C++ Cli

ECMA-372
1st Edition / December 2005
C++/CLI Language Specification
Standard ECMA-372
1st Edition / December 2005
Ecma International
Rue du Rhne 114
CH-1204 Geneva
T/F: +41 22 849 6000/01
www.ecma-international.org
Table of Contents
Table of Contents
Introduction ...................................................................................................................................................xii 1. Scope............................................................................................................................................................. 1 2. Conformance ............................................................................................................................................... 2 3. Normative references .................................................................................................................................. 3 4. Definitions .................................................................................................................................................... 4 5. Notational conventions................................................................................................................................ 7 6. Acronyms and abbreviations ..................................................................................................................... 8 7. General description ..................................................................................................................................... 9 8. Language overview.................................................................................................................................... 10 8.1 Getting started ....................................................................................................................................... 10 8.2 Types ..................................................................................................................................................... 10 8.2.1 Fundamental types and the CLI ...................................................................................................... 12 8.2.2 Conversions .................................................................................................................................... 13 8.2.3 CLI array types ............................................................................................................................... 13 8.2.4 Type system unification.................................................................................................................. 13 8.2.5 Pointers, handles, and null .............................................................................................................. 14 8.3 Parameters ............................................................................................................................................. 16 8.4 Automatic memory management........................................................................................................... 17 8.5 Expressions............................................................................................................................................ 18 8.6 Statements.............................................................................................................................................. 19 8.7 Delegates ............................................................................................................................................... 19 8.8 Native and ref classes ............................................................................................................................ 20 8.8.1 Literal fields.................................................................................................................................... 20 8.8.2 Initonly fields.................................................................................................................................. 21 8.8.3 Functions......................................................................................................................................... 21 8.8.4 Properties ........................................................................................................................................ 21 8.8.5 Events.............................................................................................................................................. 23 8.8.6 Static operators ............................................................................................................................... 24 8.8.7 Instance constructors....................................................................................................................... 25 8.8.8 Destructors and finalizers ............................................................................................................... 25 8.8.9 Static constructors........................................................................................................................... 26 8.8.10 Inheritance .................................................................................................................................... 27 8.9 Value classes ......................................................................................................................................... 28 8.10 Interfaces ............................................................................................................................................. 28 8.11 Enums .................................................................................................................................................. 30 8.12 Namespaces and assemblies ................................................................................................................ 30 8.13 Versioning ........................................................................................................................................... 31 8.14 Attributes ............................................................................................................................................. 32 8.15 Generics............................................................................................................................................... 33 8.15.1 Creating and consuming generics ................................................................................................. 33 8.15.2 Constraints .................................................................................................................................... 34 8.15.3 Generic functions .......................................................................................................................... 35 9. Lexical structure........................................................................................................................................ 37 9.1 Tokens ................................................................................................................................................... 37 9.1.1 Identifiers ........................................................................................................................................ 37 9.1.2 Keywords ........................................................................................................................................ 38
iii
9.1.3 Literals ............................................................................................................................................ 39 9.1.4 Operators and punctuators .............................................................................................................. 40 10. Basic concepts .......................................................................................................................................... 41 10.1 Assemblies........................................................................................................................................... 41 10.2 Application entry point........................................................................................................................ 41 10.3 Importing types from assemblies......................................................................................................... 41 10.4 Reserved names ................................................................................................................................... 42 10.5 Members .............................................................................................................................................. 43 10.5.1 Value class members..................................................................................................................... 43 10.5.2 Delegate members......................................................................................................................... 43 10.6 Member access .................................................................................................................................... 43 10.6.1 Declared accessibility ................................................................................................................... 43 10.7 Name lookup ....................................................................................................................................... 44 11. Preprocessor ............................................................................................................................................ 48 11.1 Conditional inclusion........................................................................................................................... 48 11.2 Predefined macro names...................................................................................................................... 48 12. Types ........................................................................................................................................................ 49 12.1 Value types .......................................................................................................................................... 50 12.1.1 Fundamental types ........................................................................................................................ 50 12.2 Class types ........................................................................................................................................... 51 12.2.1 Value classes................................................................................................................................. 51 12.2.2 Ref classes..................................................................................................................................... 51 12.2.3 Interface classes ............................................................................................................................ 51 12.2.4 Delegate types............................................................................................................................... 51 12.3 Declarator types................................................................................................................................... 52 12.3.1 Raw types...................................................................................................................................... 52 12.3.2 Pointer types ................................................................................................................................. 52 12.3.3 Handle types ................................................................................................................................. 52 12.3.4 Null type ....................................................................................................................................... 53 12.3.5 Reference types............................................................................................................................. 53 12.3.6 Interior pointers............................................................................................................................. 54 12.3.7 Pinning pointers ............................................................................................................................ 55 12.3.8 Native arrays ................................................................................................................................. 57 12.4 Top-level type visibility ...................................................................................................................... 57 13. Variables .................................................................................................................................................. 58 13.1 gc-lvalues............................................................................................................................................. 58 13.1.1 Standard conversions .................................................................................................................... 58 13.1.2 Expressions ................................................................................................................................... 59 13.1.3 Reference initializers .................................................................................................................... 60 13.1.4 Temporary objects ........................................................................................................................ 60 13.2 File-scope and namespace-scope variables ......................................................................................... 60 13.3 Direct initialization .............................................................................................................................. 60 14. Conversions.............................................................................................................................................. 62 14.1 Conversion sequences ......................................................................................................................... 62 14.2 Standard conversions........................................................................................................................... 62 14.2.1 Handle conversions....................................................................................................................... 62 14.2.2 Pointer conversions....................................................................................................................... 63 14.2.3 Lvalue conversions ....................................................................................................................... 64 14.2.4 Integral promotions....................................................................................................................... 64 14.2.5 String literal conversions .............................................................................................................. 65 14.2.6 Boxing conversions....................................................................................................................... 66
iv
Table of Contents
14.3 Implicit conversions ............................................................................................................................ 66 14.3.1 Implicit constant expression conversions ..................................................................................... 66 14.3.2 User-defined implicit conversions ................................................................................................ 66 14.3.3 Boolean Equivalence .................................................................................................................... 66 14.4 Explicit conversions ............................................................................................................................ 67 14.5 User-defined conversions .................................................................................................................... 67 14.5.1 Constructors .................................................................................................................................. 67 14.5.2 Explicit conversion functions ....................................................................................................... 67 14.5.3 Static conversion functions ........................................................................................................... 67 14.6 Parameter array conversions................................................................................................................ 67 14.7 Naming conventions............................................................................................................................ 68 15. Expressions .............................................................................................................................................. 70 15.1 Function members ............................................................................................................................... 70 15.2 Primary expressions............................................................................................................................. 71 15.3 Postfix expressions .............................................................................................................................. 71 15.3.1 Subscripting and indexed access................................................................................................... 72 15.3.2 Function call ................................................................................................................................. 72 15.3.3 Explicit type conversion (functional notation).............................................................................. 72 15.3.4 Class member access..................................................................................................................... 73 15.3.5 Increment and decrement.............................................................................................................. 73 15.3.6 Dynamic cast................................................................................................................................. 73 15.3.7 Type identification ........................................................................................................................ 74 15.3.8 Static cast ...................................................................................................................................... 75 15.3.9 Reinterpret cast ............................................................................................................................. 76 15.3.10 Const cast.................................................................................................................................... 76 15.3.11 Safe cast ...................................................................................................................................... 76 15.4 Unary expressions ............................................................................................................................... 77 15.4.1 Unary operators............................................................................................................................. 77 15.4.2 Increment and decrement.............................................................................................................. 79 15.4.3 Sizeof ............................................................................................................................................ 80 15.4.4 New............................................................................................................................................... 80 15.4.5 Delete ............................................................................................................................................ 80 15.4.6 The gcnew operator....................................................................................................................... 81 15.4.7 The throw expression.................................................................................................................... 81 15.5 Explicit type conversion (cast notation) .............................................................................................. 81 15.6 Additive operators ............................................................................................................................... 82 15.6.1 Delegate combination ................................................................................................................... 82 15.6.2 Delegate removal .......................................................................................................................... 82 15.6.3 String concatenation...................................................................................................................... 82 15.7 Shift operators ..................................................................................................................................... 83 15.8 Relational operators............................................................................................................................. 83 15.8.1 Handle equality operators ............................................................................................................. 83 15.8.2 Delegate equality operators........................................................................................................... 84 15.8.3 String equality............................................................................................................................... 85 15.9 Logical AND operator ......................................................................................................................... 85 15.10 Logical OR operator .......................................................................................................................... 85 15.11 Conditional operator .......................................................................................................................... 85 15.12 Assignment operators ........................................................................................................................ 85 15.13 Constant expressions ......................................................................................................................... 86 15.14 Property and event rewrite rules ........................................................................................................ 86 16. Statements................................................................................................................................................ 89 16.1 Selection statements ............................................................................................................................ 89 16.1.1 The switch statement..................................................................................................................... 89
16.2 Iteration statements.............................................................................................................................. 89 16.2.1 The for each statement .................................................................................................................. 89 16.3 Jump statements................................................................................................................................... 91 16.3.1 The break statement ...................................................................................................................... 91 16.3.2 The continue statement ................................................................................................................. 91 16.3.3 The return statement ..................................................................................................................... 91 16.3.4 The goto statement........................................................................................................................ 91 16.4 The try block........................................................................................................................................ 91 17. Namespaces.............................................................................................................................................. 93 17.1 Reserved namespaces .......................................................................................................................... 93 18. Functions.................................................................................................................................................. 94 18.1 <cstdarg>-style variable-argument lists............................................................................................... 94 18.2 Name lookup ....................................................................................................................................... 94 18.3 Overload resolution ............................................................................................................................. 94 18.4 Parameter arrays .................................................................................................................................. 94 18.5 Importing native functions................................................................................................................... 96 18.6 Non-member functions ........................................................................................................................ 97 18.7 Attributes ............................................................................................................................................. 97 19. Classes and members .............................................................................................................................. 98 19.1 Class definitions .................................................................................................................................. 98 19.1.1 Class modifiers ............................................................................................................................. 99 19.2 Reserved member names ................................................................................................................... 100 19.2.1 Member names reserved for properties....................................................................................... 100 19.2.2 Member names reserved for events ............................................................................................ 101 19.2.3 Member names reserved for functions........................................................................................ 101 19.2.4 Possible collision with reserved property and event names........................................................ 102 19.3 Data members.................................................................................................................................... 103 19.4 Functions ........................................................................................................................................... 103 19.4.1 Override functions ...................................................................................................................... 104 19.4.2 Sealed function modifier............................................................................................................. 107 19.4.3 Abstract function modifier.......................................................................................................... 107 19.4.4 New function modifier................................................................................................................ 108 19.5 Properties........................................................................................................................................... 109 19.5.1 Qualified names of properties and events ................................................................................... 110 19.5.2 Static and instance properties...................................................................................................... 111 19.5.3 Accessor functions ...................................................................................................................... 111 19.5.4 Virtual, sealed, abstract, and override accessor functions .......................................................... 113 19.5.5 Trivial scalar properties .............................................................................................................. 114 19.6 Events ................................................................................................................................................ 115 19.6.1 Static and instance events ........................................................................................................... 116 19.6.2 Accessor functions ...................................................................................................................... 116 19.6.3 Virtual, sealed, abstract, and override accessor functions .......................................................... 117 19.6.4 Trivial events .............................................................................................................................. 117 19.6.5 Event invocation ......................................................................................................................... 117 19.7 Static operators .................................................................................................................................. 117 19.7.1 Homogenizing the candidate overload set .................................................................................. 119 19.7.2 Operators on handles................................................................................................................... 119 19.7.3 Increment and decrement operators ............................................................................................ 120 19.7.4 Operator synthesis....................................................................................................................... 123 19.7.5 Naming conventions ................................................................................................................... 123 19.8 Non-static operators........................................................................................................................... 126 19.9 Instance constructors ......................................................................................................................... 126 19.10 Static constructors ........................................................................................................................... 127 vi
Table of Contents
19.11 Literal fields..................................................................................................................................... 128 19.12 Initonly fields................................................................................................................................... 129 19.12.1 Using static initonly fields for constants................................................................................... 130 19.12.2 Versioning of literal fields and static initonly fields................................................................. 130 19.13 Destructors and finalizers ................................................................................................................ 130 19.13.1 Destructors ................................................................................................................................ 131 19.13.2 Finalizers................................................................................................................................... 131 20. Native classes ......................................................................................................................................... 133 20.1 Functions ........................................................................................................................................... 133 20.2 Properties........................................................................................................................................... 133 20.3 Static operators .................................................................................................................................. 133 20.4 Delegates ........................................................................................................................................... 133 20.5 Friends ............................................................................................................................................... 133 20.6 Events ................................................................................................................................................ 134 20.7 Finalizer............................................................................................................................................. 134 20.8 Initonly and literal fields.................................................................................................................... 134 20.9 Static constructors ............................................................................................................................. 134 21. Ref classes .............................................................................................................................................. 135 21.1 Ref class definitions .......................................................................................................................... 135 21.1.1 Ref class base specification ........................................................................................................ 135 21.2 Ref class members ............................................................................................................................. 135 21.2.1 Variable initializers..................................................................................................................... 135 21.3 Functions ........................................................................................................................................... 136 21.4 Properties........................................................................................................................................... 136 21.5 Events ................................................................................................................................................ 136 21.6 Static operators .................................................................................................................................. 137 21.7 Non-static operators........................................................................................................................... 137 21.8 Instance constructors ......................................................................................................................... 137 21.9 Static constructor ............................................................................................................................... 137 21.10 Literal fields..................................................................................................................................... 137 21.11 Initonly fields................................................................................................................................... 137 21.12 Destructors and finalizers ................................................................................................................ 137 21.13 Delegates ......................................................................................................................................... 137 22. Value classes .......................................................................................................................................... 138 22.1 Value class definitions....................................................................................................................... 138 22.1.1 Value class base specification..................................................................................................... 138 22.2 Value class members ......................................................................................................................... 138 22.3 Ref class and value class differences................................................................................................. 139 22.3.1 Inheritance .................................................................................................................................. 139 22.3.2 Default values ............................................................................................................................. 139 22.3.3 Meaning of this ........................................................................................................................... 139 22.3.4 Destructors and finalizers ........................................................................................................... 139 22.4 Simple value classes .......................................................................................................................... 140 22.5 Constructors....................................................................................................................................... 140 22.6 Operators ........................................................................................................................................... 140 23. Mixed types ............................................................................................................................................ 141 24. CLI arrays.............................................................................................................................................. 142 24.1 CLI array types .................................................................................................................................. 142 24.1.1 The System::Array type .............................................................................................................. 142 24.2 CLI array creation.............................................................................................................................. 143 24.3 CLI array element access................................................................................................................... 143
vii
24.4 CLI array members............................................................................................................................ 144 24.5 CLI array covariance ......................................................................................................................... 144 24.6 CLI array initializers.......................................................................................................................... 144 25. Interfaces................................................................................................................................................ 146 25.1 Interface definitions........................................................................................................................... 146 25.1.1 Interface base specification......................................................................................................... 146 25.2 Interface members ............................................................................................................................. 146 25.2.1 Functions..................................................................................................................................... 147 25.2.2 Properties .................................................................................................................................... 147 25.2.3 Events.......................................................................................................................................... 147 25.2.4 Delegates..................................................................................................................................... 148 25.2.5 Member access............................................................................................................................ 148 25.2.6 Destructors and finalizers ........................................................................................................... 148 25.3 Interface implementations ................................................................................................................. 148 26. Enums..................................................................................................................................................... 150 26.1 Enum definitions ............................................................................................................................... 150 26.1.1 Enum base specification ............................................................................................................. 151 26.1.2 Initial enumerator values............................................................................................................. 151 26.1.3 CLI enum values and operations................................................................................................. 151 26.2 The System::Flags attribute ............................................................................................................... 151 27. Delegates................................................................................................................................................. 153 27.1 Delegate definitions........................................................................................................................... 153 27.2 Delegate instantiation ........................................................................................................................ 155 27.3 Delegate invocation ........................................................................................................................... 156 28. Exceptions and exception handling ..................................................................................................... 157 28.1 Common exception classes................................................................................................................ 157 28.2 Exception specifications .................................................................................................................... 158 29. Attributes ............................................................................................................................................... 159 29.1 Attribute classes................................................................................................................................. 159 29.1.1 Attribute usage ............................................................................................................................ 159 29.1.2 Positional and named parameters................................................................................................ 160 29.1.3 Attribute parameter types............................................................................................................ 161 29.2 Attribute specification ....................................................................................................................... 161 29.3 Attribute instances ............................................................................................................................. 165 29.3.1 Compilation of an attribute ......................................................................................................... 165 29.3.2 Run-time retrieval of an attribute instance.................................................................................. 166 29.4 Reserved attributes ............................................................................................................................ 166 29.4.1 The AttributeUsage attribute....................................................................................................... 166 29.4.2 The Obsolete attribute................................................................................................................. 166 29.4.3 The Conditional attribute ............................................................................................................ 167 29.4.4 Security attributes ....................................................................................................................... 167 29.5 Attributes for interoperation .............................................................................................................. 167 29.5.1 Interoperation with other CLI-based languages.......................................................................... 167 29.5.2 Interoperation with native code .................................................................................................. 167 30. Templates ............................................................................................................................................... 168 30.1 Template declarations........................................................................................................................ 168 30.2 Template specialization ..................................................................................................................... 168 30.3 Attributes ........................................................................................................................................... 168 30.4 Type deduction .................................................................................................................................. 169 30.4.1 Template argument deduction..................................................................................................... 169
viii
Table of Contents
31. Generics.................................................................................................................................................. 170 31.1 Generic declarations .......................................................................................................................... 170 31.1.1 Type parameters.......................................................................................................................... 171 31.1.2 Referencing a generic type by name ........................................................................................... 172 31.1.3 The instance type ........................................................................................................................ 172 31.1.4 Base classes and interfaces ......................................................................................................... 173 31.1.5 Class members ............................................................................................................................ 173 31.1.6 Static members............................................................................................................................ 174 31.1.7 Operators..................................................................................................................................... 175 31.1.8 Member overloading................................................................................................................... 175 31.1.9 Member overriding ..................................................................................................................... 176 31.1.10 Nested types.............................................................................................................................. 176 31.2 Constructed types .............................................................................................................................. 177 31.2.1 Open and closed constructed types ............................................................................................. 178 31.2.2 Type arguments........................................................................................................................... 178 31.2.3 Base classes and interfaces ......................................................................................................... 179 31.2.4 Class members ............................................................................................................................ 179 31.2.5 Accessibility................................................................................................................................ 180 31.3 Generic functions............................................................................................................................... 180 31.3.1 Function signature matching rules .............................................................................................. 181 31.3.2 Type deduction ........................................................................................................................... 182 31.4 Constraints......................................................................................................................................... 184 31.4.1 Satisfying constraints .................................................................................................................. 185 31.4.2 Member lookup on type parameters............................................................................................ 187 31.4.3 Type parameters and boxing....................................................................................................... 188 31.4.4 Conversions involving type parameters...................................................................................... 189 32. Standard C and C++ libraries.............................................................................................................. 190 33. CLI libraries .......................................................................................................................................... 191 33.1 Custom modifiers .............................................................................................................................. 191 33.1.1 Signature matching ..................................................................................................................... 191 33.1.2 modreq vs. modopt...................................................................................................................... 192 33.1.3 Modifier syntax........................................................................................................................... 192 33.1.4 Types having multiple custom modifiers.................................................................................... 193 33.1.5 Standard custom modifiers ......................................................................................................... 194 33.2 Standard attributes ............................................................................................................................. 199 33.2.1 NativeCppClass .......................................................................................................................... 199 34. Metadata ................................................................................................................................................ 200 34.1 Basic concepts ................................................................................................................................... 200 34.1.1 Importing types from assemblies ................................................................................................ 200 34.2 Types ................................................................................................................................................. 200 34.2.1 Reference types........................................................................................................................... 200 34.2.2 Interior pointers........................................................................................................................... 201 34.2.3 Pinning pointers .......................................................................................................................... 201 34.2.4 Native arrays ............................................................................................................................... 202 34.3 Variables............................................................................................................................................ 202 34.3.1 File-scope and namespace-scope variables................................................................................. 202 34.4 Conversions ....................................................................................................................................... 202 34.4.1 String literal conversions ............................................................................................................ 202 34.4.2 Boxing conversions..................................................................................................................... 202 34.4.3 Conversion functions .................................................................................................................. 203 34.5 Expressions........................................................................................................................................ 203 34.5.1 Class member access................................................................................................................... 203
ix
34.5.2 Dynamic cast............................................................................................................................... 204 34.5.3 Safe cast ...................................................................................................................................... 204 34.6 Functions ........................................................................................................................................... 204 34.6.1 Name lookup............................................................................................................................... 204 34.6.2 Parameter arrays.......................................................................................................................... 204 34.6.3 Importing native functions.......................................................................................................... 205 34.6.4 Non-member functions ............................................................................................................... 206 34.7 Classes and members......................................................................................................................... 206 34.7.1 Class definitions.......................................................................................................................... 206 34.7.2 Member access............................................................................................................................ 208 34.7.3 Data members ............................................................................................................................. 209 34.7.4 Functions..................................................................................................................................... 210 34.7.5 Properties .................................................................................................................................... 213 34.7.6 Events.......................................................................................................................................... 215 34.7.7 Static operators ........................................................................................................................... 217 34.7.8 Non-static operators .................................................................................................................... 218 34.7.9 Instance constructors................................................................................................................... 219 34.7.10 Static constructors..................................................................................................................... 220 34.7.11 Literal fields .............................................................................................................................. 220 34.7.12 Initonly fields............................................................................................................................ 220 34.7.13 Destructors and finalizers ......................................................................................................... 221 34.8 Native classes .................................................................................................................................... 228 34.9 Ref classes ......................................................................................................................................... 230 34.10 Value classes ................................................................................................................................... 230 34.11 CLI arrays........................................................................................................................................ 231 34.12 Interfaces ......................................................................................................................................... 232 34.13 Enums .............................................................................................................................................. 233 34.14 Delegates ......................................................................................................................................... 234 34.15 Exceptions ....................................................................................................................................... 235 34.16 Attributes ......................................................................................................................................... 236 34.17 Templates ........................................................................................................................................ 239 34.18 Generics........................................................................................................................................... 239 Annex A. Grammar .................................................................................................................................... 240 A.1 Keywords............................................................................................................................................ 240 A.2 Lexical conventions............................................................................................................................ 240 A.3 Basic concepts .................................................................................................................................... 243 A.4 Expressions......................................................................................................................................... 244 A.5 Statements........................................................................................................................................... 247 A.6 Declarations ........................................................................................................................................ 248 A.7 Declarators.......................................................................................................................................... 250 A.8 Classes ................................................................................................................................................ 252 A.9 Properties and events .......................................................................................................................... 253 A.10 Derived classes ................................................................................................................................. 254 A.11 Special member functions................................................................................................................. 254 A.12 Overloading ...................................................................................................................................... 255 A.13 Delegates .......................................................................................................................................... 255 A.14 Templates.......................................................................................................................................... 255 A.15 Generics ............................................................................................................................................ 256 A.16 Exception handling ........................................................................................................................... 257 A.17 Attributes .......................................................................................................................................... 257 A.18 Preprocessing directives ................................................................................................................... 258 Annex B. Verifiable code ............................................................................................................................ 260 Annex C. Documentation comments ......................................................................................................... 261 x
Table of Contents
C.1 Introduction......................................................................................................................................... 261 C.2 Recommended tags ............................................................................................................................. 262 C.2.1 <c> ............................................................................................................................................... 262 C.2.2 <code>.......................................................................................................................................... 263 C.2.3 <example>.................................................................................................................................... 263 C.2.4 <exception>.................................................................................................................................. 263 C.2.5 <list> ............................................................................................................................................ 264 C.2.6 <para> .......................................................................................................................................... 265 C.2.7 <param> ....................................................................................................................................... 265 C.2.8 <paramref>................................................................................................................................... 265 C.2.9 <permission>................................................................................................................................ 266 C.2.10 <remarks> .................................................................................................................................. 266 C.2.11 <returns> .................................................................................................................................... 267 C.2.12 <see> .......................................................................................................................................... 267 C.2.13 <seealso>.................................................................................................................................... 267 C.2.14 <summary> ................................................................................................................................ 268 C.2.15 <typeparam> .............................................................................................................................. 268 C.2.16 <typeparamref> .......................................................................................................................... 269 C.2.17 <value>....................................................................................................................................... 269 C.3 Processing the documentation file ...................................................................................................... 269 C.3.1 ID string format............................................................................................................................ 269 C.3.2 ID string examples ....................................................................................................................... 270 C.4 An example ......................................................................................................................................... 273 C.4.1 C++ source code........................................................................................................................... 273 C.4.2 Resulting XML............................................................................................................................. 276 Annex D. Non-normative references ......................................................................................................... 279 Annex E. CLI naming guidelines ............................................................................................................... 280 Annex F. Future directions......................................................................................................................... 281 F.1 Expressions.......................................................................................................................................... 281 F.1.1 Class member access .................................................................................................................... 281 F.1.2 Type identification........................................................................................................................ 281 F.1.3 Pointer type portability ................................................................................................................. 281 F.2 Statements ........................................................................................................................................... 281 F.2.1 The checked and unchecked statements ....................................................................................... 281 F.3 Classes................................................................................................................................................. 281 F.3.1 Delegating constructors ................................................................................................................ 281 F.3.2 Properties ...................................................................................................................................... 283 F.3.3 Events ........................................................................................................................................... 283 F.3.4 Unsupported CLS-recommended operators.................................................................................. 283 F.3.5 Operators true and false ................................................................................................................ 284 F.4 Generic types....................................................................................................................................... 284 F.5 Custom modifiers ................................................................................................................................ 284 F.5.1 IsPinned ........................................................................................................................................ 284 F.6 Attributes............................................................................................................................................. 284 Annex G. Portability issues ........................................................................................................................ 285 G.1 Undefined behavior ............................................................................................................................ 285 G.2 Implementation-defined behavior....................................................................................................... 285 G.3 Unspecified behavior.......................................................................................................................... 285 Annex H. Index............................................................................................................................................ 286
xi
Introduction
This Standard is based on a submission from Microsoft. It describes a technology, called C++/CLI, which is a binding between the Standard C++ programming language and the Common Language Infrastructure (CLI). That submission evolved from another Microsoft project, Managed Extensions for C++, the first widely distributed implementation of which was released by Microsoft in July 2000, as part of its .NET Framework initiative. The first widely distributed beta implementation of C++/CLI was released by Microsoft in July 2004. Ecma Technical Committee 39 (TC39) Task Group 5 (TG5) was formed in October 2003, to produce a standard for C++/CLI. (Another Task Group, TG3, was formed in September 2000 to produce a standard for a library and execution environment called Common Language Infrastructure. The current version of that standard is ECMA-335, 3rd edition, June 2005. CLI is based on a subset of the .NET Framework.) The goals used in the design of C++/CLI were as follows: Provide an elegant and uniform syntax and semantics that give a natural feel for C++ programmers. Provide first-class support for CLI features (e.g., properties, events, garbage collection, and generics) for all types including existing Standard C++ classes. Provide first-class support for Standard C++ features (e.g., deterministic destruction, templates) for all types including CLI classes. Preserve the meaning of existing Standard C++ programs by specifying pure extensions wherever possible.
The development of this standard started in December 2003. It is expected there will be future revisions to this standard, primarily to add new functionality.
xii
Scope
1. Scope
This Standard specifies requirements for implementations of the C++/CLI binding. The first such requirement is that they implement the binding, and so this Standard also defines C++/CLI. Other requirements and relaxations of the first requirement appear at various places within this Standard. C++/CLI is an extension of the C++ programming language as described in ISO/IEC 14882:2003, Programming languages C++. In addition to the facilities provided by C++, C++/CLI provides additional keywords, classes, exceptions, namespaces, and library facilities, as well as garbage collection.
2. Conformance
Clause 1.4, Implementation compliance, of the C++ Standard applies to this Standard.
Normative references
3. Normative references
The following normative documents contain provisions, which, through reference in this text, constitute provisions of this Standard. For dated references, subsequent amendments to, or revisions of, any of these publications do not apply. However, parties to agreements based on this Standard are encouraged to investigate the possibility of applying the most recent editions of the normative documents indicated below. For undated references, the latest edition of the normative document referred to applies. Members of ISO and IEC maintain registers of currently valid International Standards. ECMA-335, 3rd edition, June 2005, Common Language Infrastructure (CLI), all Partitions and the accompanying library XML. ISO/IEC 2382.1:1993, Information technology Vocabulary Part 1: Fundamental terms. ISO/IEC 10646 (all parts), Information technology Universal Multiple-Octet Coded Character Set (UCS). ISO/IEC 14882:2003, Programming languages C++. [Note: Revision of the C++ Standard is currently underway, and changes proposed in that revision will affect future versions of this C++/CLI standard. For an example, see 9.1.1. end note] IEC 60559:1989, Binary floating-point arithmetic for microprocessor systems (previously designated IEC 559:1989). (This standard is widely known by its U.S. national designation, ANSI/IEEE Standard 754-1985, IEEE Standard for Binary Floating-Point Arithmetic.) This Standard supports the same version of Unicode as the CLI standard.
4. Definitions
For the purposes of this Standard, the following definitions apply. Other terms are defined where they appear in italic type or on the left side of a syntax rule. Terms explicitly defined in this Standard are not to be presumed to refer implicitly to similar terms defined elsewhere. Terms not defined in this Standard are to be interpreted according to the C++ Standard, ISO/IEC 14882:2003. application Refers to an assembly that has an entry point. When an application is run, a new application domain is created. Several different instantiations of an application can exist on the same machine at the same time, and each has its own application domain. application domain An entity that enables application isolation by acting as a container for application state. An application domain acts as a container and boundary for the types defined in the application and the class libraries it uses. A type loaded into one application domain is distinct from the same type loaded into another application domain, and objects on the CLI heap are not directly shared between application domains. Each application domain has its own copy of static variables for these types, and a static constructor for a type is run at most once per application domain. Implementations are free to provide implementation-specific policy or mechanisms for the creation and destruction of application domains. assembly Refers to one or more files that are output by the compiler as a result of program compilation. An assembly is a configured set of loadable code modules and other resources that together implement a unit of functionality. An assembly can contain types, the executable code used to implement these types, and references to other assemblies. The physical representation of an assembly is defined by the CLI Standard (3). Essentially, an assembly is the output of the compiler. An assembly that has an entry point is called an application. (See also metadata.) attribute A characteristic of a type and/or its members that contains descriptive information. While the most common attributes are predefined, and have a specific encoding in the metadata associated with them, user-defined attributes can also be added to the metadata. boxing An explicit or implicit conversion from any value class type V to type V^, in which a V box is allocated on the CLI heap, and the value is copied into that box. (See also unboxing.) CIL Common Intermediate Language, the instruction set of the Virtual Execution System. This instruction set is defined in Partition III of the CLI Standard (3). CLI array A CLI-specific array. A Standard C++-style array is referred to as a native array or, more simply, array, whenever the distinction is needed. A CLI array differs from a native array in that the former is allocated on the CLI heap, and can have a rank other than one. CLS compliance The Common Language Specification (CLS) defines language interoperability rules, which apply only to items that are visible outside of their defining assembly. CLS compliance is described in Partition I of the CLI Standard (3). definition, out-of-class A synonym for what Standard C++ calls a non-inline definition. delegate A ref class such that an instance of it can encapsulate one or more functions. Given a delegate instance and an appropriate set of arguments, one can invoke all of that delegate instances functions with that set of arguments. event A member that enables a class or a CLI object to provide notifications. field A synonym for what Standard C++ calls a data member. function, abstract A synonym for what Standard C++ calls a pure virtual function.
Definitions
garbage collection The process by which memory allocated from the CLI heap is automatically reclaimed on the CLI heap. gc-lvalue An expression that refers to an entity that might be allocated on the CLI heap. (See also lvalue.) handle A handle is called an object reference in the CLI specification. For any CLI class type T, the declaration T^ h declares a handle h to type T, where the object to which h is capable of pointing resides on the CLI heap. A handle tracks, is rebindable, and can point to a whole object only. (See also type, reference, tracking.) heap, CLI The storage area (accessed by gcnew) that is under the control of the garbage collector of the Virtual Execution System as specified in the CLI. (See also heap, native.) heap, native The storage area (accessed by new) as defined in the C++ Standard (18.4). (See also heap, CLI.) instance An instance of a type. lvalue This has the same meaning as that defined in the C++ Standard (3.10). (See also gc-lvalue.) metadata Data that describes and references the types defined by the Common Type System (CTS). Metadata is stored in a way that is independent of any particular programming language. Thus, metadata provides a common interchange mechanism for use between tools that manipulate programs (such as compilers and debuggers) as well as between these tools and the Virtual Execution System. pinning The process of (temporarily) keeping constant the location of an object that resides on the CLI heap, so that objects address can be taken with that address remaining constant. property A member that defines a named value and the functions that access that value. A property definition defines the accessing contracts on that value. Hence, the property definition specifies the accessing functions that exist and their respective function contracts. rebinding The act of making a handle or pointer refer to the same or another object on the CLI heap. rvalue This has the same meaning as that defined in the C++ Standard (3.10). tracking The act of keeping track of the location of an object that resides on the CLI heap; this is necessary because such objects can move during their lifetime (unlike objects on the native heap, which never move). Tracking is maintained by the Virtual Execution System during garbage collection. Tracking is an inherent property of handles and tracking references. type, boxed See type, value class, boxed. type, class, any Any CLI or native class type. type, class, CLI class A ref class type, a value class type, or an interface class type. type, class, interface A type that declares a set of virtual members that an implementing class shall define. An interface class type is a CLI type. type, class, native An ordinary Standard C++ class (declared using class, struct, or union). type, class, ref A type that can contain fields, function members, and nested types. A ref class type is a CLI type. type, class, value A type that can contain fields, function members, and nested types. Instances of a value class type are values. Since they directly contain their data, no heap allocation is necessary. A value class type is a CLI type. type, class, value, boxed A boxed value class is an instance of a value class on the CLI heap. For a value class V, a boxed value class is always of the form V^. type, class, value, simple The subset of value class types that can be embedded in a native class type and allocated with the new operator.
type, fundamental The arithmetic types as defined by the C++ Standard (3.9.1), and that each have a corresponding value class type provided by the implementation. (These include bool, char, and wchar_t, but exclude enumerations.) type, handle Longhand for handle. type, pointer, native The pointer types as defined by the C++ Standard (8.3.1). (Unlike a handle, a native pointer doesnt track, since objects on the native heap never move.) type, reference, native The reference types as defined by the C++ Standard (8.3.2). type, reference, tracking A reference that can keep track of an object on the CLI heap when that object is moved by the garbage collector. For any type T, the declaration T% r declares a tracking reference r to type T. (See also handle.) unboxing An explicit conversion from type System::Object^ to any value class type, from type System::ValueType^ to any value class type, from V^ (the boxed form of a value class type) to V (the value class type), or from any interface class type handle to any value class type that implements that interface class. (See also boxing.) Virtual Execution System (VES) This system implements and enforces the Common Type System (CTS) model. The VES is responsible for loading and running programs written for the CLI. It provides the services needed to execute CIL and data, using the metadata to connect separately generated modules together at runtime. For example, given an address inside the code for a function, it must be able to locate the metadata describing that function. It must also be able to walk the stack, handle exceptions, and store and retrieve security information. The VES is also known as the Execution Engine.
Notational conventions
5. Notational conventions
Various pieces of text from the C++ Standard appear verbatim in this standard. The C++ Standard is augmented by this C++/CLI Standard, with additions indicated by underlining, and deletions indicated using strike-through. For example: The rules for operators remain largely unchanged from Standard C++; however, the following rule in Standard C++ (13.5/6) is augmented to allow static member functions: A static member or a non-member operator function shall either be a non-static member function or be a non-member function and have at least one parameter whose type is a class, a reference to a class, a handle to a class, an enumeration, a reference to an enumeration, or a handle to an enumeration. Unless otherwise noted, the following names are used as shorthand to refer to a type of their corresponding kind:
I for interface class N for native type R for ref class S for simple value class V for value class
The CLI has its own set of naming conventions, some of which differ from established C++ programming practice. The CLI conventions have been used throughout this Standard; see Annex E. Many source code examples use facilities provided by the CLI namespace System; however, that namespace is not explicitly referenced. Instead, there is an implied using namespace System; at the beginning of each of those examples. Similarly, examples using cout also assume that the iostream header has been included and there is an implied using namespace std; at the beginning of each of those examples. In a number of examples, C++/CLI source code is shown with corresponding metadata. For expository purposes, a specific mapping between primitive C++ types and metadata types is assumed; however, that mapping need not be used by a conforming implementation. For example, type int is shown to map to System::Int32 (which, in metadata, is referred to as int32). In the examples, C++/CLI source code is written in a constant-width font, and the corresponding metadata it written in the same font, but with a greyshaded background. For example,
public ref struct D : B { ref class R { }; }; .class public auto ansi D extends B { .class auto ansi nested public R extends [mscorlib]System.Object { } }
6. Acronyms and abbreviations

