanf( (char *)str.c_str(), "%llu", &v );
                //printf( "%llu %llu\n", v, v );
        } // if

        if ( v <= INT_MAX ) {                                                           // signed int
                size = 2;
        } else if ( v <= UINT_MAX && ! dec ) {                          // unsigned int
                size = 2;
                Unsigned = true;                                                                // unsigned
        } else if ( v <= LONG_MAX ) {                                           // signed long int
                size = 3;
        } else if ( v <= ULONG_MAX && ( ! dec || LONG_MAX == LLONG_MAX ) ) { // signed long int
                size = 3;
                Unsigned = true;                                                                // unsigned long int
        } else if ( v <= LLONG_MAX ) {                                          // signed long long int
                size = 4;
        } else {                                                                                        // unsigned long long int
                size = 4;
                Unsigned = true;                                                                // unsigned long long int
        } // if

        // At least one digit in integer constant, so safe to backup while looking for suffix.

        if ( checkU( str[last] ) ) {                                            // suffix 'u' ?
                Unsigned = true;
                if ( checkL( str[last - 1] ) ) {                                // suffix 'l' ?
                        size = 3;
                        if ( checkL( str[last - 2] ) ) {                        // suffix "ll" ?
                                size = 4;
                        } // if
                } else if ( checkH( str[last - 1] ) ) {                 // suffix 'h' ?
                        size = 0;
                        if ( checkH( str[last - 2] ) ) {                        // suffix "hh" ?
                                size = 1;
                        } // if
                        str.erase( last - size - 1, size + 1 );         // remove 'h'/"hh"
                } // if
        } else if ( checkL( str[ last ] ) ) {                           // suffix 'l' ?
                size = 3;
                if ( checkL( str[last - 1] ) ) {                                // suffix 'll' ?
                        size = 4;
                        if ( checkU( str[last - 2] ) ) {                        // suffix 'u' ?
                                Unsigned = true;
                        } // if
                } else if ( checkU( str[last - 1] ) ) {                 // suffix 'u' ?
                        Unsigned = true;
                } // if
        } else if ( checkH( str[ last ] ) ) {                           // suffix 'h' ?
                size = 0;
                if ( checkH( str[last - 1] ) ) {                                // suffix "hh" ?
                        size = 1;
                        if ( checkU( str[last - 2] ) ) {                        // suffix 'u' ?
                                Unsigned = true;
                        } // if
                } else if ( checkU( str[last - 1] ) ) {                 // suffix 'u' ?
                        Unsigned = true;
                } // if
                str.erase( last - size, size + 1 );                             // remove 'h'/"hh"
        } else if ( checkZ( str[last] ) ) {                                     // suffix 'z' ?
                size = 5;
                str.erase( last, 1 );                                                   // remove 'z'
        } // if

        ret = new ConstantExpr( Constant( new BasicType( noQualifiers, kind[Unsigned][size] ), str, v ) );
        if ( Unsigned && size < 2 ) {                                           // less than int ?
                // int i = -1uh => 65535 not -1, so cast is necessary for unsigned, which eliminates warnings for large values.
                ret = new CastExpr( ret, new BasicType( Type::Qualifiers(), kind[Unsigned][size] ) );
        } else if ( size == 5 ) {                                                       // explicit cast to size_t
                ret = new CastExpr( ret, new TypeInstType( Type::Qualifiers(), "size_t", false ) );
        } // if
  CLEANUP:
        if ( units.length() != 0 ) {
                ret = new UntypedExpr( new NameExpr( units ), { ret } );
        } // if

        delete &str;                                                                            // created by lex
        return ret;
} // build_constantInteger

Expression * build_constantFloat( string & str ) {
        static const BasicType::Kind kind[2][3] = {
                { BasicType::Float, BasicType::Double, BasicType::LongDouble },
                { BasicType::FloatComplex, BasicType::DoubleComplex, BasicType::LongDoubleComplex },
        };

        string units;                                                                           // units
        sepNumeric( str, units );                                                       // separate constant from units

        bool complx = false;                                                            // real, complex
        int size = 1;                                                                           // 0 => float, 1 => double (default), 2 => long double
        // floating-point constant has minimum of 2 characters: 1. or .1
        size_t last = str.length() - 1;
        double v;

        sscanf( str.c_str(), "%lg", &v );

        if ( checkI( str[last] ) ) {                                            // imaginary ?
                complx = true;
                last -= 1;                                                                              // backup one character
        } // if

        if ( checkF( str[last] ) ) {                                            // float ?
                size = 0;
        } else if ( checkD( str[last] ) ) {                                     // double ?
                size = 1;
        } else if ( checkL( str[last] ) ) {                                     // long double ?
                size = 2;
        } // if
        if ( ! complx && checkI( str[last - 1] ) ) {            // imaginary ?
                complx = true;
        } // if