This clause is informative The following acronyms and abbreviations are used throughout this Standard: IEC the International Electrotechnical Commission IEEE the Institute of Electrical and Electronics Engineers ISO the International Organization for Standardization The following terms are defined in the CLI standard. BCL Base Class Library, which provides types to represent the built-in data types of the CLI, simple file access, custom attributes, security attributes, string manipulation, formatting, streams, and collections. CIL Common Intermediate Language CLI Common Language Infrastructure CLS Common Language Specification CTS Common Type System VES Virtual Execution System
End of informative text
General description
7. General description
This Standard is intended for implementers, academics, and application programmers. As such, it contains a considerable amount of explanatory material that, strictly speaking, is not necessary in a formal language specification. This standard is divided into the following subdivisions: 1. Front matter (clauses 17); 2. Language overview (clause 8); 3. The language syntax, constraints, semantics, and library (clauses 932); 4. Metadata generation (clauses 3334); 5. Annexes Examples are provided to illustrate possible forms of the constructions described. References are used to refer to related clauses. Notes are provided to give advice or guidance to implementers or programmers. Rational provides explantory material as to why something is or is not in this standard. Annexes provide additional information and summarize the information contained in this Standard. Clauses 15, 7, and 934 form a normative part of this standard; Introduction, clauses 6 and 8, annexes, notes, examples, rationale, and the index, are informative. Except for whole clauses or annexes that are identified as being informative, informative text that is contained within normative text is indicated in the following ways: 1. [Example: code fragment, possibly with some narrative end example] 2. [Note: narrative end note] 3. [Rationale: narrative end rationale]
8. Language overview
This clause is informative. This specification is a superset of Standard C++. This clause describes the essential features of this specification. While later clauses describe rules and exceptions in detail, this clause strives for clarity and brevity at the expense of completeness. The intent is to provide the reader with an introduction to the language that will facilitate the writing of early programs and the reading of later clauses.
8.1 Getting started

The canonical hello, world program can be written as follows:
int main() { System::Console::WriteLine("hello, world"); }
The source code for a C++/CLI program is typically stored in one or more text files with a file extension of .cpp, as in hello.cpp. Using a command-line compiler (called cl, for example), such a program can be compiled with a command line like
cl hello.cpp
which produces an application named hello.exe. The output produced by this application when it is run is:
hello, world
where the WriteLine function automatically adds a terminating newline. The CLI library is organized into a number of namespaces, the most commonly used being System. That namespace contains a ref class called Console, which provides a family of functions for performing console I/O. One of these functions is WriteLine, which when given a string, writes that string plus a trailing newline to the console. (Examples from this point on assume that the namespace System has been the subject of a using-declaration.)
8.2 Types
Value class types differ from handle types in that variables of value class types directly contain their data, whereas variables of the handle types store handles to objects. With handle types, it is possible for two variables to reference the same CLI object, and thus possible for operations on one variable to affect the object referenced by the other variable. With value classes, the variables each have their own copy of the data, and it is not possible for operations on one to affect the other. The example
ref class Class1 { public: int Value; Class1() { Value = 0; } }; int main() { int val1 = 0; int val2 = val1; val2 = 123;
10
Language overview Class1^ ref1 = gcnew Class1; Class1^ ref2 = ref1; ref2->Value = 123; Console::WriteLine("Values: {0}, {1}", val1, val2); Console::WriteLine("Refs: {0}, {1}", ref1->Value, ref2->Value);
shows this difference. The output produced is

Values: 0, 123 Refs: 123, 123
The assignment to the local variable val1 does not affect the local variable val2 because both local variables have primitive types (which are also value class types), and each local variable of a primitive type has its own storage. In contrast, the assignment ref2->Value = 123; affects the CLI object that both ref1 and ref2 reference. The lines
Console::WriteLine("Values: {0}, {1}", val1, val2); Console::WriteLine("Refs: {0}, {1}", ref1->Value, ref2->Value);
deserve further comment, as they demonstrate some of the string formatting behavior of Console::WriteLine, which, in fact, takes a variable number of arguments. The first argument is a string, which can contain numbered placeholders like {0} and {1}. Each placeholder refers to a trailing argument with {0} referring to the second argument, {1} referring to the third argument, and so on. Before the output is sent to the console, each placeholder is replaced with the formatted value of its corresponding argument. Developers can define new value class types through enum and value class definitions. The following code shows an example of each kind of type definition. Later clauses describe type definitions in detail.
public enum class Color { Red, Blue, Green }; public value struct Point { int x, y; }; public interface class IBase { void F(); }; public interface class IDerived : IBase { void G(); }; public ref class A { protected: virtual void H() { Console::WriteLine("A.H"); } }; public ref class B : A, IDerived { public: virtual void F() { Console::WriteLine("B::F, implementation of IBase::F"); } virtual void G() { Console::WriteLine("B::G, implementation of IDerived::G"); }
11
C++/CLI Language Specification protected: virtual void H() override { Console::WriteLine("B::H, override of A::H"); } }; public delegate void MyDelegate();
Types like Color, Point, and IBase above, which are not defined inside other types (i.e., they are toplevel types), can have a type visibility specifier of either public or private. The use of public in this context indicates that the type is visible outside its parent assembly. Conversely, private indicates that the type is not visible outside its parent assembly. The default visibility for a top-level type is private. 8.2.1 Fundamental types and the CLI Each of the fundamental types has a corresponding value class type provided by the implementation; the correspondence is implementation-defined. For example, one implementation might specify that int has the corresponding type System::Int32, while another specifies it has the corresponding type System::Int64. Using the keyword name has the usual Standard C++ meaning, while the corresponding CLI name indicates a particular CLI platform type. [Example: int specifies the implementation-defined natural integer type, whereas Int32 specifies an integer type that is exactly 32 bits on any CLI platform. end example] The table below lists the fundamental types and their corresponding CLI-provided type in one implementation. For consistency, the examples in this Standard use the values in this table without continually re-stating implementation-defined.
Type
bool char signed char unsigned char short unsigned short int unsigned int long unsigned long long long int
Description Boolean type; a bool value is either true or false 8-bit signed/unsigned integral type 8-bit signed integral type 8-bit unsigned integral type 16-bit signed integral type 16-bit unsigned integral type 32-bit signed integral type 32-bit unsigned integral type 32-bit signed integral type 32-bit unsigned integral type
Corresponding CLI Value class type

System::Boolean System::SByte
(with modopt for IsSignUnspecifiedByte)

System::SByte System::Byte System::Int16 System::UInt16 System::Int32 System::UInt32 System::Int32
(with modopt IsLong)

System::UInt32
(with modopt IsLong)

System::Int64 System::Uint64 System::Single System::Double System::Double (with
64-bit signed integral type unsigned long long int 64-bit unsigned integral type float Single-precision floating point type double Double-precision floating point type
long double wchar_t
Extra-precision floating point type A 16-bit Unicode code unit
modopt IsLong)
System::Char
Although they are not fundamental types, three other types provided in the CLI library are worth mentioning. They are: 12
Language overview System::Object, which is the ultimate base type of all value and handle types System::String, a sequence of Unicode code units System::Decimal, a precise decimal type with at least 28 significant digits
C++/CLI has no keyword type names that can correspond to these. 8.2.2 Conversions A number of new kinds of conversion have been defined. These include handle and parameter array conversion, among others. 8.2.3 CLI array types A CLI array differs from a native array (C++ Standard 8.3.4) in that the former is allocated on the CLI heap, and can have a rank other than one. The rank determines the number of indices associated with each array element. The rank of a CLI array is also referred to as the dimensions of the CLI array. A CLI array with a rank of one is called a single-dimensional CLI array, and a CLI array with a rank greater than one is called a multi-dimensional CLI array. Throughout this Standard, the term CLI array is used to mean an array in the CLI. A C++-style array is referred to as a native array or, more simply, array, whenever the distinction is needed. A CLI array type is declared using a built-in pseudo-template ref class having the following declaration:
namespace cli { template<typename T, int rank = 1> ref class array : System::Array { }; }
An example of using this pseudo-template is:

int main() { array<int>^ arr1D = gcnew array<int>(4) {10, 42, 30, 12}; Console::Write("The {0} elements are:", arr1D->Length); for each (int i in arr1D) { Console::Write("{0,3}", i); } Console::WriteLine(); array<int, 3>^ arr3D = gcnew array<int, 3>(10, 20, 30); }
The output produced is:

The 4 elements are: 10 42 30 12
Handle arr1D can be made to refer to any one-dimensional array of int. It currently refers to one containing four int elements. The read-only property Array::Length contains the element count. Handle arr3D can be made to refer to any three-dimensional array of int. It currently refers to one of size 10x20x30, all of whose elements have the default value for int; that is, zero. 8.2.4 Type system unification C++/CLI provides a unified type system. All value and handle types derive from the type System::Object. It is possible to call instance functions on any value, even values of fundamental types such as int. The example
int main() { Console::WriteLine((3).ToString()); }
calls the instance function ToString from type System::Int32 on an integer literal, resulting in the string 3 being output. (Note that the seemingly redundant grouping parentheses around the literal 3, are not redundant; they are needed to get the tokens 3 and . instead of 3..) The example 13
C++/CLI Language Specification int main() int i = Object^ int j = } { 123; o = i; safe_cast<int>(o);
// boxing // unboxing
is more interesting. An int value can be converted to System::Object^ and back again to int. This example shows both boxing and unboxing. When a variable of a value class type needs to be converted to a handle type, a System::Object box is allocated to hold the value, and the value is copied into the box. Unboxing is just the opposite. When a System::Object box handle is cast back to its original value class type, the value is copied out of the box and into the appropriate storage location. This type system unification provides value classes with the benefits of object-ness without introducing unnecessary overhead. For programs that dont need int values to act like CLI objects, int values are simply 32-bit values. For programs that need int values to behave like CLI objects, this capability is available on demand. This ability to treat instances of value class types as CLI objects bridges the gap between value classes and ref classes that exists in most languages. For example, a Stack class can provide Push and Pop functions that take and return Object^ values.
public ref class Stack { public: Object^ Pop() { } void Push(Object^ o) { } };
Because C++/CLI has a unified type system, the Stack class can be used with elements of any type, including value class types like int. 8.2.5 Pointers, handles, and null Standard C++ supports pointer types and null pointer constants. C++/CLI adds handle types and null values. To help integrate handles, and to have a universal null, C++/CLI defines the keyword nullptr. This keyword represents a literal having the null type. nullptr is referred to as the null value constant. (No instances of the null type can ever be created, and the only way to obtain a null value constant is via this keyword.) The definition of null pointer constant (which Standard C++ requires to be a compile-time expression that evaluates to zero) is augmented to include nullptr. The null value constant can be implicitly converted to any pointer or handle type, in which case it becomes a null pointer value or null value, respectively. This allows nullptr to be used in relational, equality, conditional, and assignment expressions, among others.
Object^ obj1 = nullptr; String^ str1 = nullptr; if (obj1 == 0); if (obj1 == 0L); if (obj1 == nullptr); char* pc1 = nullptr; if (pc1 == 0); if (pc1 == 0L); if (pc1 == nullptr); int n1 = 0; n1 = nullptr; if (n1 == 0); if (n1 == 0L); if (n1 == nullptr); if (nullptr); if (nullptr == 0); if (nullptr == 0L); nullptr = 0; nullptr + 2; // // // // // // // // // handle obj1 has the null value handle str1 has the null value false (0 is boxed, the two handles differ) false true pc1 is the null pointer value true as zero is a null pointer value true true as nullptr is a null pointer constant
// error, no implicit conversion to int // true, performs integer comparison // // error, no implicit conversion to int // // // // // error error, no implicit conversion to int error, nullptr is not an lvalue error, nullptr cant take part in arithmetic
14
Language overview Object^ obj2 = 0; Object^ obj3 = 0L; String^ str2 = 0; String^ str3 = 0L; char* pc2 = 0; char* pc3 = 0L; // // // // // // obj2 is a handle to a boxed zero obj3 error, no conversion from int to String^ pc2 is the null pointer value pc3 // obj4 is the null value // error, no composite type
Object^ obj4 = expr ? nullptr : nullptr; Object^ obj5 = expr ? 0 : nullptr; char* pc4 = expr ? nullptr : nullptr; char* pc5 = expr ? 0 : nullptr; int n2 = expr ? nullptr : nullptr; int int n3 = expr ? 0 : nullptr; sizeof(nullptr); typeid(nullptr); throw nullptr; void f(Object^); void f(String^); void f(char*); void f(int); f(nullptr); f(0); void g(Object^, Object^); void g(Object^, char*); void g(Object^, int); g(nullptr, nullptr); g(nullptr, 0); g(0, nullptr); void h(Object^, int); void h(char*, Object^); h(nullptr, nullptr); h(nullptr, 2);
// pc4 is the null pointer value // error, no composite type
// error, no implicit conversion to // error, no composite type
// error, the null type has no size, per se // error // error // // // // // // // // // // // // 1 2 3 4 error, ambiguous (1, 2, 3 possible) calls f(int) 1 2 3 error, ambiguous (1, 2 possible) calls g(Object^, int) error, ambiguous (1, 2 possible)
// calls h(char*, Object^); // calls h(Object^, int);
template<typename T> void k(T t); k(0); // specializes k, T = int k(nullptr); // error, cant instantiate null type k((Object^)nullptr); // specializes k, T = Object^ k<int*>(nullptr); // specializes k, T = int*
Since objects allocated on the native heap do not move, pointers and references to such objects need not track an objects location. However, objects on the CLI heap can move, so they require tracking. As such, native pointers and references are not sufficient for dealing with them. To track objects on the CLI heap, C++/CLI defines handles (using the punctuator ^) and tracking references (using the punctuator %).
N* pn = new N; N& rn = *pn; R^ hr = gcnew R; R% rr = *hr; // allocate on native heap // bind ordinary reference to native object // allocate on CLI heap // bind tracking reference to gc-lvalue
In general, the punctuator % is to ^ as the punctuator & is to *. Just as Standard C++ has a unary & operator, C++/CLI provides a unary % operator. While &t yields a T* or an interior_ptr<T> (see below), %t yields a T^. Rvalues and lvalues continue to have the same meaning as with Standard C++, with the following rules applying: An entity declared with type T*, a native pointer to T, points to an lvalue. Applying unary * to an entity declared with type T*, dereferencing a T*, yields an lvalue. An entity declared with type T&, a native reference to T, is an lvalue. The expression &lvalue yields a T*.
15
The expression %lvalue yields a T^.
A gc-lvalue is an expression that refers to an object that might be on the CLI heap, or to a value member contained within such an object. The following rules apply to gc-lvalues: Standard conversions exist from cv-qualified lvalue of type T to cv-qualified gc-lvalue of type T, and from cv-qualified gc-lvalue of type T to cv-qualified rvalue of type T. An entity declared with type T^, a handle to T, points to a gc-lvalue. Applying unary * to an entity declared with type T^, dereferencing a T^, yields a gc-lvalue. An entity declared with type T%, a tracking reference to T, is a gc-lvalue. The expression &gc-lvalue yields an interior_ptr<T> (see below). The expression %gc-lvalue yields a T^.
The garbage collector is permitted to move objects that reside on the CLI heap. In order for a pointer to refer correctly to such an object, the runtime needs to update that pointer to the objects new location. An interior pointer (which is defined using interior_ptr) is a pointer that is updated in this manner.
8.3 Parameters
A parameter array is a type-safe alternative to parameter lists that end with an ellipsis. A parameter array is declared with a leading ... punctuator, followed by a CLI array type. There can be only one parameter array for a given function, and it shall always be the last parameter specified. The type of a parameter array is always a single-dimensional CLI array type. A caller can either pass a single argument of this CLI array type, or any number of arguments of the element type of this CLI array type. For instance, the example
void F(... array<int>^ args) { Console::WriteLine("# of arguments: {0}", args->Length); for (int i = 0; i < args->Length; i++) Console::WriteLine("\targs[{0}] = {1}", i, args[i]); } int main() { F(); F(1); F(1, 2); F(1, 2, 3); F(gcnew array<int> {1, 2, 3, 4}); }
shows a function F that takes a variable number of int arguments, and several invocations of this function. The output is:
# of arguments: # of arguments: args[0] = 1 # of arguments: args[0] = 1 args[1] = 2 # of arguments: args[0] = 1 args[1] = 2 args[2] = 3 # of arguments: args[0] = 1 args[1] = 2 args[2] = 3 args[3] = 4 0 1 2 3
By declaring the parameter array to be a CLI array of type System::Object^, the parameters can be heterogeneous; for example:
16
Language overview void G(... array<Object^>^ args) { } G(10, Hello, 1.23, X); // arguments 1, 3, and 4 are boxed
A number of examples presented in this Standard use the WriteLine function of the Console class. The argument substitution behavior of this function, as exhibited in the example
int a = 1, b = 2; Console::WriteLine("a = {0}, b = {1}", a, b);
is accomplished using a parameter array. The Console class provides several overloaded versions of the WriteLine function to handle the common cases in which a small number of arguments are passed, and one general-purpose version that uses a parameter array, as follows:
namespace System { public ref class Object { }; public ref class String { }; public ref class Console { public: static void WriteLine(String^ static void WriteLine(String^ static void WriteLine(String^ static void WriteLine(String^ { } static void WriteLine(String^ }; }
s) s, s, s,
{ } Object^ a) { } Object^ a, Object^ b) { } Object^ a, Object^ b, Object^ c)
s, ... array<Object^>^ args) { }
The CLI library specification shows library functions using C# syntax, in which case, the C# keyword params indicates a parameter array. For example, the declaration of the final WriteLine function above is written in C#, as follows:
public static void WriteLine(string s, params object[] args)
8.4 Automatic memory management

The example
public ref class Stack { public: Stack() { first = nullptr; } property bool IsEmpty { bool get() { return (first == nullptr); } } Object^ Pop() { if (first == nullptr) throw gcnew Exception("Can't Pop from an empty Stack."); else { Object^ temp = first->Value; first = first->Next; return temp; } } void Push(Object^ o) { first = gcnew Node(o, first); }
17
C++/CLI Language Specification ref struct Node { Node^ Next; Object^ Value; Node(Object^ value) { Next = nullptr; Value = value; } Node(Object^ value, Node^ next) { Next = next; Value = value; } }; private: Node^ first; };
shows a Stack class implemented as a linked list of Node instances. Node instances are created in the Push function and are garbage-collected when no longer needed. A Node instance becomes eligible for garbage collection when it is no longer possible for any code to access it. For instance, when an item is removed from the Stack, the associated Node instance becomes eligible for garbage collection. The example
int main() { Stack^ s = gcnew Stack; for (int i = 0; i < 10; i++) s->Push(i); s = nullptr; }
shows code that uses the Stack class. A Stack is created and initialized with 10 elements, and then the handle to it is assigned the value nullptr. Once the variable s is assigned the null value, the Stack and the associated 10 Node instances become eligible for garbage collection. The garbage collector is permitted to clean up immediately, but is not required to do so. The garbage collector underlying C++/CLI can work by moving objects on the CLI heap around in memory, but this motion is invisible to most C++/CLI developers. For developers who are generally content with automatic memory management, but sometimes need fine-grained control or that extra bit of performance, C++/CLI provides the ability to pin objects on the CLI heap, to prevent temporarily the garbage collector from moving them. For example,
void f(int* p) { *p = 100; } int main() { array<int>^ arr = gcnew array<int>(100); pin_ptr<int> pinp = &arr[0]; // pin arrs location f(pinp); // change arr[0]s value }
8.5 Expressions
C++/CLI augments the C++ Standard with respect to operators. For example: The addition of delegates requires the use of the function-call operator to invoke the functions encapsulated by a delegate. A new use of typeid has been added. For example, Int32::typeid results in a handle to a CLI object of type System::Type that describes the CLI type Int32. The cast operators are augmented to accommodate handle types. The safe_cast operator has been added. The operator gcnew has been added. This allocates memory from the CLI heap. The binary + and operators are augmented to accommodate delegate addition and removal, respectively.
18
Language overview
Simple assignment is augmented to accommodate properties and events as the left operand. Compound assignment operators are synthesized from the corresponding binary operator (19.7.4).
8.6 Statements
A new statement, for each, has been added. This statement enumerates the elements of a collection, executing a block for each element of that collection. For example:
void display(array<int>^ args) { for each (int i in args) Console::WriteLine(i); }
A type is said to be a collection type if it implements the System::Collections::IEnumerable interface or implements some collection pattern by meeting a number of criteria.
8.7 Delegates
Delegates enable scenarios that Standard C++ programmers typically address with function adapters from the Standard C++ Library. A delegate definition implicitly defines a class that is derived from the class System::Delegate. A delegate instance encapsulates one or more functions in an invocation list, each member of which is referred to as a callable entity. For instance functions, a callable entity is an instance and a member function on that instance. For static functions or global- or namespace-scope functions, a callable entity is just a member, global-, or namespace-scope function, respectively. Given a delegate instance and an appropriate set of arguments, one can invoke all of that delegate instances callable entities with that set of arguments. Consider the following example:
delegate void MyFunction(int value); // define a delegate type public ref struct A { static void F(int i) { Console::WriteLine("F:{0}", i); } }; public ref struct B { void G(int i) { Console::WriteLine("G:{0}", i); } };
The static function A::F and the instance function B::G both have the same parameter types and return type as MyFunction, so they can be encapsulated by a delegate of that type. Note that even though both functions are public, their accessibility is irrelevant when considering their compatibility with MyFunction. Such functions can also be defined in the same or different classes, as the programmer sees fit.
int main() { MyFunction^ d; d = gcnew MyFunction(&A::F); d(10); // create a delegate reference // invocation list is A::F
B^ b = gcnew B; d += gcnew MyFunction(b, &B::G); // invocation list is A::F B::G d(20); d += gcnew MyFunction(&A::F); d(30); // invocation list is A::F B::G A::F
d -= gcnew MyFunction(b, &B::G); // invocation list is A::F A::F d(40);
19
C++/CLI Language Specification F:10 F:20 G:20 F:30 G:30 F:30 F:40 F:40
The constructor for a delegate needs two arguments when it is bound to a non-static member function: the first is a handle to an instance of a ref class, and the second designates the non-static member function within that ref classs type, using the syntax of a pointer to member. The constructor for a delegate needs only one argument when it is bound to a static function, or a global- or namespace-scope function; the argument designates that function, using the syntax of a pointer to member or pointer to function, as appropriate. The invocation lists of two compatible delegates can be combined via the += operator, as shown. In addition, callable entities can be removed from an invocation list via the -= operator, as shown. However, an invocation list cannot be changed once it has been created. Specifically, these operators create new invocation lists. Once a delegate instance has been initialized, it is possible to indirectly call the functions it encapsulates just as if they were called directly (in the same order in which they were added to the delegate's invocation list), except the delegate instances name is used instead. The value (if any) returned by the delegate call is that returned by the final function in that delegate's invocation list. If a delegate instance is null and an attempt is made to call the encapsulated functions, an exception of type NullReferenceException results.
8.8 Native and ref classes