        Expression * ret = new ConstantExpr( Constant( new BasicType( noQualifiers, kind[complx][size] ), str, v ) );
        if ( units.length() != 0 ) {
                ret = new UntypedExpr( new NameExpr( units ), { ret } );
        } // if

        delete &str;                                                                            // created by lex
        return ret;
} // build_constantFloat

static void sepString( string & str, string & units, char delimit ) {
        string::size_type posn = str.find_last_of( delimit ) + 1;
        if ( posn != str.length() ) {
                units = "?" + str.substr( posn );                               // extract units
                str.erase( posn );                                                              // remove units
        } // if
} // sepString

Expression * build_constantChar( string & str ) {
        string units;                                                                           // units
        sepString( str, units, '\'' );                                          // separate constant from units

        Expression * ret = new ConstantExpr( Constant( new BasicType( noQualifiers, BasicType::Char ), str, (unsigned long long int)(unsigned char)str[1] ) );
        if ( units.length() != 0 ) {
                ret = new UntypedExpr( new NameExpr( units ), { ret } );
        } // if

        delete &str;                                                                            // created by lex
        return ret;
} // build_constantChar

Expression * build_constantStr( string & str ) {
        string units;                                                                           // units
        sepString( str, units, '"' );                                           // separate constant from units

        Type * strtype;
        switch ( str[0] ) {                                                                     // str has >= 2 characters, i.e, null string "" => safe to look at subscripts 0/1
          case 'u':
                if ( str[1] == '8' ) goto Default;                              // utf-8 characters => array of char
                // lookup type of associated typedef
                strtype = new TypeInstType( Type::Qualifiers( Type::Const ), "char16_t", false );
                break;
          case 'U':
                strtype = new TypeInstType( Type::Qualifiers( Type::Const ), "char32_t", false );
                break;
          case 'L':
                strtype = new TypeInstType( Type::Qualifiers( Type::Const ), "wchar_t", false );
                break;
          Default:                                                                                      // char default string type
          default:
                strtype = new BasicType( Type::Qualifiers( Type::Const ), BasicType::Char );
        } // switch
        ArrayType * at = new ArrayType( noQualifiers, strtype,
                                                                        new ConstantExpr( Constant::from_ulong( str.size() + 1 - 2 ) ), // +1 for '\0' and -2 for '"'
                                                                        false, false );
        Expression * ret = new ConstantExpr( Constant( at, str, (unsigned long long int)0 ) ); // constant 0 is ignored for pure string value
        if ( units.length() != 0 ) {
                ret = new UntypedExpr( new NameExpr( units ), { ret } );
        } // if

        delete &str;                                                                            // created by lex
        return ret;
} // build_constantStr

Expression * build_field_name_FLOATING_FRACTIONconstant( const string & str ) {
        if ( str.find_first_not_of( "0123456789", 1 ) != string::npos ) throw SemanticError( "invalid tuple index " + str );
        Expression * ret = build_constantInteger( *new string( str.substr(1) ) );
        delete &str;
        return ret;
} // build_field_name_FLOATING_FRACTIONconstant

Expression * build_field_name_FLOATING_DECIMALconstant( const string & str ) {
        if ( str[str.size()-1] != '.' ) throw SemanticError( "invalid tuple index " + str );
        Expression * ret = build_constantInteger( *new string( str.substr( 0, str.size()-1 ) ) );
        delete &str;
        return ret;
} // build_field_name_FLOATING_DECIMALconstant

Expression * build_field_name_FLOATINGconstant( const string & str ) {
        // str is of the form A.B -> separate at the . and return member expression
        int a, b;
        char dot;
        stringstream ss( str );
        ss >> a >> dot >> b;
        UntypedMemberExpr * ret = new UntypedMemberExpr( new ConstantExpr( Constant::from_int( b ) ), new ConstantExpr( Constant::from_int( a ) ) );
        delete &str;
        return ret;
} // build_field_name_FLOATINGconstant

Expression * make_field_name_fraction_constants( Expression * fieldName, Expression * fracts ) {
        if ( fracts ) {
                if ( UntypedMemberExpr * memberExpr = dynamic_cast< UntypedMemberExpr * >( fracts ) ) {
                        memberExpr->set_member( make_field_name_fraction_constants( fieldName, memberExpr->get_aggregate() ) );
                        return memberExpr;
                } else {
                        return new UntypedMemberExpr( fracts, fieldName );
                } // if
        } // if
        return fieldName;
} // make_field_name_fraction_constants