8.8.1 Literal fields A literal field is a field that represents a compile-time constant rvalue. The value of a literal field is permitted to depend on the value of other literal fields within the same program as long as they have been previously defined. The example
ref class X { literal int A = 1; public: literal int B = A + 1; }; ref class Y { public: literal double C = X::B * 5.6; };
shows two classes that, between them, define three literal fields, two of which are public while the other is private. Even though literal fields are accessed like static members, a literal field is not static and its definition neither requires nor allows the keyword static. Literal fields can be accessed through the class, as in
int main() { cout << "B = " << X::B << "\n"; cout << "C = " << Y::C << "\n"; }
which produces the following output:

B = 2 C = 11.2
Literal fields are only permitted in ref, value, and interface classes.
20
Language overview
8.8.2 Initonly fields The initonly identifier declares a field that is an lvalue only within the ctor-initializer and the body of an instance constructor, or within a static constructor, and thereafter is an rvalue. Such a field is called an initonly field. For example:
public ref class Data { initonly static double coefficient1; initonly static double coefficient2; static Data() { // read in the value of the coefficients from some source coefficient1 = ; // ok coefficient2 = ; // ok } public: static void F() { coefficient1 = ; // error coefficient2 = ; // error } };
Assignments to an initonly field can only occur as part of its definition, or in an instance constructor or static constructor in the same class. (A static initonly field can be assigned to in a static constructor, and a nonstatic initonly field can be assigned to in an instance constructor.) Initonly fields are only permitted in ref and value classes. 8.8.3 Functions Member functions in CLI class types are defined and used just as in Standard C++. However, C++/CLI does have some differences in this regard. For example: The const and volatile qualifiers are not permitted on instance member functions. The function modifier override and override specifiers provide the ability to indicate explicit overriding and named overriding (8.8.10.1). Marking a virtual member function as sealed prohibits that function from being overridden in a derived class. The function modifier abstract provides an alternate way to declare an abstract function. The function modifier new allows the function to which it applies to hide the base class function of the same name, parameter-type-list, and cv-qualification. Such a hiding function does not override any base class function, even if the hiding function is declared virtual. Type-safe variable-length argument lists are supported via parameter arrays.
8.8.4 Properties A property is a member that behaves as if it were a field. There are two kinds of properties: scalar and indexed. A scalar property enables field-like access to a class or CLI object. Examples of scalar properties include the length of a string, the size of a font, the caption of a window, and the name of a customer. An indexed property enables array-like access to a CLI object. An example of an index property is a bit-array class. Properties are an evolutionary extension of fieldsboth are named members with associated types, and the syntax for accessing scalar fields and scalar properties is the same, as is that for accessing CLI arrays and indexed properties. However, unlike fields, properties do not denote storage locations. Instead, properties have accessor functions that specify the statements to be executed when their values are read or written. Properties are defined with property definitions. The first part of a property definition looks quite similar to a field definition. The second part includes a get accessor function and/or a set accessor function. Properties that can be both read and written include both get and set accessor functions. In the example below, the Point class defines two read-write properties, X and Y. 21
C++/CLI Language Specification public value class Point { int Xor; int Yor; public: property int X { int get() void set(int value) } property int Y { int get() void set(int value) } Point(int x, int y) { Move(x, y); } void Move(int x, int y) { X = x; Y = y; } void Translate(int x, int y) { X += x; Y += y; } }; // absolute move
{ return Xor; } { Xor = value; }
{ return Yor; } { Yor = value; }
// relative move
The get accessor function is called when the propertys value is read; the set accessor function is called when the propertys value is written. The definition of properties is relatively straightforward, but the real value of properties is seen when they are used. For example, the X and Y properties can be read and written as though they were fields. In the example above, the properties are used to implement data hiding within the class itself. The following application code (directly and indirectly) also uses these properties:
Point p1; p1.X = 10; p1.Y = 5; p1.Move(5, 7); Point p2(9, 1); p2.Translate(-4, 12); // // // // // // set to (0,0) set to (10,0) set to (10,5) move to (5,7) set to (9,1) move 4 left and 12 up, to (5,13)
For a trivial property declaration such as

property String^ Name;
the compiler automatically provides the default implementations of the accessor functions. A default-indexed property allows array-like access directly on an instance. [Note: Other languages refer to default-indexed properties as indexers. end note] As an example, consider a Stack class. The designer of this class might want to expose array-like access so that it is possible to inspect or alter the items on the stack without performing unnecessary Push and Pop operations. That is, class Stack is implemented as a linked list, but it also provides the convenience of array access. Default-indexed property definitions are similar to property definitions, with the main differences being that default-indexed properties are nameless and that they include indexing parameters. The indexing parameters are provided between square brackets. The example
22
Language overview public ref class Stack { public: ref struct Node { Node^ Next; Object^ Value; Node(Object^ value) : Next(nullptr), Value(value) {} Node(Object^ value, Node^ next) { Next = next; Value = value; } }; private: Node^ first; Node^ GetNode(int index) { Node^ temp = first; while (index > 0) { temp = temp->Next; index--; } return temp; } bool ValidIndex(int index) { } public: property Object^ default[int] { // default-indexed property Object^ get(int index) { if (!ValidIndex(index)) throw gcnew Exception("Index out of range."); else return GetNode(index)->Value; } void set(int index, Object^ value) { if (!ValidIndex(index)) throw gcnew Exception("Index out of range."); else GetNode(index)->Value = value; } } Object^ Pop() { } void Push(Object^ o) { } }; int main() { Stack^ s = gcnew Stack; s->Push(1); s->Push(2); s->Push(3); s[0] = 33; s[1] = 22; s[2] = 11; } // The top item now refers to 33 instead of 3 // The middle item now refers to 22 instead of 2 // The bottom item now refers to 11 instead of 1
shows a default-indexed property for the Stack class. [Note: A more efficient implementation of Stack would make use of generics. end note] 8.8.5 Events An event is a member that enables a class or CLI object to provide notifications. A class defines an event by providing an event declaration (which resembles a field declaration, though with an added event identifier) and an optional set of event accessor functions. The type of this declaration must be a handle to a delegate type (8.7). In the example
23
C++/CLI Language Specification public delegate void EventHandler(Object^ sender, EventArgs^ e); public ref class Button { public: event EventHandler^ Click; };
the Button class defines a Click event of type EventHandler. The Click member is only used on the left-hand side of the += and = operators, or with the function-call operator (in which case, all the functions in the event's delegate list are called). The += operator adds a handler for the event, and the -= operator removes a handler for the event. The example
public ref class Form1 { Button^ Button1; void Button1_Click(Object^ sender, EventArgs^ e) { Console::WriteLine("Button1 was clicked!"); } public: Form1() { Button1 = gcnew Button; // Add Button1_Click as an event handler for Button1s Click event Button1->Click += gcnew EventHandler(this, &Form1::Button1_Click); } void Disconnect() { Button1->Click -= gcnew EventHandler(this, &Form1::Button1_Click); } };
shows a class, Form1, that adds Button1_Click as an event handler for Button1s Click event. In the Disconnect function, that event handler is removed. Programmers who wants more control can get it by explicitly providing add and remove accessor functions. For example, the Button class could be rewritten as follows:
public ref class Button { EventHandler^ handler; public: event EventHandler^ Click { void add(EventHandler^ e) { handler += e; } void remove(EventHandler^ e) { handler -= e; } } };
This change has no effect on client code, but it allows the Button class more implementation flexibility. For example, the event handler for Click need not be represented by a field. For a trivial event declaration such as
event EventHandler^ Click;
the compiler automatically provides the default implementations of the accessor functions. 8.8.6 Static operators In addition to Standard C++ operator overloading, C++/CLI provides the ability to define operators that are static and/or take parameters of ^ type. The following example shows part of an integer vector class:
public ref class IntVector { array<int>^ values; public: property int Length { // property int get() { return values->Length; } }
24
Language overview property int default[int] { // default-indexed property int get(int index) { return values[index]; } void set(int index, int value) { values[index] = value; } } IntVector(int length); IntVector(int length, int value); // unary (negation) static IntVector^ operator-(IntVector^ iv) { IntVector^ temp = gcnew IntVector(iv->Length); for (int i = 0; i < iv->Length; ++i) { temp[i] = -iv[i]; } return temp; } static IntVector^ operator+(IntVector^ iv, int val) { IntVector^ temp = gcnew IntVector(iv->Length); for (int i = 0; i < iv->Length; ++i) { temp[i] = iv[i] + val; } return temp; } static IntVector^ operator+(int val, IntVector^ iv) { return iv + val; } }; int main() { IntVector^ iv1 = gcnew IntVector(4); IntVector^ iv2 = gcnew IntVector(7, 2); iv1 = -2 + iv2 + 5; iv2 = -iv1; } // 4 elements with value 0 // 7 elements with value 2
8.8.7 Instance constructors Unlike Standard C++, C++/CLI supports static constructors (8.8.9). As such, this specification refers to constructors as defined by the C++ Standard as being instance constructors. 8.8.8 Destructors and finalizers In Standard C++, cleanup code has traditionally been encapsulated by the destructor. While this approach provides a convenient and powerful way to abstract resources, resource leaks can occur if the destructor is not called. By having a garbage collector, C++/CLI provides a mechanism to write cleanup code that can be executed instead when an object is no longer referenced. As a result, a ref class can have two special member functions responsible for cleaning up resources held by an instance of that type: a destructor and a finalizer. Destructor: The destructor provides deterministic cleanup and ends the lifetime of the object. As in Standard C++, the destructor cleans up the bases and members of an object in the reverse order of the completion of their constructor. Within each ref class, in order, the destructor executes the user-written code, calls the destructors for each embedded member of the class, and calls the destructor for each base class. The main advantage of a destructor is that it is called at deterministic points in the program, which has the advantage of freeing resources earlier than if one waited for garbage collection. Finalizer: The finalizer provides non-deterministic cleanup. A finalizer is a last-chance function that is executed during garbage collection, typically on an object whose destructor was not executed. Finalizers are particularly useful to ensure resources that are represented by data members having value types (such as native pointers referring to allocation from the native heap) are cleaned up even if the destructor is not executed. The finalizer executes sometime
25
after the garbage collector determines there are no active references to the object. (There can be a performance penalty for having a finalizer.) A ref class whose instances own resources should always have a destructor. A class that has a finalizer should always have a destructor as well, to enable deterministic cleanup and early resource release. However, a class that has a destructor need not necessarily have a finalizer.
ref struct R { ~R() { } }; // destructor, but no finalizer
A ref class whose instances have resources represented by value types (such as a pointer) should have a finalizer. (There may be a performance penalty for introducing a finalizer to a class that does not already have some finalizable ancestor class. As such, a well-designed class hierarchy will limit resources represented by value types to the leaves of the class hierarchy.) A ref class whose instances have no value types representing resources can still have a destructor, but should not have a finalizer.
ref struct R { ~R() { } !R() { } }; // destructor // finalizer
C++/CLI implements the destructor and finalizer semantics in any ref class T by using the CLI dispose pattern, which makes use of five functions (Dispose(), Dispose(bool), Finalize(), __identifier(~T)(), and __identifier(!T)()), all of whose definitions are generated by the compiler, as required. These cleanup mechanisms are hidden from the C++/CLI programmer. In C++/CLI, the proper way to do cleanup is to place all of the cleanup code in the destructor and finalizer, as follows: The finalizer should clean up any resources that are represented by value types. The destructor should do the maximal cleanup possible. To facilitate this, the programmer should call the finalizer from the destructor and write any other cleanup code in the destructor. A destructor can safely access the state of ref classes with references from the object, whereas a finalizer cannot.
For ref classes, both the finalizer and destructor must be written so they can be executed multiple times and on objects that have not been fully constructed. 8.8.9 Static constructors A static constructor is a ref or value class static member function that implements the actions required to initialize the static members of a class, rather than the instance members of that class. Static constructors cannot have parameters, they must be private, and they cannot be called explicitly. The static constructor for a class is called automatically by the runtime. [Note: A static constructor is required to be private to prevent the static constructor from being invoked more than once. end note] The example
public ref class Data { private: initonly static double coefficient1; initonly static double coefficient2; static Data() { // read in the value of the coefficients from some source coefficient1 = ; coefficient2 = ; } public: };
shows a Data class with a static constructor that initializes two initonly static fields.
26
Language overview
8.8.10 Inheritance When using ref classes, C++/CLI supports single inheritance of ref classes only. However, multiple inheritance of interfaces is permitted. 8.8.10.1 Function overriding In Standard C++, given a derived class with a function having the same name, parameter-type-list, and cvqualification as a virtual function in a base class, the derived class function always overrides the one in the base class, even if the derived class function is not declared virtual.
struct B { virtual void f(); virtual void g(); }; struct D : B { virtual void f(); void g(); };
// D::f overrides B::f // D::g overrides B::g
We refer to this as implicit overriding. (As the virtual specifier on D::f is optional, the presence of virtual there really isnt an indication of explicit overriding.) Since implicit overriding gets in the way of versioning (8.13), implicit overriding must be diagnosed by a C++/CLI compiler. C++/CLI supports two virtual function-overriding features not available in Standard C++. These features are available in ref class types. They are explicit overriding and named overriding. Explicit overriding: In C++/CLI, it is possible to state that 1. A derived class function explicitly overrides a base class virtual function having the same name, parameter-type-list, and cv-qualification, by using the function modifier override, with the program being ill-formed if no such base class virtual function exists; and 2. A derived class function explicitly does not override a base class virtual function having the same name, parameter-type-list, and cv-qualification, by using the function modifier new.
ref struct B { virtual void F() {} virtual void G() {} }; ref struct D : B { virtual void F() override {} // D::F overrides B::F virtual void G() new {} // D::G doesnt override B::G, it hides it }; D::F must be virtual, and must be marked as such. On the other hand, D::G doesn't have to be virtual, and if it isn't, it shouldn't be marked as such.
Named overriding: Instead of using the override modifier, we can achieve the same thing by using an override-specifier, which involves naming the function we are overriding. This approach also allows us to override a function having a different name, provided the parameter lists are the same.
ref struct B { virtual void F() {} }; interface struct I { virtual void G(); }; ref struct D : B, I { virtual void X() = B::F, I::G {} // D::X overrides B::F and I::G };
The use of virtual in all function declarations having an override-specifier is mandatory. Explicit and named overriding can be combined, as follows:
27
C++/CLI Language Specification ref struct B { virtual void F() {} virtual void G() {} }; ref struct D : B { virtual void F() override = B::G {} };
A function can only be overridden once in any given class. Therefore, if an implicit or explicit override does the same thing as a named override, the program is ill-formed.
ref struct B { virtual void F() {} virtual void G() {} }; ref struct virtual virtual virtual }; D : B { void F() override = B::F {} // Error: B::F is overridden twice void G() override {} // B::G is overridden void H() = B::G {} // Error: B::G is overridden twice
[Note: If a base class is dependent on a template type parameter, a named override of a virtual function from that base class does not happen until the point of instantiation. In the following
template<typename T> ref struct R : T { virtual void F() = T::G {} }; T::G is a dependent name. end note]
8.9 Value classes

Value classes are similar to ref classes in that the former represent data structures that can contain fields and function members. However, unlike ref classes, value classes do not require heap allocation. A variable of a value class directly contains the data of the value class, whereas a variable of a ref class contains a handle to the data. Value classes are particularly useful for small data structures that have value semantics. Complex numbers, points in a coordinate system, or key-value pairs in a dictionary are all good examples of value classes. Key to these data structures is that they have few fields, they do not require the use of inheritance or referential identity, and they can be conveniently implemented using value semantics where assignment copies the value instead of the reference. The primitive typessuch as int, double, and boolare, in fact, all value class types. It is possible to use value class types and operator overloading to implement new primitive types.
value struct Point { int x, y; Point(int x, int y) { this->x = x; this->y = y; } };
8.10 Interfaces
An interface defines a contract. A class that implements an interface must adhere to its contract by implementing all of the functions, properties, and events that interface declares. The example
delegate void EventHandler(Object^ sender, EventArgs^ e);
28
Language overview interface class IExample { void F(int value); property bool P { bool get(); } property double default[int] { double get(int); void set(int, double); } event EventHandler^ E; };
shows an interface that contains a function F, a read-only scalar property P, a default-indexed property, and an event E, all of which are implicitly public. Interfaces are implemented using inheritance syntax.
interface class I1 { void F(); }; // F is implicitly virtual abstract ref struct R1 : I1 { virtual void F() { /* implement I1::F */ } };
An interface can require implementation of one or more other interfaces. For example
interface class IControl { void Paint(); }; interface class ITextBox : IControl { void SetText(String^ text); }; interface class IListBox : IControl { void SetItems(array<String^>^ items); }; interface class IComboBox : ITextBox, IListBox {};
A class that implements IComboBox must also implement ITextBox, IListBox, and IControl. Classes can implement multiple interfaces. In the example
interface class IDataBound { void Bind(Binder^ b); }; public ref class EditBox : Control, IControl, IDataBound { public: virtual void Paint() { } virtual void Bind(Binder^ b) { } };
the class EditBox derives from the ref class Control and implements both IControl and IDataBound. In the previous example, interface functions were implicitly implemented. C++/CLI provides an alternative way of implementing these functions that allows the implementing class to avoid having these members be public. Interface functions can be explicitly implemented using the named overriding syntax shown in 8.8.10.1. For example, the EditBox class could instead be implemented by providing IControl::Paint and IDataBound::Bind functions.
public ref class EditBox : IControl, IDataBound { private: virtual void Paint() = IControl::Paint { } virtual void Bind(Binder^ b) = IDataBound::Bind { } };
Interface members implemented in this way are called explicit interface members because each member explicitly designates the interface member being implemented.
int main() { EditBox^ editbox = gcnew EditBox; editbox->Paint(); // error: Paint is private IControl^ control = editbox; control->Paint(); // calls EditBoxs Paint implementation }
29
8.11 Enums
Standard C++ already supports enumerated types. However, C++/CLI provides some interesting extensions to this facility. For example: An enum can be declared public or private, so its visibility outside its parent assembly can be controlled. The underlying type for an enum can be specified. An enum type and/or its enumerators can have attributes. A new syntax is available for defining enums that are strongly typed and thus do not have integral promotions.
8.12 Namespaces and assemblies

The programs presented so far have stood on their own except for dependence on a few system-provided classes such as System::Console. It is far more common, however, for real-world applications to consist of several different pieces, each compiled separately. For example, a corporate application might depend on several different components, including some developed internally and some purchased from independent software vendors. Namespaces and assemblies enable this component-based system. Namespaces provide a logical organizational system. Namespaces are used both as an internal organization system for a program, and as an external organization systema way of presenting program elements that are exposed to other programs. Assemblies are used for physical packaging and deployment. An assembly can contain types, the executable code used to implement these types, and references to other assemblies. To demonstrate the use of namespaces and assemblies, this subclause revisits the hello, world program presented earlier, and splits it into two pieces: a class library that contains a function that displays the greeting, and a console application that calls that function. The class library will contain a single class named DisplayMessage. For example:
// DisplayHelloLibrary.cpp namespace MyLibrary { public ref struct DisplayMessage { static void Display() { Console::WriteLine("hello, world"); } }; }
The next step is to write a console application that uses the DisplayMessage class; for example:
// HelloApp.cpp #using <DisplayHelloLibrary.dll> int main() { MyLibrary::DisplayMessage::Display(); }
No headers need to be included when using CLI library classes and functions. Instead, library assemblies are referenced via #using directives, with the assembly name enclosed in <>, as shown. The code written can be compiled into a class library containing the class DisplayMessage and an application containing the function main. The details of this compilation step might differ based on the compiler or tool being used. A command-line compiler might enable compilation of a class library and an application that uses that library with the following command-line invocations:
cl /LD DisplayHelloLibrary.cpp cl HelloApp.cpp
which produce a class library named DisplayHelloLibrary.dll and an application named HelloApp.exe.
30
Language overview
8.13 Versioning
Versioning is the process of evolution of a component over time in a compatible manner. A new version of a component is source-compatible with a previous version if code that depends on the previous version can, when recompiled, work with the new version. In contrast, a new version of a component is binarycompatible if an application that depended on the old version can, without recompilation, work with the new version. Consider the situation of a base class author who ships a class named Base. In the first version, Base contains no function F. A component named Derived derives from Base, and introduces an F. This Derived class, along with the class Base on which it depends, is released to customers, who deploy to numerous clients and servers.
public ref struct Base { }; // version 1
public ref struct Derived : Base { virtual void F() { Console::WriteLine("Derived::F"); } };
So far, so good, but now the versioning trouble begins. The author of Base produces a new version, giving it its own function F.
public ref struct Base { // version 2 virtual void F() { // added in version 2 Console::WriteLine("Base::F"); } };
This new version of Base should be both source and binary compatible with the initial version. (If it werent possible simply to add a function then a base class could never evolve.) Unfortunately, the new F in Base makes the meaning of Deriveds F unclear. Did Derived mean to override Bases F? This seems unlikely, since when Derived was compiled, Base did not even have an F! Further, if Deriveds F does override Bases F, then it must adhere to the contract specified by Basea contract that was unspecified when Derived was written. In some cases, this is impossible. For example, Bases F might require that overrides of it always call the base. Deriveds F could not possibly adhere to such a contract. C++/CLI addresses this versioning problem by allowing developers to state their intent clearly. In the original code example, the code was clear, since Base did not even have an F. Clearly, Deriveds F is intended as a new function rather than an override of a base function, since no base function named F exists. If Base adds an F and ships a new version, then the intent of a binary version of Derived is still clear Deriveds F is semantically unrelated, and should not be treated as an override. However, when Derived is recompiled, the meaning is unclearthe author of Derived might intend its F to override Bases F, or to hide it. By default, the compiler makes Deriveds F override Bases F. However, this course of action does not duplicate the semantics for the case in which Derived is not recompiled. If Deriveds F is semantically unrelated to Bases F, then Deriveds author can express this intent by using the function modifier new in the declaration of F.
public ref struct Base { // version 2 virtual void F() { // added in version 2 Console::WriteLine("Base::F"); } }; public ref struct Derived : Base { // version 2a: new virtual void F() new { Console::WriteLine("Derived::F"); } };
31
On the other hand, Deriveds author might investigate further, and decide that Deriveds F should override Bases F. This intent can be specified explicitly by using the function modifier override, as shown below.
public ref struct Base { // version 2 virtual void F() { // added in version 2 Console::WriteLine("Base::F"); } }; public ref struct Derived : Base { // version 2b: override virtual void F() override { Base::F(); Console::WriteLine("Derived::F"); } };
The author of Derived has one other option, and that is to change the name of F, thus completely avoiding the name collision. Although this change would break source and binary compatibility for Derived, the importance of this compatibility varies depending on the scenario. If Derived is not exposed to other programs, then changing the name of F is likely a good idea, as it would improve the readability of the programthere would no longer be any confusion about the meaning of F.
8.14 Attributes
Standard C++ has certain declarative elements. For example, the accessibility of a function in a class can be specified by declaring it public, protected, or private. C++/CLI generalizes this capability, so that programmers can invent new kinds of declarative information, attach this declarative information to various program entities, and retrieve this declarative information at run-time. Programs specify this additional declarative information by defining and using attributes. For instance, a framework might define a HelpAttribute attribute that can be placed on program elements such as classes and functions, enabling developers to provide a mapping from program elements to documentation for them. The example
[AttributeUsage(AttributeTargets::All)] public ref class HelpAttribute : Attribute { String^ url; public: HelpAttribute(String^ url) { this->url = url; } String^ Topic; property String^ Url { String^ get() { return url; } } };
defines an attribute class named HelpAttribute that has one positional parameter (String^ url) and one named parameter (String^ Topic). Positional parameters are defined by the formal parameters for public instance constructors of the attribute class, and named parameters are defined by public non-static read-write fields and properties of the attribute class. For convenience, usage of an attribute name when applying an attribute is allowed to drop the Attribute suffix from the name. The example
[Help("http://www.mycompany.com//Class1.htm")] public ref class Class1 { public: [Help("http://www.mycompany.com//Class1.htm", Topic = "F")] void F() {} };
shows several uses of the attribute Help.
32
Language overview
Attribute information for a given program element can be retrieved at run-time by using reflection support. The example
int main() { Type^ type = Class1::typeid; array<Object^>^ arr = type->GetCustomAttributes(HelpAttribute::typeid, true); if (arr->Length == 0) Console::WriteLine("Class1 has no Help attribute."); else { HelpAttribute^ ha = (HelpAttribute^) arr[0]; Console::WriteLine("Url = {0}, Topic = {1}", ha->Url, ha->Topic); } }
checks to see if Class1 has a Help attribute, and writes out the associated Topic and Url values if that attribute is present.
8.15 Generics
Generic types and functions are a set of featurescollectively called genericsdefined by the CLI to allow parameterized types. Generics differ from templates in that generics are instantiated by the Virtual Execution System (VES) at runtime rather than by the compiler at compile-time. A generic definition must be a ref class, value class, interface class, delegate, or function. 8.15.1 Creating and consuming generics Below, we create a Stack generic class definition where we specify a type parameter, ItemType, using the same notation as with templates, except that the keyword generic is used instead of template. This type parameter acts as a placeholder until an actual type is specified at use.
generic<typename ItemType> public ref class Stack { array<ItemType>^ items; public: Stack(int size) { items = gcnew array<ItemType>(size); } void Push(ItemType data) { } ItemType Pop() { } };
When we use the generic class definition Stack, we specify the actual type to be used by the generic class. In this case, we instruct the Stack to use an int type by specifying it as a type argument using the angle brackets after the name:
Stack<int>^ s = gcnew Stack<int>(5);
In so doing, we have created a new constructed type, Stack<int>, for which every ItemType inside the definition of Stack is replaced with the supplied type argument int. If we wanted to store items other than an int into a Stack, we would have to create a different constructed type from Stack, specifying a new type argument. Suppose we had a simple Customer type and we wanted to use a Stack to store it. To do so, we simply use the Customer class as the type argument to Stack and easily reuse our code:
Stack<Customer^>^ s = gcnew Stack<Customer^>(10); s->Push(gcnew Customer); Customer^ c = s->Pop();
Of course, once weve created a Stack with a Customer type as its type argument, we are now limited to storing only Customer objects (or objects of a class derived from Customer). Like templates, generics provide strong typing.
33
Generic type definitions can have any number of type parameters. Suppose we created a simple
Dictionary generic class definition that stored values alongside keys. We could define a generic version of a Dictionary by declaring two type parameters, as follows: generic<typename KeyType, typename ElementType> public ref class Dictionary { public: void Add(KeyType key, ElementType val) { } property ElementType default[KeyType] { // indexed property ElementType get(KeyType key) { } void set(KeyType key, ElementType value) { } } };
When we use Dictionary, we need to supply two type arguments within the angle brackets. Then when we call the Add function or use the indexed property, the compiler checks that we supplied the right types:
Dictionary<String^, Customer^>^ dict = gcnew Dictionary<String^, Customer^>; dict->Add("Peter", gcnew Customer); Customer^ c = dict["Peter"];
8.15.2 Constraints In many cases, we will want to do more than just store data based on a given type parameter. Often, we will also want to use members of the type parameter to execute statements within our generic type definition. For example, suppose in the Add function of our Dictionary we wanted to compare items using the CompareTo function of the supplied key, as follows:
generic<typename KeyType, typename ElementType> public ref class Dictionary { public: void Add(KeyType key, ElementType val) { if (key->CompareTo(val) < 0) { } // compile-time error
} };
Unfortunately, at compile-time the type parameter KeyType is, as expected, generic. As written, the compiler will assume that only the operations available to System::Object, such as calls to the function ToString, are available on the variable key of type KeyType. As a result, the compiler will issue a diagnostic because the CompareTo function would not be found. However, we can cast the key variable to a type that does contain a CompareTo function, such as an IComparable interface, allowing the program to compile:
generic<typename KeyType, typename ElementType> public ref class Dictionary { public: void Add(KeyType key, ElementType val) { if (static_cast<IComparable^>(key)->CompareTo(val) < 0) { }
} };
However, if we now construct a type from Dictionary and supply a key type argument which does not implement IComparable, we will encounter a run-time error (in this case, a System::InvalidCastException). Since one of the objectives of generics is to provide strong typing and to reduce the need for casts, a more elegant solution is needed. We can supply an optional list of constraints for each type parameter. A constraint indicates a requirement that a type must fulfill in order to be accepted as a type argument. (For example, it might have to implement a given interface or be derived from a given base class.) A constraint is declared using the word where, followed by a type parameter and colon (:), followed by a comma-separated list of class or interface types.
34
Language overview
In order to satisfy our need to use the CompareTo function inside Dictionary, we can impose a constraint on KeyType, requiring any type passed as the first argument to Dictionary to implement IComparable, as follows:
generic<typename KeyType, typename ElementType> where KeyType : IComparable public ref class Dictionary { public: void Add(KeyType key, ElementType val) { if (key->CompareTo(val) < 0) { } }; }
When compiled, this code will now be checked to ensure that each time we construct a Dictionary type we are passing a first type argument that implements IComparable. Further, we no longer have to explicitly cast variable key to an IComparable interface before calling the CompareTo function. Constraints are most useful when they are used in the context of defining a framework, i.e., a collection of related classes, where it is advantageous to ensure that a number of types support some common signatures and/or base types. Constraints can be used to help define generic algorithms that plug together functionality provided by different types. This can also be achieved by subclassing and runtime polymorphism, but static, constrained polymorphism can, in many cases, result in more efficient code, more flexible specifications of generic algorithms, and more errors being caught at compile-time rather than runtime. However, constraints need to be used with care and taste. Types that do not implement the constraints will not easily be usable in conjunction with generic code. For any given type parameter, we can specify any number of interfaces as constraints, but no more than one base class. Each constrained type parameter has a separate where clause. In the example below, the KeyType type parameter has two interface constraints, while the ElementType type parameter has one class constraint:
generic<typename KeyType, typename ElementType> where KeyType : IComparable, IEnumerable where ElementType : Customer public ref class Dictionary { public: void Add(KeyType key, ElementType val) { if (key->CompareTo(val) < 0) { }
} };
8.15.3 Generic functions In some cases, a type parameter is not needed for an entire class, but only when calling a particular function. Often, this occurs when creating a function that takes a generic type as a parameter. For example, when using the Stack described earlier, we might often find ourselves pushing multiple values in a row onto a stack, and decide to write a function to do so in a single call. We do this by writing a generic function. Like a generic class definition, a generic function is preceded by the keyword generic and a list of type parameters enclosed in angle brackets. As in a template function, the type parameters of a generic function can be used within the parameter list, return type, and body of the function. A generic PushMultiple function might look like this:
generic<typename StackType, typename ItemType> where ItemType : StackType void PushMultiple(Stack<StackType>^ s, ... array<ItemType>^ values) { for each (ItemType v in values) { s->Push(v); } }
35
Using this generic function, we can now push multiple items onto a Stack of any kind. Furthermore, because a constraint exists, the compiler type checking will ensure that the pushed items have the correct type for the kind of Stack being used. When calling a generic function, we place type arguments to the function in angle brackets; for example:
Stack<int>^ s = gcnew Stack<int>(5); PushMultiple<int,int>(s, 1, 2, 3, 4);
The call to this function supplies the desired StackType and ItemType as type arguments to the function. In many cases, however, the compiler can deduce the correct type argument from the other arguments passed to the function, using a process called type deduction. In the example above, since the first regular argument is of type Stack<int>, and the subsequent arguments are of type int, the compiler can reason that the type parameter must also be int. Thus, the generic PushMultiple function can be called without specifying the type parameter, as follows:
Stack<int>^ s = gcnew Stack<int>(5); PushMultiple(s, 1, 2, 3, 4);
End of informative text.
36
Lexical structure
9. Lexical structure
9.1 Tokens
9.1.1 Identifiers Certain places in the Standard C++ grammar do not allow identifiers. However, C++/CLI allows a defined set of identifiers to exist in those places, with these identifiers having special meaning. [Note: Such identifiers are colloquially referred to as context-sensitive keywords; nonetheless, they are identifiers. end note] The identifiers that carry special meaning in certain contexts are:
abstract initonly where delegate internal event literal finally override generic property in sealed
When referred to in the grammar, these identifiers are used explicitly rather than using the identifier grammar production. Ensuring that the identifier is meaningful is a semantic check rather than a syntax check. An identifier is considered a keyword in a given context if and only if there is no valid parse if the token is taken as an identifier. That is, if it can be an identifier, it is an identifier. Some naming patterns are reserved for function names in certain contexts (19.2, 19.7.5). When the token generic is found, it has special meaning if and only if it is not preceded by the token :: or typename, and is followed by the token < and then either of the keywords class or typename. [Note: In rare cases, a valid Standard C++ program could contain the token sequence generic followed by < followed by class, where generic should be interpreted as a type name. For example:
template<typename T> struct generic { typedef int I; }; class X {}; generic<class X> x1; generic<class X()> x2;
In such cases, use typename to indicate that the occurrence of generic is a type name:
typename generic<class X> x1; typename generic<class X()> x2;
or, in these particular cases, an alternative would be to remove the keyword class (that is, to not use the elaborated-type-specifier), for example:
generic<X> x1; generic<X()> x2;
end note] The grammar productions for elaborated-type-specifier (C++ Standard 7.1.5.3, 14.6, and A.6) that mention typename are augmented as follows, to make nested-name-specifier optional in the first of the two applicable productions: elaborated-type-specifier: attributesopt class-key ::opt nested-name-specifieropt identifier attributesopt class-key ::opt nested-name-specifieropt templateopt template-id attributesopt enum-key ::opt nested-name-specifieropt identifier attributesopt typename ::opt nested-name-specifieropt identifier attributesopt typename ::opt nested-name-specifier templateopt template-id
37
[Note: Revision of the C++ Standard is currently underway, and changes proposed in that revision alter this production. end note] attributes is described in 29. The C++ Standard (14.6/3) is augmented, as follows: An qualified-ididentifier that refers to a type and in which the nested-name-specifier depends on a template-parameter (14.6.2) shall be prefixed by the keyword typename to indicate that the qualified-ididentifier denotes a type, forming an elaborated-type-specifier (7.1.5.3). and 14.6/5 is deleted: The keyword typename shall only be used in template declarations and definitions, including in the return type of a function template or member function template, in the return type for the definition of a member function of a class template or of a class nested within a class template, and in the typespecifier for the definition of a static member of a class template or of a class nested within a class template. The keyword typename shall be applied only to qualified names, but those names need not be dependent. The keyword typename shall be used only in contexts in which dependent names can be used. This includes template declarations and definitions but excludes explicit specialization declarations and explicit instantiation declarations. The keyword typename is not permitted in a base-specifier or in a mem-initializer; in these contexts a qualified-id that depends on a templateparameter (14.6.2) is implicitly assumed to be a type name. [Note: The presence of typename lets the programmer disambiguate otherwise ambiguous cases such as the token sequence property :: X x;. The declaration property :: X x; declares a member variable named x of type property::X, as it does in Standard C++. The token sequence property typename :: X x; declares a property named x of type ::X. end note] When name lookup for any of array, interior_ptr, pin_ptr, or safe_cast fails to find the name, and the name is not followed by a left angle bracket (<), the name is interpreted as though it were qualified with cli:: and the lookup succeeds, finding the name in namespace ::cli. When name lookup for any of array, interior_ptr, pin_ptr, or safe_cast succeeds and finds the name in namespace ::cli, the name is not a normal identifier, but has special meaning as described in this Standard. Tokens that are not identifiers can be used as identifiers. This is achieved via __identifier(T), where T shall be an identifier, a keyword, or a string-literal. The string-literal form is reserved for use by C++/CLI implementations. It is unspecified whether this replacement takes place before or after translation phase 4. [Note: Therefore, this construct should not be used in place of the first or only identifier in a #define preprocessing directive. end note] [Example:
__identifier(totalCost) __identifier(delete) __identifier("<Special Name #3>")
end example] 9.1.2 Keywords The list of keywords in the C++ Standard (2.11) is augmented by the following:
enumclass interfaceclass refstruct enumstruct interfacestruct valueclass foreach nullptr valuestruct gcnew refclass
The symbol is used in the grammar to signify that white-space appears within the keyword. Any white space that appears in the program text after translation phase 1 is permitted in the position signified by the symbol. It is unspecified whether white space generated by comments, documentation comments, and macro invocations is permitted in the position signified by the symbol. Following translation phase 4, a keyword with will be a single token. [Note: The symbol is only used in the grammar of the language.
38
Lexical structure
Examples will include white-space as is required in a well-formed program. end note] [Note: Keywords that include the symbol can be produced by macros, but are never considered to be macro names. end note] Translation phase 4 in the C++ Standard (2.1/4) is augmented as follows: Preprocessing directives are executedparsed and stored. Then, in the translation unit and in each macro replacement-list, starting with the first token, each pair of adjacent tokens token1 and token2 is successively considered, and if token1token2 is a keyword, then token1 and token2 are replaced with the single token token1token2. and Then macro invocations are expanded. ... In some places in the grammar, certain identifiers have special meaning, but are not keywords. [Note: For example, within a virtual function declaration, the identifiers abstract and sealed have special meaning. Ordinary user-defined identifiers are never permitted in these locations, so this use does not conflict with a use of these words as identifiers. For a complete list of these special identifiers, see 9.1.1. end note] 9.1.3 Literals The grammar for literal in the C++ Standard (2.13) is augmented as follows: literal: integer-literal character-literal floating-literal string-literal boolean-literal null-literal 9.1.3.1 Integer literals To accommodate the addition of the types long long int and unsigned long long int, the grammar for integer-suffix in the C++ Standard (2.13.1) is augmented as follows: integer-suffix: unsigned-suffix long-suffixopt unsigned-suffix long-long-suffixopt long-suffix unsigned-suffixopt long-long-suffix unsigned-suffixopt long-long suffix: one of
ll LL
The C++ Standard (2.13.1/2) is augmented as follows: The type of an integer literal depends on its form, value, and suffix. If it is decimal and has no suffix, it has the first of these types in which its value can be represented: int, long int, long long int; if the value cannot be represented as a long int, the behavior is undefined. If it is octal or hexadecimal and has no suffix, it has the first of these types in which its value can be represented: int, unsigned int, long int, unsigned long int, long long int, unsigned long long int. If it is suffixed by u or U, its type is the first of these types in which its value can be represented: unsigned int, unsigned long int, unsigned long long int. If it is decimal and is suffixed by l or L, its type is the first of these types in which its value can be represented: long int, unsigned long intlong long int. If it is octal or hexadecimal and is suffixed by l or L, its type is the first of these types in which its value can be represented: long int, unsigned long int, long long int, unsigned long long int. If it is suffixed by ul, lu, uL, Lu, Ul, lU, UL, or LU, its type is the first of these types in which its value can be represented: unsigned long int, unsigned long long int. If it is decimal and is suffixed by ll or LL, its type is long long int. If it is octal or hexadecimal and is suffixed by ll or LL, its type is the first of these types in which its value can be represented: long long int, unsigned long long int. If it is suffixed by both u or U and ll or LL, its type is unsigned long long int. 39
To accommodate the addition of extended integer types, the C++ Standard (2.13.1/3) is augmented as follows: If an integer constant cannot be represented by any type in its list and an extended integer type can represent its value, then it has an extended integer type. If all of the types in the list for the constant are signed, the extended integer type shall be signed. If all of the types in the list for the constant are unsigned, the extended integer type shall be unsigned. If the list contains both signed and unsigned types, the extended integer type may be signed or unsigned. A program is ill-formed if one of its translation units contains an integer literal that cannot be represented by any of the allowed types. 9.1.3.2 The null literal null-literal:
nullptr
The null-literal is the keyword nullptr, whose type is the null type (12.3.4). nullptr represents the null value constant and is unique. This literal is not an lvalue. The null value constant can be converted to any handle type, with the result being a null handle. The null value constant can also be converted to any pointer type, with the result being a null pointer. 9.1.3.3 String literals The C++ Standard (2.13.4/1) is augmented as follows: An ordinary string literal has type <narrow-string-literal-type>. This type cannot be named in the language, but it can be converted implicitly to either System::String^ or array of n const char, as described in 14.2.5. array of n const char and static storage duration (3.7), where n is the size of the string as defined below, and is initialized with the given characters. A wide string literal has type <wide-string-literal-type>. This type cannot be named in the language, but it can be converted implicitly to either System::String^ or array of n const wchar_t, as described in 14.2.5. array of n const wchar_t and has static storage duration, where n is the size of the string as defined below, and is initialized with the given characters. 9.1.4 Operators and punctuators C++/CLI requires that template and generic constructs such as List<List<int>> be permitted, where >> is treated as two tokens instead of one. This requires augmentations to a number of places in the C++ Standard, as specified in this subclause and the subclauses 15.3, 30.1, and 30.2. The C++ Standard (2.1/1), translation phase 7, is augmented by adding the following text just prior to the existing note: [Note: The process of analyzing and translating the tokens may occasionally result in one token being replaced by a sequence of other tokens (14.2). end note]
40
Basic concepts
10. Basic concepts