Expression * build_field_name_fraction_constants( Expression * fieldName, ExpressionNode * fracts ) {
        return make_field_name_fraction_constants( fieldName, maybeMoveBuild< Expression >( fracts ) );
} // build_field_name_fraction_constants

NameExpr * build_varref( const string * name ) {
        NameExpr * expr = new NameExpr( *name, nullptr );
        delete name;
        return expr;
} // build_varref

// TODO: get rid of this and OperKinds and reuse code from OperatorTable
static const char * OperName[] = {                                              // must harmonize with OperKinds
        // diadic
        "SizeOf", "AlignOf", "OffsetOf", "?+?", "?-?", "?\\?", "?*?", "?/?", "?%?", "||", "&&",
        "?|?", "?&?", "?^?", "Cast", "?<<?", "?>>?", "?<?", "?>?", "?<=?", "?>=?", "?==?", "?!=?",
        "?=?", "?@=?", "?\\=?", "?*=?", "?/=?", "?%=?", "?+=?", "?-=?", "?<<=?", "?>>=?", "?&=?", "?^=?", "?|=?",
        "?[?]", "...",
        // monadic
        "+?", "-?", "AddressOf", "*?", "!?", "~?", "++?", "?++", "--?", "?--",
}; // OperName

Expression * build_cast( DeclarationNode * decl_node, ExpressionNode * expr_node ) {
        Type * targetType = maybeMoveBuildType( decl_node );
        if ( dynamic_cast< VoidType * >( targetType ) ) {
                delete targetType;
                return new CastExpr( maybeMoveBuild< Expression >(expr_node) );
        } else {
                return new CastExpr( maybeMoveBuild< Expression >(expr_node), targetType );
        } // if
} // build_cast

Expression * build_virtual_cast( DeclarationNode * decl_node, ExpressionNode * expr_node ) {
        return new VirtualCastExpr( maybeMoveBuild< Expression >( expr_node ), maybeMoveBuildType( decl_node ) );
} // build_virtual_cast

Expression * build_fieldSel( ExpressionNode * expr_node, Expression * member ) {
        return new UntypedMemberExpr( member, maybeMoveBuild< Expression >(expr_node) );
} // build_fieldSel

Expression * build_pfieldSel( ExpressionNode * expr_node, Expression * member ) {
        UntypedExpr * deref = new UntypedExpr( new NameExpr( "*?" ) );
        deref->location = expr_node->location;
        deref->get_args().push_back( maybeMoveBuild< Expression >(expr_node) );
        UntypedMemberExpr * ret = new UntypedMemberExpr( member, deref );
        return ret;
} // build_pfieldSel

Expression * build_addressOf( ExpressionNode * expr_node ) {
        return new AddressExpr( maybeMoveBuild< Expression >(expr_node) );
} // build_addressOf

Expression * build_sizeOfexpr( ExpressionNode * expr_node ) {
        return new SizeofExpr( maybeMoveBuild< Expression >(expr_node) );
} // build_sizeOfexpr

Expression * build_sizeOftype( DeclarationNode * decl_node ) {
        return new SizeofExpr( maybeMoveBuildType( decl_node ) );
} // build_sizeOftype

Expression * build_alignOfexpr( ExpressionNode * expr_node ) {
        return new AlignofExpr( maybeMoveBuild< Expression >(expr_node) );
} // build_alignOfexpr

Expression * build_alignOftype( DeclarationNode * decl_node ) {
        return new AlignofExpr( maybeMoveBuildType( decl_node) );
} // build_alignOftype

Expression * build_offsetOf( DeclarationNode * decl_node, NameExpr * member ) {
        Expression * ret = new UntypedOffsetofExpr( maybeMoveBuildType( decl_node ), member->get_name() );
        delete member;
        return ret;
} // build_offsetOf

Expression * build_and_or( ExpressionNode * expr_node1, ExpressionNode * expr_node2, bool kind ) {
        return new LogicalExpr( notZeroExpr( maybeMoveBuild< Expression >(expr_node1) ), notZeroExpr( maybeMoveBuild< Expression >(expr_node2) ), kind );
} // build_and_or

Expression * build_unary_val( OperKinds op, ExpressionNode * expr_node ) {
        list< Expression * > args;
        args.push_back( maybeMoveBuild< Expression >(expr_node) );
        return new UntypedExpr( new NameExpr( OperName[ (int)op ] ), args );
} // build_unary_val

Expression * build_unary_ptr( OperKinds op, ExpressionNode * expr_node ) {
        list< Expression * > args;
        args.push_back(  maybeMoveBuild< Expression >(expr_node) ); // xxx -- this is exactly the same as the val case now, refactor this code.
        return new UntypedExpr( new NameExpr( OperName[ (int)op ] ), args );
} // build_unary_ptr

Expression * build_binary_val( OperKinds op, ExpressionNode * expr_node1, ExpressionNode * expr_node2 ) {
        list< Expression * > args;
        args.push_back( maybeMoveBuild< Expression >(expr_node1) );
        args.push_back( maybeMoveBuild< Expression >(expr_node2) );
        return new UntypedExpr( new NameExpr( OperName[ (int)op ] ), args );
} // build_binary_val

Expression * build_binary_ptr( OperKinds op, ExpressionNode * expr_node1, ExpressionNode * expr_node2 ) {
        list< Expression * > args;
        args.push_back( maybeMoveBuild< Expression >(expr_node1) );
        args.push_back( maybeMoveBuild< Expression >(expr_node2) );
        return new UntypedExpr( new NameExpr( OperName[ (int)op ] ), args );
} // build_binary_ptr

Expression * build_cond( ExpressionNode * expr_node1, ExpressionNode * expr_node2, ExpressionNode * expr_node3 ) {
        return new ConditionalExpr( notZeroExpr( maybeMoveBuild< Expression >(expr_node1) ), maybeMoveBuild< Expression >(expr_node2), maybeMoveBuild< Expression >(expr_node3) );
} // build_cond

Expression * build_attrexpr( NameExpr * var, ExpressionNode * expr_node ) {
        return new AttrExpr( var, maybeMoveBuild< Expression >(expr_node) );
} // build_attrexpr

Expression * build_attrtype( NameExpr * var, DeclarationNode * decl_node ) {
        return new AttrExpr( var, maybeMoveBuildType( decl_node ) );
} // build_attrtype

Expression * build_tuple( ExpressionNode * expr_node ) {
        list< Expression * > exprs;
        buildMoveList( expr_node, exprs );
        return new UntypedTupleExpr( exprs );;
} // build_tuple

Expression * build_func( ExpressionNode * function, ExpressionNode * expr_node ) {
        list< Expression * > args;
        buildMoveList( expr_node, args );
        return new UntypedExpr( maybeMoveBuild< Expression >(function), args, nullptr );
} // build_func

Expression * build_range( ExpressionNode * low, ExpressionNode * high ) {
        return new RangeExpr( maybeMoveBuild< Expression >( low ), maybeMoveBuild< Expression >( high ) );
} // build_range

Expression * build_compoundLiteral( DeclarationNode * decl_node, InitializerNode * kids ) {
        Declaration * newDecl = maybeBuild< Declaration >(decl_node); // compound literal type
        if ( DeclarationWithType * newDeclWithType = dynamic_cast< DeclarationWithType * >( newDecl ) ) { // non-sue compound-literal type
                return new CompoundLiteralExpr( newDeclWithType->get_type(), maybeMoveBuild< Initializer >(kids) );
        // these types do not have associated type information
        } else if ( StructDecl * newDeclStructDecl = dynamic_cast< StructDecl * >( newDecl )  ) {
                if ( newDeclStructDecl->has_body() ) {
                        return new CompoundLiteralExpr( new StructInstType( Type::Qualifiers(), newDeclStructDecl ), maybeMoveBuild< Initializer >(kids) );
                } else {
                        return new CompoundLiteralExpr( new StructInstType( Type::Qualifiers(), newDeclStructDecl->get_name() ), maybeMoveBuild< Initializer >(kids) );
                } // if
        } else if ( UnionDecl * newDeclUnionDecl = dynamic_cast< UnionDecl * >( newDecl )  ) {
                if ( newDeclUnionDecl->has_body() ) {
                        return new CompoundLiteralExpr( new UnionInstType( Type::Qualifiers(), newDeclUnionDecl ), maybeMoveBuild< Initializer >(kids) );
                } else {
                        return new CompoundLiteralExpr( new UnionInstType( Type::Qualifiers(), newDeclUnionDecl->get_name() ), maybeMoveBuild< Initializer >(kids) );
                } // if
        } else if ( EnumDecl * newDeclEnumDecl = dynamic_cast< EnumDecl * >( newDecl )  ) {
                if ( newDeclEnumDecl->has_body() ) {
                        return new CompoundLiteralExpr( new EnumInstType( Type::Qualifiers(), newDeclEnumDecl ), maybeMoveBuild< Initializer >(kids) );
                } else {
                        return new CompoundLiteralExpr( new EnumInstType( Type::Qualifiers(), newDeclEnumDecl->get_name() ), maybeMoveBuild< Initializer >(kids) );
                } // if
        } else {
                assert( false );
        } // if
} // build_compoundLiteral

// Local Variables: //
// tab-width: 4 //
// mode: c++ //
// compile-command: "make install" //
// End: //