10.1 Assemblies
The CLI defines an assembly as a configured set of loadable code modules and resources that together implement a unit of functionality. A C++/CLI program recognizes an assembly by the name of the file containing the assembly manifest. An assembly manifest describes all the constituent parts of the assembly such as the name of the assembly in metadata, other files that contribute to the assembly, and any hash codes that validate constituent parts. An assembly can be an application or a library. An application has an application entry point, whereas a library does not.
10.2 Application entry point

In addition to the two definitions of the main function allowed in Standard C++ (see 3.6.1), C++/CLI allows the following definition:
int main(array<System::String^>^ args) { /* ... */ }
The value of args shall be a CLI array that represents the arguments to the program, where index 0 contains the first argument. If no arguments were passed to the program, args shall be a zero-length array; args shall never be null. The array passed to main is generated by the CLI runtime. [Note: Application entry points are described in 15.4.1.2 of the CLI Standard. end note]
10.3 Importing types from assemblies

Each type definition resides in some assembly, and an assembly can contain one or more types. The CLI Standard defines many types, each of which is defined in one of the three following assemblies: mscorlib.dll, System.dll, and System.Xml.dll. An application programmer can create any number of other assemblies, as needed. A #using directive makes types from an assembly available in a source file; that is, it imports types from the metadata, and does not cause any types to be defined in the current translation unit. This directive has the following forms, which are equivalent:
#using < assembly-name > #using " assembly-name "
[Note: Despite its appearance, #using is not a preprocessing directive. end note] The types in assembly mscorlib.dll shall be implicitly imported by the compiler. [Example:
#using <mscorlib.dll> // redundant #using <System.dll> // needed for Socket #using <System.Xml.dll> // needed for XmlTextReader int main() { System::Text::StringBuilder^ strBld; System::Net::Sockets::Socket^ soc; System::Xml::XmlTextReader^ xtr; }
Each type has a namespace, a parent assembly, and a parent library; all three characteristics are separate and unrelated. For example, the type Socket is in the namespace System::Net::Sockets, the assembly System.dll, and the Networking library. end example] For metadata details, see 34.1.1.
41
When a #using directive imports a type from an assembly, that type continues to belong to that assembly regardless of the number of other assemblies into which it is imported. On the other hand, when a #include preprocessing directive brings in a header containing a type definition, it brings in source code, which, when compiled, defines that type in the current translation unit. When #using an assembly, if an imported type has a function with a signature that contains a modopt (33.1) not defined by this Standard or one that has been used in a manner not defined by this Standard (for instance, using IsSignUnspecifiedByte (33.1.5.7) on something other than a System::Byte or System::SByte), the following rules apply: If no other signature in the type is the same when ignoring the modopt, the compiler shall use the signature as if the modopt did not exist. Then if the function is virtual, any overriding function shall repeat the modopt. If when ignoring the modopt the functions signature is the same as another functions signature in the type, the compiler shall ignore the function with the unknown modopt, treating that function as if it did not exist. If there are two or more signatures with unknown modopts, and no signatures without modopts, all of the functions are ignored.
When #using an assembly, any value class type that has the NativeCppClass attribute (33.2.1, 34.8), is treated as a native class, as described below. (If a type other than a value class has this attribute applied to it, the attribute is ignored and the type is treated as though the attribute had not been present.) A value class brought in from another assembly via #using is a forward declaration for that type. If a definition of the class is in source code, it is treated as the same class as that being brought it if the following criteria are met: o o o The source code definition has the same name as the encoding that came from #using. The size of the source code definition is identical to the size in the encoding. The visibility of the two need not be the same.
Being treated as "the same" means the following: Whenever the type from another assembly is used, the type defined in source code (in the current assembly) can be substituted. This is not a conversion. Whenever type information is needed for instructions such as call, the type used will match the function being called, but the type being supplied can be substituted by an object of the matching type in the current assembly. Whenever type information must be introduced in the current assembly (i.e., function parameter metadata), the type used shall be the type from the current assembly. The only exception is virtual overriding in a ref class. The signature of the virtual function shall match the original. Thus if the signature includes a native type, any function overriding it shall use the same type in its encoding.
All access to the native type using non-virtual functions shall be with functions from the current assembly. Member functions shall be private to each assembly. When #using an assembly, if that assembly cannot be found or it is found but has an invalid format according to the CLI Standard, the compiler shall behave as if a corresponding #error directive was encountered.
10.4 Reserved names

There are certain functions that a programmer can never write in C++/CLI, but which may need to be imported from metadata created by translators of other languages. [Example: This can happen when a name 42
Basic concepts
is reserved and cannot be written by the programmer; for example, Finalize, Dispose, or any of the operator function names. end example]
#using can import types with names that cannot be authored in C++/CLI. A C++/CLI programmer can use
such a name in an expression when the reserved name does not have the meaning C++/CLI gives it. [Example: If a function named Finalize does not override the Finalize method from System::Object, a C++/CLI programmer can call the function Finalize without using the !T syntax (19.13.2). A second example involves the following C# class:
public class C : IDisposable { void IDisposable.Dispose() {} public void Dispose() {} }
the function C::Dispose can be called from C++/CLI when #using that C# class because C::Dispose does not implement the IDisposable::Dispose function or override any function that does implement IDisposable::Dispose. A third example is when an imported class has an implicit and explicit conversion operator that do the same thing. In this case, the compiler should just fall back to allowing the developer to write op_Implicit or op_Explicit. end example] See also __identifier (9.1.1).
10.5 Members
10.5.1 Value class members The members of a value class are the members declared in that value class, and the members inherited from the value classs direct base class System::ValueType and the indirect base class System::Object. The members of a fundamental type are the members of the corresponding value class type provided by the implementation (12.1). [Example: The members of signed char are the members of the System::SByte value class. end example] 10.5.2 Delegate members The members of a delegate are the members inherited from class System::Delegate, a public instance constructor, and the public methods BeginInvoke, EndInvoke, and Invoke (34.14).
10.6 Member access

10.6.1 Declared accessibility In the C++ Standard (10), an access-specifier is used to define member access control. This grammar is augmented to accommodate the notion of assemblies, as follows: access-specifier:
private protected public internal protected public public protected private protected protected private
In the C++ Standard (11/1), member access control for each access-specifier is defined. To accommodate the addition of assemblies, the list of definitions is augmented, as follows: A member of a class can be 43
private; that is, its name can be used only by members and friends of the class in which it is declared. This is referred to as private access. protected; that is, its name can be used only by members and friends of the class in which it
is declared, and by members and friends of classes derived from this class (see 11.5). The parent assembly of derived classes does not affect protected access. This is referred to as family access.
public; that is, its name can be used anywhere without access restriction. This is referred to as
public access.
internal; that is, its name can be used in its parent assembly. This is referred to as assembly
access.
public protected or protected public; that is, its name can be used in its parent
assembly or by types derived from the containing class. This is referred to as family or assembly access.
private protected or protected private; that is, its name can be used only by types
derived from the containing class within its parent assembly. This is referred to as family and assembly access. [Note: For access-specifiers containing two keywords, the more restrictive of the two applies outside the parent assembly while the less restrictive of the two applies within the parent assembly. end note] An overriding name is allowed to have a different accessibility than the name it is overriding. An ordering is applied to distinguish between greater accessibility. Given the two accessibilities A and B, A has narrower access than B if A permits less access than A within the assembly and outside the assembly. A has wider access than B if A permits more access than A within the assembly and outside the assembly. Narrowing and widening of accessibilities implies a total ordering of accessibilities. For example, protected is wider than private, protected is narrower than public, protected private is narrower than public protected, and no ordering exists between internal and protected. [Note: In general, widening and narrowing accessibility is not CLS compliant. end note] When no ordering exsts between two accessibilities, one shall not be used to override the other. When requirements are placed on wider or narrower accessibility, only the directly associated access specifier is considered. While accessiblity to a class member or type is determined by first checking accessibility of the enclosing entity, widening and narrowing rules do not consider the enclosing entity. [Example: The following code is valid.
public ref struct B { ref struct NB { virtual void F(); }; }; private ref class D : B { ref class ND : B::NB { public: virtual void F() override; }; };
The overriding virtual function F in ND cannot have narrower accessibility than the virtual function F in NB. Since NB::F has public accessibility, ND::F must also have public accessibility. Both D and ND having private accessibility do not affect the narrowing rules. end example] For metadata details, see 34.7.2.
10.7 Name lookup

The CLI (Partition I, 8.10.4) supports two different approaches to name lookup in base classes: If a derived member is marked hide-by-name, then functions in the base class with the same name are not visible in the derived class. This approach is referred to as hidebyname. 44
Basic concepts
If a derived member is marked hide-by-name-and-signature, then functions in the base class with the same name and signature are not visible in the derived class. This approach is referred to as hidebysig.
Implementation of the distinction between these two forms of hiding is provided entirely by source language compilers and the reflection library; it has no direct impact on the VES itself. [Note: As in Standard C++, during lookup, whether the functions in a candidate set are static, virtual, or nonvirtual, has no effect on overload resolution. end note] The C++ Standard requires hidebyname lookup. As such, member functions of native classes use hidebyname lookup. [Example: Given the following program:
struct B { void F(int i) { } }; struct D : B { void F(String^ d) { } }; int main() { D d; d.F(100); }
the function F(String^) is found, it's incompatible, and results in an error. end example] On the other hand, member functions of ref classes, value classes, interface classes, and delegates use hidebysig lookup. [Example: Given the following program:
ref struct B { void F(int i) { } }; ref struct D : B { void F(String^ d) { } }; int main() { D d; d.F(100); }
the function F(int) is called. end example] If lookup for a name begins in a class, base interfaces are ignored. If lookup for a name begins in an interface, when lookup proceeds to the bases of that interface, it shall continue searching for names in those interfaces. The C++ Standard (3.4/1) states: The access rules (clause 11) are considered only once name lookup and function overload resolution (if applicable) have succeeded. In C++/CLI, that rule applies only to native classes. Otherwise, for CLI class types, inaccessible functions are not visible to name lookup. [Note: In Standard C++, a private name can hide names in a base class, whereas, in a CLI class type, a private name cannot hide names in a base class. end note] [Note: In hidebyname, name lookup stops as soon as the name is found in a scope. In hidebysig, lookup continues unless the signature also matches. end note] For qualified name lookup, lookup begins in the scope specified. If that scope uses hidebysig rules, then lookup uses hidebysig rules to find all names in the specified scope and other scopes. [Example: an expression such as expr->R::F, if R is a hidebysig class, lookup begins in R. Normal hidebysig rules apply, and thus a name set including names found in base classes of R is possible. end example]
45
Because hidebysig rules can create ambiguities between functions in a base class and a function in a derived class, the overload resolution rules are augmented to prefer functions in a derived class. [Note: Overload resolution is the same for candidate overload sets produced by hidebyname and hidebysig lookup. This can lead to ambiguity. end note] In C++/CLI, functions in derived classes are preferred. To accomplish this, the C++ Standard (13.3.3) is augmented, as follows: Given these definitions, a viable function F1 is defined to be a better function than another viable function F2 if for all arguments i, ICSi(F1) is not a worse conversion sequence than ICSi(F2), and then F1 is a member of a more derived class than F2 and neither F1 nor F2 are conversion functions, or if not that, for some argument j, ICSj(F1) is a better conversion sequence than ICSj(F2), or, if not that, [Note: With that rule, the program below will print float. end note] [Example:
ref struct B { void F(double) { Console::WriteLine("double"); } }; ref struct D : B { void F(float) { Console::WriteLine("float"); } }; int main() { D d; d.F(3.14); }
The conversions from (D^, double) to (B^, double) and (D^, float) are equally ranked. Thus, with no additional rules the call would be ambiguous. end example] If lookup in a class finds an entity that is not a function, lookup does not continue in the base classes. If lookup originated in a derived class, and the lookup set already contains a function, the entity in the base class is not included in the name set. (For the purpose of lookup, properties and events are treated as fields.) [Example:
ref struct A { void F(Object^) { Console::WriteLine("A::F"); } }; ref struct B : A { int F; }; ref struct C : B { void F(String^) { Console::WriteLine("C::F"); } }; int main() { C c; c.F(4); }
// error
No function F will be found because when lookup starts in C, it finds a function, then stops in B because a field with the same name exists. The same would happen if B::F were a property or event. end example] A function scope is always hidebyname. As such, if lookup finds a name in function scope, it does not continue looking further. [Example:
46
Basic concepts ref struct R { void F(Object^) { Console::WriteLine("R::F(Object^)"); } void F() { extern void F(String^); F(4); // error Console::WriteLine("R::F()"); } }; int main() { R r; r.F(); } void F(String^) { Console::WriteLine("::F(String^)"); }
The program is ill-formed because the argument 4 cannot be converted to String^, which is the only viable function that lookup finds. end example] A program that contains the definitions of two or more generic types with the same name and different arity (31) in the same namespace, is ill-formed. However, a C++/CLI program can import such types from other assemblies with #using. When this happens, the ambiguity shall be resolved by counting the number of type arguments.
47
11. Preprocessor
11.1 Conditional inclusion
To accommodate the addition of the types long long int and unsigned long long int, and extended integer types, the C++ Standard (16.1/4) is augmented, as follows: The resulting tokens comprise the controlling constant expression which is evaluated according to the rules of 5.19 using arithmetic that has at least the ranges specified in 18.2, except that int and unsigned int all signed and unsigned integer types act as if they have the same representation as, respectively, the largest signed integer type or unsigned integer type.
11.2 Predefined macro names

In addition to the macros specified in the C++ Standard (16.8), the following macro name shall be defined by the implementation:
__cplusplus_cli The name __cplusplus_cli is defined to the value 200509L when compiling a C++/CLI translation unit. [Note: It is intended that future versions of this standard will replace the value of this macro with a greater value. end note]
The value of this predefined macro remains constant throughout the translation unit. If this pre-defined macro name is the subject of a #define or a #undef preprocessing directive, the behavior is implementation-defined.
48
Types
12. Types
All values in C++/CLI have a type. Types are grouped into seven categories as described in the following table. Type Category Native Class Ref Class Type Subcategory POD Union Boxed Value Type Delegate CLI Array Fundamental Type Enum Pointer Value Class
Value Type Interface Native Array Handle Reference
Native Reference Tracking Reference
Ref class types, value class types, and interface types are collectively known as CLI class types. The C++ Standard (3.9/10) definition for scalar types is augmented, as follows: Arithmetic types (3.9.1), enumeration types, handle types, pointer types, and pointer to member types (3.9.2), and cv-qualified versions of these types (3.9.3) are collectively called scalar types. The C++ Standard (7.1.5) definition for type-specifier is augmented, as follows: type-specifier: simple-type-specifier class-specifier enum-specifier elaborated-type-specifier cv-qualifier delegate-specifier To accommodate the addition of the types long long int and unsigned long long int, the C++ Standard (7.1.5.2/Table 7) is augmented by the following rows: Specifier(s)
long long signed long long long long int signed long long int unsigned long long unsigned long long int
Type "signed long long int" "signed long long int" "signed long long int" "signed long long int" "unsigned long long int" "unsigned long long int"
49
12.1 Value types

Value types consist of the fundamental types, enums, pointers, and value classes. [Note: Standard C++ distinguishes between class types and non-class types; in C++/CLI, the fundamental types and enums have characteristics of both (see 12.1.1). All value types, with the exception of pointers, have the ability to be boxed through a boxing conversion (14.2.6). end note] Fundamental types are those that are built-into the language and have keywords associated with them. Enums are declared with the enum, enum class, or enum struct keywords. Pointers are declared using the asterisk in a declarator. Value classes are declared with the value class or value struct keywords. 12.1.1 Fundamental types To accommodate the addition of the types long long int and unsigned long long int, and extended integer types, Standard C++ (3.9.1) is augmented, as follows: 3.9.1/2: "There are fourfive standard signed integer types: signed char, short int, int, and long int, and long long int. In this list, each type provides at least as much storage as those preceding it in the list. Plain ints have the natural size suggested by the architecture of the execution environment; the other signed integer types are provided to meet special needs. There may also be implementation-defined extended signed integer types. The standard and extended signed integer types are collectively called signed integer types." 3.9.1/3: "For each of the standard signed integer types, there exists a corresponding (but different) standard unsigned integer type: unsigned char, unsigned short int, unsigned int, and unsigned long int, and unsigned long long int, each of which occupies the same amount of storage and has the same alignment requirements (3.9) as the corresponding signed integer type; that is, each signed integer type has the same object representation as its corresponding unsigned integer type. Likewise, for each of the extended signed integer types there exists a corresponding extended unsigned integer type with the same amount of storage and alignment requirements. The standard and extended unsigned integer types are collectively called unsigned integer types. The range of nonnegative values of a signed integer type is a subrange of the corresponding unsigned integer type, and the value representation of each corresponding signed/unsigned type shall be the same. The standard signed integer types and standard unsigned integer types are collectively called the standard integer types, and the extended signed integer types and extended unsigned integer types are collectively called the extended integer types." 3.9.1, footnote 43): Therefore, enumerations (7.2) are not integral; however, enumerations can be promoted to int, unsigned int, long, or unsigned long, integral types as specified in 4.5." For all fundamental types (not just character types), all bits of the object representation participate in the value representation. An object of type char shall have exactly 8 bits. The value of an object having a signed integer type shall be stored using twos-complement representation.
signed char maps to System::SByte. unsigned char maps to System::Byte.
The fundamental types map to corresponding value class types provided by the implementation, as follows:
If a plain char is signed, char maps to System::SByte; otherwise, it maps to System::Byte. For all other fundamental types, the mapping is implementation-defined.
The representation of the bool value false shall be all-bits-zero.
50
Types
In the C++ Standard, fundamental types are not considered class types; however, C++/CLI introduces class members to all fundamental types as every fundamental type shall map to a CLI class determined by the implementation. In C++/CLI, when a member selection operator is applied to an expression of fundamental type, or the scope resolution operator is applied to that fundamental types keyword or typedef, in the scope of the expression containing the member selection operator or scope resolution operator, that fundamental type is treated as a class type. [Note: If a fundamental type is represented by more than one keyword, such as unsigned int, the scope resolution operator shall be applied to a typedef or the CLI class name to access static members. end note] As soon as the member selection operator or the scope resolution operator are used, C++/CLI uses the fundamental types equivalent value class to resolve members. As member access and scope resolution are not allowed on fundamental types in the C++ Standard, all scenarios that distinguish between class and non-class types in the C++ Standard will always consider fundamental types as non-classes. [Example: In the following example, the scope resolution operator applied to the keyword int results in looking for the name Parse in the associated CLI value class type. The member selection operator applied to the expression x with type int results in looking for the name ToString in the associated CLI value class type.
int x = int::Parse("42"); String^ s = x.ToString();
end example]
12.2 Class types

Ref class types, value class types, interface types, and delegate types shall not be declared at block scope. 12.2.1 Value classes [Note: A value class is a data structure that can contain fields, function members, and nested types. Unlike other class types, value classes do not support user-defined destructors, finalizers, default constructors, copy constructors, or copy assignment operators. Value classes are designed to allow the CLI execution engine to efficiently copy value class objects. All value class types implicitly inherit from the class System::ValueType, which, in turn, inherits from class System::Object. System::ValueType is not itself a value class type. Rather, it is a ref class type, from which all value class types are automatically derived. Value classes are described in 22. end note] 12.2.2 Ref classes [Note: A ref class defines a data structure that can contain fields, function members (functions, properties, events, operators, instance constructors, destructors, finalizers, and static constructors), and nested types. Ref classes support inheritance. Instances of ref classes are created using new-expressions (15.4.6). Ref classes are described in 21. end note] 12.2.3 Interface classes [Note: An interface defines a contract. A ref or value class that implements an interface shall adhere to its contract. An interface can inherit from multiple base interfaces, and a ref or value class can implement multiple interfaces. Interface classes are described in 25. end note] 12.2.4 Delegate types [Note: A delegate is a data structure that refers to one or more functions, and for instance functions, it also refers to their corresponding instances. Delegate types are described in 27. end note]
51
12.3 Declarator types

The C++ Standard (8.3.5/3) is augmented, as follows: The resulting list of transformed parameter types and the presence or absence of the ellipsis is the functions parameter-type-list. 12.3.1 Raw types A raw type is a class or fundamental type. [Note: This excludes "handle to" and "pointer to" types. end note] 12.3.2 Pointer types It is possible to declare a pointer to a function that takes a parameter array (18.4). [Example:
void F(double, ... array<int>^); void (*p)(double, ... array<int>^) = &F;
end example] A native pointer cannot point to a CLI heap-based object unless that object has been pinned (12.3.7). 12.3.3 Handle types For any CLI class type T, the declaration T^ h declares a handle h to type T, where the object to which h is capable of pointing resides on the CLI heap. A handle tracks, is rebindable, and can point to a whole CLI heap-based object only. [Note: In general, handles are to the gc heap as pointers are to the native heap. end note] The default initial value of a handle shall be nullptr. Objects of CLI class type are allocated on the CLI heap via gcnew, and such objects are referred to by handles. [Example:
R^ r1 = gcnew R; R^ r2 = r1; // allocate an object on the CLI heap // handles r1 and r2 refer to the same object
end example] If an object allocated using gcnew is never destroyed (using delete or by an explicit destructor call), that objects destructor will never be run; however, the garbage collector will reclaim the objects memory, and the objects finalizer (19.13), if one exists, will be run. [Example:
{ R^ r3 = gcnew R; } // the object will be garbage-collected and // finalized, but its destructor will not be run // allocate an object on the CLI heap
end example] Unlike pointers, handles track; that is, a handles value can change as the CLI heap-based object to which it refers is moved by the garbage collector. This has the following implications: [Example: A handle cannot be converted to and from void*. (A handle can, however, be converted to and from Object^.) [Note: There is no void^. end note] A handle cannot be converted to and from an integral type. (A handle cannot be hidden from the garbage collector.) Handles cannot be ordered. A handle can only point to a whole CLI heap-based object.
52
Types R^ r4 = gcnew R; Object^ o = r4; // ok R^ r5 = dynamic_cast<R^>(o); // ok, r4 and r5 point to the same object long l = reinterpret_cast<long>(r5); // error, cant convert to integer R^ r6 = reinterpret_cast<R^>(l); // error, cant convert from integer std::set<R^> s; // error, R^s cant be compared with less
end example] All handles to the same CLI heap-based object compare equal, even if that object is moved by the garbage collector. A handle can have any storage duration. The representation of a handle with value nullptr shall be all-bits-zero. 12.3.4 Null type The null type is a special type that exists solely to support the null-literal, nullptr (also referred to as the null value constant). No instances of this type can be created; the only way to obtain a value of this type is via the nullptr literal, whose type is the null type. 12.3.5 Reference types A native reference can bind to any lvalue. As an object on the CLI heap can be moved by the garbage collector, its location must be tracked. As such, a reference to such an object is called a tracking reference (%), and it can bind to any gc-lvalue. Whenever an object is definitively not on the CLI heap (as is the case if the object is an instance of a native class, a pinning pointer, or an interior pointer), the instance is an lvalue. [Note: As such, a native class does not need a copy assignment operator or copy constructor that takes gc-lvalues. An N% can be passed to these functions safely, since instances of native class types are never allocated on the CLI heap. An N% is an lvalue to begin with, so taking the address of an N% results in a native pointer, not an interior pointer. end note] [Note: Because there is a standard conversion from lvalue to gc-lvalue, a tracking reference can therefore bind to any gc-lvalue or lvalue. end note] For any type T, the declaration T% r declares a tracking reference r to type T. [Example:
R^ h = gcnew R; R% r = *h; void F(V% r); F(*gcnew V); // allocate on CLI heap // bind tracking reference to ref class object // bind tracking reference to value class object
end example] A tracking reference can refer to an instance of a ref class type, a cv-qualified value class type, a cvqualified handle type, a cv-qualified native class type, or a cv-qualified native pointer. A program containing tracking references that refer to other types is ill-formed. Like a native reference, a tracking reference is not rebindable; once set, its value cannot be changed. A program containing a tracking reference that has storage duration other than automatic is ill-formed. (This precludes having a tracking reference as a data member.) [Note: This limitation directly reflects that of the CLI, because, in general, tracking references are implemented in terms of CLI managed pointers. end note] Given an instance v of a value type V, v cannot be used as the object of a reference initialization if the reference is to a base class of V. (That is, v cannot reference bind to System::Object%, to System::ValueType%, or to any reference to an interface that V implements.) [Rationale: The reason for this is that such a reference binding would require boxing, yet binding a reference to a boxed value rather than to the original value defeats the purpose of reference binding. end rationale] For metadata details, see 34.2.1.
53
12.3.6 Interior pointers The garbage collector is permitted to move objects that reside on the CLI heap. In order for a pointer to refer correctly to such an object, the runtime needs to update that pointer to the objects new location. An interior pointer (declared using interior_ptr) is a pointer that is updated in this manner. For metadata details, see 34.2.2. 12.3.6.1 Definitions The compiler processes an interior pointer as follows: The compiler performs a lookup in the current context for the name interior_ptr. If the name refers unambiguously to ::cli::interior_ptr, or the name is not found, then the expression is processed by the compiler according to the following grammar, and interpreted according to the rules specified herein.
interior_ptr < type-id >
An interior pointer shall have an implicit or explicit auto storage-class-specifier. An interior pointer can be used as a parameter and return type. An interior pointer shall not be a class member or a base class. The default initial value for an interior pointer shall be nullptr. 12.3.6.2 Target type restrictions In the expression interior_ptr<T>, the target type T shall be a cv-qualified value class type, a cvqualified handle type, a cv-qualified native class type, or a cv-qualified native pointer. A program containing other target types is ill-formed. [Example:
interior_ptr<int> p1; interior_ptr<int*> p2 = nullptr; interior_ptr<System::String> p3; interior_ptr<System::String^> p4; interior_ptr<interior_ptr<int>> p5; interior_ptr<int^> p6 = nullptr; // // // // // // OK OK error, String is a ref class OK; is a handle to ref class error, not a native pointer OK
end example] 12.3.6.3 Operations An interior pointer can be involved in the same set of operations as native pointers, as defined by the C++ Standard. [Note: This includes comparison and pointer arithmetic. end note] 12.3.6.4 Data access An interior pointer exhibits the usual pointer semantics for data access: [Example:
value struct V { int data; }; V v; interior_ptr<V> pv = &v; pv->data = 42; interior_ptr<int> pi = &v.data; assert(*pi == 42);
Operator -> is used to access a member of a CLI heap-based object pointed to by an interior pointer; Operator * is used to dereference an interior pointer.
54
Types
end example] Taking the address of an interior pointer yields a native pointer. Interior pointers can point to objects inside the CLI heap. As such, taking the address of an object pointed to by an interior pointer yields an interior pointer that cannot be converted to T*. [Example:
value struct V { int data; }; V v; interior_ptr<V> pv = &v; V** p = &pv; interior_ptr<V>* pi = &pv; int* p2 = &(pv->data); int* p3 = &(v.data);
// // // //
error OK, pv is on the stack and so is an lvalue error OK, v is on the stack, v.data is an lvalue
end example] 12.3.6.5 The this pointer In the body of a non-static member-function of a value class V, this is an rvalue expression of type interior_ptr<V>, whose value is the address of the CLI heap-based object for which the function is called. [Example:
value struct V { int data; void f(); }; void V::f() { interior_ptr<V> pv1 = this; V* pv2 = this; } // OK // error
end example] 12.3.7 Pinning pointers Ordinarily, the garbage collector is permitted to move objects that reside on the CLI heap. However, such movement can be blocked temporarily, on a per object basis. A pinning pointer (declared using pin_ptr) is a pointer that prevents the garbage collector from moving the CLI heap-based object to which that pointer points. This makes it possible for code not under the control of the runtime to manipulate memory within the bounds of the CLI heap without corrupting that heap. Although a pinning pointer can be initialized from an interior pointer, the value of a pinning pointer is never changed by the runtime. A pinning pointer can point to an object anywhere in memory; it need not point to an object on the CLI heap. For metadata details, see 34.2.3. 12.3.7.1 Definitions The compiler processes a pinning pointer as follows: The compiler performs a lookup in the current context for the name pin_ptr. If the name refers unambiguously to ::cli::pin_ptr, or the name is not found, then the expression is processed by the compiler according to the following grammar, and interpreted according to the rules specified herein.
pin_ptr < type-id >
55
A pinning pointer is an interior pointer that is a handle to type type-specifier; it is a type-id. A pinning pointer shall have an implicit or explicit auto storage-class-specifier. A pinning pointer shall not be used as a parameter type or return type. [Note: As a pinning pointer is an interior pointer, the default initial value for a pinning pointer is nullptr. (12.3.6.1) end note] 12.3.7.2 Target type restrictions The target type restrictions for pinning pointers are the same as for interior pointers (12.3.6.2). 12.3.7.3 Operations The operations that can be formed on pinning pointers are the same as for interior pointers (12.3.6.3) except that a pinning pointer cannot be the target of a cast. 12.3.7.4 Data access With two exceptions, pinning pointers follow the same data access semantic as interior pointers (12.3.6.4). Since a pinning pointer points to an unmovable object inside the CLI heap, pin_ptr<T> can be converted to T*. Dereferencing a pinning pointer yields an lvalue. [Example:
value struct V { int data; void f(); }; void V::f() { int* pi; interior_ptr<V> ipv = this; pi = &(ipv->data); pin_ptr<V> ppv = this; pi = &(ppv->data); V* pv; pv = ipv; pv = ppv; } V v; pin_ptr<V> pv = &v; V** p = &pv; int* pi = &pv->data;
// error // OK // error // OK
// error // OK
end example] 12.3.7.5 Duration of pinning As soon as a pinning pointer is initialized or assigned the address of a CLI heap-based object, that object is guaranteed to remain at its location. If the pinning pointer is then made to point to another CLI heap-based object, that object is guaranteed to remain at its location, and the object previously pointed to is no longer pinned by that pointer, allowing it to be moved. If a pinning pointer is assigned the value nullptr, the object previously pointed to (if any) is no longer considered pinned When the block in which a pinning pointer is defined exits, any CLI heap-based object pointed to by that pinning pointer is no longer considered pinned by that pinning pointer; however, it might still be pinned by another pinning pointer. With the exception of the functionality provided by the class System::Runtime::InteropServices::GCHandle, if no pinning pointer points to a CLI heap-based object, it is not safe to assume that object is pinned. [Example:
56
Types ref struct R { int data; }; R^ r = gcnew R; { pin_ptr<int> ppi = &r->data; } // ppis parent block has exited, so object is free to move // object referenced by r is pinned
end example] 12.3.8 Native arrays A program that contains a native array of elements having CLI class type or handle type, is ill-formed. [Note: Allowing elements of such types would make the array type a mixed type (23). end note] A native array type is local to its parent assembly (i.e., it is internal), and that type is not verifiable. Thus, a virtual function taking a native array type as a parameter cannot be overridden from another assembly. For metadata details, see 34.2.4.
12.4 Top-level type visibility

A non-nested class, interface, delegate, or enum definition can optionally specify the visibility of the class, interface, delegate, or enum: top-level-visibility:
public private
The public top-level-visibility specifier indicates that the non-nested class, interface, delegate, or enum is visible outside its parent assembly. Conversely, the private top-level-visibility specifier indicates that the class, interface, delegate, or enum is not visible outside its parent assembly. However, private types are visible within their parent assembly. The default visibility for a class, interface, delegate, or enum is private. [Example:
public class VisibleClass {}; private class InternalClass {}; // visible outside the assembly // visible only within the assembly
end example] Those class, interface, delegate, or enum definitions nested within another type definition have the accessibility specified within that type. The use of a top-level-visibility modifier on a nested type definition causes the program to be ill-formed.
57
13. Variables
This part of this clause is informative. In Standard C++, the term variable is used to designate a named object (C++ Standard 3/4, "Basic concepts"): A name is a use of an identifier (2.10) that denotes an entity or label (6.6.4, 6.1). A variable is introduced by the declaration of an object. The variable's name denotes the object. In Standard C++, the term object refers to a region of data storage. (C++ Standard 1.8/1, "The C++ object model "): The constructs in a C++ program create, destroy, refer to, access, and manipulate objects. An object is a region of storage. [Note: A function is not an object, regardless of whether or not it occupies storage in the way that objects do.] The term CLI object refers to any instance of CLI class type. The term native object refers to an instance of a native class. End of informative text.
13.1 gc-lvalues
In Standard C++, every expression is either an lvalue or rvalue. In C++/CLI, an expression can also be a gclvalue, which refers to an object that might be tracked by the garbage collector. Except where noted below, expectations for lvalues and rvalues based on Standard C++, are unchanged. In C++/CLI, every expression is either an lvalue, a gc-lvalue, or rvalue. Some built-in operators yield gc-lvalues. [Example: If E is an expression of type "handle to type", then *E is a gc-lvalue. As the function int% f(); yields a gc-lvalue, the call f() is a gc-lvalue. end example] Some operators produce results that depend on whether the operand is an lvalue or gc-lvalue. [Example: One such operator is unary &. end example] The result of calling a function returning a tracking reference, is a gc-lvalue, unless the tracking reference refers to a native class. Whenever an lvalue appears in a context where a gc-lvalue is expected, the lvalue is converted to a gclvalue. Likewise, whenever a gc-lvalue appears in a context where an rvalue is expected, the gc-lvalue is converted to an rvalue. Reference initialization and temporaries shall have semantics that make allowance for gc-lvalues, as well as lvalues and rvalues. Like an lvalue, a gc-lvalue can have any complete type, the void type, or an incomplete type. Like with an lvalue, to modify an object, a gc-lvalue for that object shall be used. A program that attempts to modify an object through a nonmodifiable gc-lvalue is ill-formed. The list of restrictions in the C++ Standard (3.10/15) for accessing the stored value of an object through an lvalue also applies to gc-lvalues. 13.1.1 Standard conversions The C++ Standard (4.1) is augmented by the following:
58
Variables
Any lvalue can be converted to a gc-lvalue. A gc-lvalue can convert to an rvalue in exactly the same cases as a conversion from lvalue to an rvalue. A program that necessitates any other lvalue to gclvalue or gc-lvalue to rvalue conversion is ill-formed. 13.1.2 Expressions The C++ Standard (5/6) is augmented by the following: If an expression initially has the type reference to T (8.3.2, 8.5.3), the type is adjusted to T prior to any further analysis, the expression designates the object or function denoted by the reference, and the expression is an lvalue. If an expression initially has the type tracking reference to T, the type is adjusted to T prior to any further analysis, the expression designates the object or function denoted by the reference, and the expression is a gc-lvalue. In general, in any context in which this clause determines the result of an expression is an lvalue because the resulting entity is a function, variable, or data member, it is an lvalue only if the entity is a function, or it is a variable or data member that is not on the CLI heap. If the entity is a variable or data member that is, or could be, on the CLI heap, the result is a gc-lvalue. This applies to cases mentioned in the C++ Standard, 5.1/4, 5.1/7, and 5.1/8. An entity of an expression might not always be on the CLI heap, but it might be. [Example: A member function of a value class referring to a data member of that value class shall assume that the class is allocated on the CLI heap, and is, therefore, a gc-lvalue. end example] The C++ Standard (5.2.2/10) is augmented as follows: A function call is an lvalue if and only if the result type is a native reference. A function call is a gclvalue if and only if the result type is a tracking reference. The C++ Standard (5.2.5/4) is augmented as follows: If E2 is a member enumerator, and the type of E2 is T, the expression E1.E2 isnot an lvalue an rvalue. The type of E1.E2 is T. The following rules have been added to the requirements of the C++ Standard (5.2.5/4): If E2 is a static data member of a ref class or value class, and the type of E2 is T, then E1.E2 is a gc-lvalue; the expression designates the named member of the class. The type of E1.E2 is T. If E2 is a non-static data member, the expression designates the named member of the object designated by the first expression. If E1 is a gc-lvalue, then E1.E2 is a gc-lvalue. The C++ Standard (5.2.6/1) is augmented as follows: The operand shall be a modifiable gc-lvalue. The C++ Standard (5.3.1/1) is augmented as follows: The unary * operator performs indirection: the expression to which it is applied shall be a pointer or handle to an object type, or a pointer to a function type. and tThe result of applying indirection to a pointer is an lvalue referring to the object or function to which the expression points. The result of applying indirection to a handle is a gc-lvalue referring to the object. If the type of the expression is pointer to T, the type of the result is T. If the type of the expression is T^, the type of the result is T. [Note: a pointer to an incomplete type (other than cv void) can be dereferenced. The lvalue thus obtained can be used in limited ways (to initialize a reference, for example); this lvalue shall not be converted to an rvalue, see 4.1. ] The C++ Standard (5.3.1/2) is augmented as follows: The result of the unary & operator is a pointer to its operand. The operand shall be an lvalue, gclvalue, or a qualified-id. If the operand is an lvalue, given the type of the expression is T, the result is an rvalue and its type is pointer to T. If the operand is a gc-lvalue, given the type of the expression is T, the result is an rvalue and its type is interior_ptr to T.In the first case, if the type of the expression is T, the type of the result is pointer to T. In particular, the address of an object of type cv T is pointer to cv T, with the same cv-qualifiers. For a qualified-id, if the
59
member is a static member of type T, the type of the result is plain pointer to T. If the member is a nonstatic member of class C of type T, the type of the result is pointer to member of class C of type T. The C++ Standard (5.3.2/1) is augmented as follows: The operand shall be a modifiable gc-lvalue. The primary list in the C++ Standard (5.16/3) is augmented by the following: If E2 is a gc-lvalue, E1 can be converted to match E2 if E1 can be implicitly converted to the type tracking reference to T2, subject to the constraint that in the conversion the reference shall bind directly to E1. The C++ Standard (5.16/4) is augmented by the following: If the second and third operands are lvalues and have the same type, the result is of that type and is an lvalue. If the second and third operands are gc-lvalues and have the same type, the result is of that type and is a gc-lvalue. The C++ Standard (5.17/1) is augmented as follows: There are several assignment operators, all of which group right-to-left. All require a modifiable gclvalue or lvalue as their left operand, and the type of an assignment expression is that of its left operand. The result of the assignment operation is the value stored in the left operand after the assignment has taken place; the result is an lvalue. The result of an assignment operator is an lvalue if the left operand was an lvalue. Likewise, the result of an assignment operator is a gc-lvalue if the left operand was a gc-lvalue. The C++ Standard (5.18/1) is augmented by the following: The type and value of the result are the type and value of the right operand; the result is an lvalue if its right operand is. The result is a gc-lvalue if its right operand is a gc-lvalue. 13.1.3 Reference initializers The C++ Standard (8.5.3) is augmented by the following: A native reference cannot bind to a gc-lvalue. If a native reference is bound to an rvalue, a temporary of the initializer expression shall be created (as described in Standard C++ 8.5.3/5). The temporary shall be allocated in memory not under control of the CLI heap. A tracking reference can bind to an lvalue or a gc-lvalue. Unlike native references, a tracking reference need not be const to bind to an rvalue. That is, int% r = 42; is well-formed. Binding of tracking references otherwise follows the same rules as native references. A native reference expression is always considered an lvalue. A tracking reference expression is always considered a gc-lvalue, except when the tracking reference refers to a native class, in which case, it is an lvalue. 13.1.4 Temporary objects The C++ Standard (12.2) is augmented by the following: A temporary object is an rvalue, which shall not be allocated on the native heap.
13.2 File-scope and namespace-scope variables

For metadata details, see 34.3.1.
13.3 Direct initialization

Direct initialization in the C++ Standard (8.5) occurs in new expressions, static_cast expressions, functional notation type conversions, and base and member initializers. Direct initialization considers both
60
Variables
constructors and user-defined conversion functions. C++/CLI makes a distinction amongst these different forms of direct initialization for CLI class types and limits usage of constructors and conversion functions to specific cases. If the initialization is taking place in a new expression and the destination type is a CLI class type, only constructors of the destination type are considered. [Note: Such a new expression, will only use the gcnew form of the grammar. end note] The C++ Standard (8.5/14) is augmented for this case to remove any reference to conversion functions. If the initialization is taking place in a static_cast expression and the destination type is a CLI class type, only conversion functions of both the source type and destination type are considered. The C++ Standard (8.5/14) is augmented for this case to remove any reference to constructors. If the initialization is taking place in a functional notation type conversion and the destination type is a CLI class type, only constructors of the destination type are considered. The C++ Standard (8.5/14) is augmented for this case to remove any reference to conversion functions. This is further described in 15.3.3. If the initialization is taking place in base or member initializer and the destination type is a CLI class type, only constructors of the destination type are considered. The C++ Standard (8.5/14) is augmented for this case to remove any reference to conversion functions.
61
14. Conversions
14.1 Conversion sequences
To accommodate the addition of boxing conversions and parameter array conversions, 13.3.3.2 of the C++ Standard is augmented, as follows: When comparing the basic forms of implicit conversion sequences (as defined in 13.3.3.1) a standard conversion sequence (13.3.3.1.1) is a better conversion sequence than a boxing conversion sequence, a user-defined conversion sequence, a parameter array conversion sequence, or an ellipsis conversion sequence, and a boxing conversion sequence is a better conversion sequence than a user-defined conversion sequence, a parameter array conversion sequence, or an ellipsis conversion sequence, and a user-defined conversion sequence (13.3.3.1.2) is a better conversion sequence than a parameter array conversion sequence or an ellipsis conversion sequence (13.3.3.1.3). a parameter array conversion sequence is a better conversion sequence than an ellipsis conversion sequence (13.3.3.1.3).
14.2 Standard conversions

The standard conversions in the C++ Standard apply to C++/CLI. C++/CLI has the following standard conversions as well. 14.2.1 Handle conversions A handle conversion is similar to a pointer conversion as defined in the C++ Standard (4.10). To accommodate the addition of handle conversions, Table 9, "conversions", in the C++ Standard, 13.3.3.1.1, "Standard conversion sequences", is augmented by the addition of a "Handle conversion" row, as shown in 18.3. An rvalue of type handle to cv D, where D is a type, can be converted to an rvalue of type handle to cv B, where B is a base class of D. The result of the conversion is a handle to the same object. Since the type void^ is ill-formed, there is no handle conversion to it. A handle to a type array<S^,n> has a handle conversion to a handle to type array<T^,n> provided S^ has a handle conversion to T^ and n (the rank of both CLI arrays) is the same. Such a conversion is better than a conversion from type array<S^,n> to System::Array^. This relationship is known as array covariance. Because array covariance can allow a variable to refer to a base class of the arrays element type, assignments to elements of handle type arrays include a run-time check performed by the CLI (see CLI Partion III, 4.26 and 4.27). The run-time check ensures that the value being assigned to the array element is of a permitted type. Array covariance specifically does not extend to CLI arrays of value types. For example, no conversion permits an array<int> to be treated as array<Object^>. A handle can be used as the first operand of a conditional operator. The null value constant can be converted to any handle type; the result is a handle with null value of that type, and is distinguishable from every other value that is a handle to an CLI heap-based object. To support this, the C++ Standard is augmented, as follows: 4/2: [Note: When used in the condition of an if statement or iteration statement (6.4, 6.5). If the condition is a handle, and conversion from the handle to bool is not possible, the destination
62
Conversions
type is the handle type; otherwise, the destination type is bool. If the condition is not a handle type, the destination type is bool. end note] 5.16/1: The first expression is implicitly converted to bool (clause 4). If that conversion is illformed and the expression is a handle type or a type given by a generic type parameter not constrained by the value class constraint, the expression is tested for the null value, returning true if not null and false if it is null. Otherwise, if the conversion to bool is ill-formed and the expression is not a handle type or a type given by a generic type parameter not constrained by the value class constraint, the program is ill-formed. 6.4/4: The value of a condition that is an initialized declaration in a statement other than a switch statement is the value of the declared variable implicitly converted to type bool. If that conversion is ill-formed, the program is ill-formed. The value of a condition that is an initialized declaration in a switch statement is the value of the declared variable if it has integral or enumeration type, or of that variable implicitly converted to integral or enumeration type otherwise. The value of a condition that is an expression is the value of the expression, implicitly converted to bool for statements other than switch; if that conversion is ill-formed, the program is ill-formed. The value of the condition will be referred to as simply the condition where the usage is unambiguous. The value of a condition that is an expression is the value of the expression, implicitly converted to bool for statements other than switch. If that conversion is ill-formed and the expression is a handle type or a type given by a generic type parameter not constrained by the value class constraint, the expression is tested for the null value, returning true if not null and false if it is null. Otherwise, if the conversion to bool is illformed and the expression is not a handle type or a type given by a generic type parameter not constrained by the value class constraint, the program is ill-formed. [Note: If there is no conversion to bool and the declared variable or expression is not a handle type, a conversion to a handle type is not considered. end note.] 6.5.2/1: The expression is implicitly converted to bool; if that is not possible, and the expression is a handle type or a type given by a generic type parameter not constrained by the value class constraint, it is tested for null. If there is no conversion to bool, and the expression is not a handle type or a type given by a generic type parameter not constrained by the value class constraint, the program is ill-formed. 14.2.1.1 Ranking handle conversions Of the additional standard conversion C++/CLI adds, only handle conversions can require further ranking to determine whether one conversion is better than another. In addition to the rules in the C++ Standard 13.3.3.2/4, the following rules apply: If class B is derived directly or indirectly from class A and class C is derived directly or indirectly from B, o o Conversion of C^ to B^ is better than conversion of C^ to A^. Conversion of B^ to A^ is better than conversion of C^ to A^.
14.2.2 Pointer conversions The definition of null pointer constant in the C++ Standard (4.10/1) is augmented, as follows: A null pointer constant is either an integral constant expression rvalue of integer type that evaluates to zero, or the null value constant nullptr. [Note: The implication of this is that the null value constant can be converted to any pointer type. end note] The following conversion rules apply to interior pointers: Conversion from interior_ptr<T1> to interior_ptr<T2> is allowed if and only if conversion from T1* to T2* is allowed; In conversions between types where exactly one type is interior_ptr<T1>, the interior pointer behaves exactly as if it were pointer to cv T1, with two exceptions: 63
[Example:
Conversion to any other type pointer to cv T1 is not allowed. In particular, conversion from interior_ptr<T> to T* is not allowed. Conversion from the null pointer constant to interior_ptr<T> is not allowed, but conversion from the null value constant is allowed.
array<int>^ arr = gcnew array<int>(100); interior_ptr<int> ipi = &arr[0]; int* p = ipi; // error; no conversion from interior to noninterior int k = 10; ipi = &k; // OK; k is an auto variable ipi = 0; // error; must use nullptr instead ipi = nullptr; // OK ipi = p; // OK if (ipi) { } // OK
end example] The following conversion rules apply to pinning pointers: Conversion from pin_ptr<T1> to pin_ptr<T2> is allowed if and only if conversion from T1* to T2* is allowed; In conversions between types where exactly one type is cv pin_ptr<T>, the pinning pointer behaves exactly as if it were pointer to cv T, with the exception that conversion from a null pointer constant to pin_ptr<T> is not allowed, but conversion from the null value constant is allowed. [Note: In particular, conversion from pin_ptr<T> to T* is allowed as a standard conversion. end note] [Example:
array<int>^ arr = gcnew array<int>(100); pin_ptr<int> ppi = &arr[0]; int* p = ppi; // OK int k = 10; ppi = &k; // OK; k is an auto variable ppi = 0; // error; must use nullptr instead ppi = nullptr; // OK pin_ptr<int> ppi2 = p; // OK
end example] 14.2.3 Lvalue conversions There is a standard conversion for each of the following: cv-qualified lvalue of type T to cv-qualified gclvalue of type T, and cv-qualified gc-lvalue of type T to cv-qualified rvalue of type T. If a cv-qualified lvalue would not convert to an rvalue in a given context, it is ill-formed for a gc-lvalue to convert to an rvalue. [Rationale: Conversion from a gc-lvalue to an rvalue when binding a native reference to an integer on the CLI heap results in loss of type safety. end rationale] 14.2.4 Integral promotions To accommodate the addition of extended integer types, the C++ Standard (4.5/1) is is augmented, as follows: An rvalue of type char, signed char, unsigned char, short int, or unsigned short int an integer type whose integer conversion rank (4.13) is less than the rank of int and unsigned int can be converted to an rvalue of type int if int can represent all the values of the source type; otherwise, the source rvalue can be converted to an rvalue of type unsigned int. and the C++ Standard is augmented by the following new clause, 4.13: 4.13 Integer conversion rank Every integer type has an integer conversion rank defined as follows:
64
Conversions
No two signed integer types shall have the same rank, even if they have the same representation. The rank of a signed integer type shall be greater than the rank of any signed integer type with less precision. The rank of long long int shall be greater than the rank of long int, which shall be greater than the rank of int, which shall be greater than the rank of short int, which shall be greater than the rank of signed char. The rank of any unsigned integer type shall equal the rank of the corresponding signed integer type, if any. The rank of any standard integer type shall be greater than the rank of any extended integer type with the same width. The rank of char shall equal the rank of signed char and unsigned char. The rank of bool shall be less than the rank of all other standard integer types. The rank of any enumerated type shall equal the rank of its underlying type (7.2). The rank of any extended signed integer type relative to another extended signed integer type with the same precision is implementation-defined, but still subject to the other rules for determining the integer conversion rank. For all integer types T1, T2, and T3, if T1 has greater rank than T2 and T2 has greater rank than T3, then T1 has greater rank than T3. [Note: The integer conversion rank is used in the definition of the integral promotions (4.5) and the usual arithmetic conversions (5).] To accommodate the addition of the types long long int and unsigned long long int, the C++ Standard (4.5/2) is augmented, as follows: An rvalue of type wchar_t (3.9.1) or System::Char can be converted to an rvalue of the first of the following types that can represent all the values of its underlying type: int, unsigned int, long, or unsigned long, long long int, or unsigned long long int. An rvalue of an enumeration type (7.2) can be converted to an rvalue of the first of the following types that can represent all the values of the enumeration (i.e., the values in the range bmin to bmax as described in 7.2): int, unsigned int, long, or unsigned long, long long int, or unsigned long long int. 14.2.5 String literal conversions An rvalue of type <narrow-string-literal-type> can be converted to one of two types: System::String^ or array of n const char. When a <narrow-string-literal-type> is converted to System::String^, the result is treated as a CLI string literal (34.4.1). When a <narrow-string-literal-type> is converted to an array, n is the size of the string (as defined in the C++ Standard, 2.13.4/5), the array has static storage duration, and the array is initialized with the given characters. A conversion from <narrow-string-literaltype> to System::String^ is better than a conversion from <narrow-string-literal-type> to array of n const char. An rvalue of type <wide-string-literal-type> can be converted to one of two types: System::String^ or array of n const wchar_t. When a <wide-string-literal-type> is converted to System::String^, the result is treated as a CLI string literal (34.4.1). When a <wide-string-literal-type> is converted to an array, n is the size of the string (as defined in the C++ Standard, 2.13.4/5), the array has static storage duration, and the array is initialized with the given characters. A conversion from <wide-string-literal-type> to System::String^ is better than a conversion from <wide-string-literal-type> to array of n const wchar_t. For conversion in the presence of the subscript operator, see 15.3.1; for the unary * operator, see 15.4.1.2; for the binary -> operator, see 15.3.4; and with the binary + operator, see 15.6.3.
65
Consider the case in which a function, whose parameter-declaration-clause terminates with an ellipsis, is called with a string literal as the argument that corresponds to the ellipsis. If the string literal is a narrow string literal, it is converted to an array of n char; if it is a wide string literal, it is converted to an array of n wchar_t. 14.2.6 Boxing conversions A boxing conversion involves the creation of a new object on the CLI heap. A boxing conversion shall be applied only to instances of value types, with the exception of pointers. For any given value type V, the conversion results in a V^. [Note: Boxing in some other CLI-based languages goes directly from V to Object^. This can be achieved in C++/CLI via a boxing conversion followed by a handle conversion. end note] Although the value type expression can be cv-qualified, the resulting boxed value type is not. To accommodate the addition of boxing conversions, Table 9, "conversions", in the C++ Standard, 13.3.3.1.1, "Standard conversion sequences", is augmented by the addition of a "Boxing conversion" row, as shown in 18.3. [Example: Note that the positioning of the boxing conversion in that table means that given a choice between a narrowing conversion and boxing, boxing is preferred. Given the following,
void F(float f) { Console::WriteLine("F(float)"); } void F(Object^ o) { Console::WriteLine("F(Object^)"); } int main() { F(3.14); }
the output is "F(Object^)". end example] A boxing conversion cannot be rewritten by the user; it is reserved to the implementation. A boxing conversion follows the exact same sequence of operations as user-defined conversions (C++ Standard 13.3.3.1.2). Boxing conversions are considered before user-defined conversions, and a boxing conversion sequence never invokes a user-defined conversion. In other words, given a choice between applying a boxing conversion or a user-defined conversion, the boxing conversion is selected. Thus, 13.3.3.2 of the C++ Standard is augmented, as shown in 14.1 . [Note: One can write a user-defined conversion operator that performs the same conversion as a boxing conversion. Although the compiler would not call this user-defined conversion in boxing contexts, the programmer could call the user -defined conversion using explicit operator function syntax. end note] For metadata details, see 34.4.2.
14.3 Implicit conversions

14.3.1 Implicit constant expression conversions The following implicit constant expression conversions are permitted: The null value constant can be converted to any pointer type. The null value constant can be converted to any handle type.
14.3.2 User-defined implicit conversions 14.3.3 Boolean Equivalence Whether or not bool maps to System::Boolean, an rvalue of type bool can be converted to an rvalue of type System::Boolean, and an rvalue of type System::Boolean can be converted to an rvalue of type bool.
66
Conversions
14.4 Explicit conversions

The following explicit conversions are permitted: The null value constant can be converted to any pointer type. The null value constant can be converted to any handle type.
14.5 User-defined conversions

Generic conversion functions are allowed. [Note: However, the need to check generic constraints after overload resolution makes it difficult to write a generic conversion that is useful. A template conversion function will usually be more useful. end note] 14.5.1 Constructors Although the explicit keyword is permitted on a constructor in a ref class or value class, it has no effect. Constructors in these classes are never used for conversions or casts (see 13.3). 14.5.2 Explicit conversion functions C++/CLI allows the explicit keyword on conversion functions. Thus, C++ Standard 7.1.2 is augmented, as follows: The explicit specifier shall be used only in declarations of constructors within a class declaration, or on declarations of conversion functions within a class declaration; see 12.3.1 and 12.3.2. A conversion function that is declared with the explicit keyword is known as an explicit conversion function. A conversion function that is declared without the explicit keyword (i.e., every conversion function in Standard C++) is known as an implicit conversion function. Like an explicit constructor, an explicit conversion function can only be invoked by direct-initialization syntax (C++ Standard 8.5) and casts (C++ Standard 5.2.9, 5.4). A type shall not contain an implicit conversion function and an explicit conversion function that perform the same conversion. Only one of these is allowed. It is possible to write a class that has both an explicit converting constructor and a conversion function that can perform the same conversion. In this case, the explicit conversion function is preferred. 14.5.3 Static conversion functions C++/CLI allows conversion functions, both implicit and explicit, to be static. Conversion functions shall not have namespace scope. A static conversion function shall take only one parameter, which is the type to convert from (a non-static member conversion function shall have no parameters). Neither static nor nonstatic conversion functions shall specify return types. Either the source type (parameter type) or the target type (type-specifier-seq) is required to be T, T^, T&, T%, T^%, or T^&, where T is the type of the containing class. (T* is not allowed because conversions are not looked up through pointers.) Implicit conversions can now be found in more than one place: the scope of the type of the source expression and the scope of all potential target types. If overload resolution results in a set of conversion functions (and possibly converting constructors) that can perform the same conversion, the program is ambiguous and ill-formed.
14.6 Parameter array conversions

The parameter array conversion sequence occurs when overload resolution chooses a function that takes a parameter array as its last argument. Such overloads are preferred to C-style variable-argument functions, and are not preferred to any other overloads.
67
A parameter array overload is chosen by overload resolution. For the purpose of overload resolution, the compiler creates signatures for the parameter array functions by replacing the parameter array argument with n arguments of the CLI arrays element type, where n matches the number of arguments in the function call. These synthesized signatures have higher cost than other non-synthesized signatures, and they have lower cost than functions whose parameter-declaration-clause terminates with an ellipsis. [Note: This is similar to the tiebreaker rules for template functions and non-template functions in the C++ Standard (13.3.3). end note] For example, for the function call f(var1, var2, , varm, val1, val2, , valn)
void f(T1 arg1, T2 arg2, , Tm argm, ... array<T>^ arr)
is replaced with
void f(T1 arg1, T2 arg2, , Tm argm, T t1, T t2, , T tn)
Overload resolution is performed with the set containing the synthesized signatures according to the rules of Standard C++. If overload resolution selects a C-style variable-argument conversion, it means that none of the synthesized signatures was chosen. If overload resolution selects one of the synthesized signatures, the conversion sequences needed for each argument to satisfy the call is performed. For the synthesized parameter array arguments, the compiler constructs a CLI array of length n and initializes it with the converted values. Then the function call is made with the constructed parameter array. [Note: User-defined conversions are better than parameter array conversions.
ref class A {}; ref class B { public: static operator A^(B^ b) { return gcnew A; } }; void F(... array<B^>^ arr) { Console::WriteLine("array<B^>^"); } void F(A^ a) { Console::WriteLine("A^"); } int main() { B^ b = gcnew B; F(b); }
The program prints A^. end note]
14.7 Naming conventions

During compilation, the name of the conversion function is the C++ identifier used in source code for that function. For example, the conversion function from A to B could be the static member function of either A or B, operator B(A), or the instance function of A, operator B(). [Example:
public value struct Decimal { static operator Decimal(int value); static explicit operator double(Decimal value); }; explicit operator float();
end example] A program that declares or defines a member function within a ref class, value class, or interface class using the names op_Implicit or op_Explicit, is ill-formed. A program shall not directly refer to these names. Operator functions are either CLS-compliant or C++-dependent. A conversion function is CLS-compliant when all of the following conditions occur: The conversion function is a static member of a ref class or a value class.
68
Conversions
If a value class is a parameter or a target value of the conversion function, the value class shall not be passed by reference nor passed by pointer or handle. If a ref class is a parameter or a target value of the operator function, the ref class shall be passed by handle. The handle shall not be passed by reference.
If a conversion function does not match these criteria, it is C++-dependent.
69
15. Expressions
To accommodate the addition of the types long long int and unsigned long long int, and extended integer types, the C++ Standard (5/9) is augmented as follows: Many binary operators that expect operands of arithmetic or enumeration type cause conversions and yield result types in a similar way. The purpose is to yield a common type, which is also the type of the result. This pattern is called the usual arithmetic conversions, which are defined as follows: If either operand is of type long double, the other shall be converted to long double. Otherwise, if either operand is double, the other shall be converted to double. Otherwise, if either operand is float, the other shall be converted to float. Otherwise, the integral promotions (4.5) shall be performed on both operands. Then, if either operand is unsigned long the other shall be converted to unsigned long. Otherwise, if one operand is a long int and the other unsigned int, then if a long int can represent all the values of an unsigned int, the unsigned int shall be converted to a long int; otherwise both operands shall be converted to unsigned long int. Otherwise, if either operand is long, the other shall be converted to long. Otherwise, if either operand is unsigned, the other shall be converted to unsigned. [Note: otherwise, the only remaining case is that both operands are int ] Otherwise, the integer promotions are performed on both operands. Then the following rules are applied to the promoted operands: If both operands have the same type, then no further conversion is needed. Otherwise, if both operands have signed integer types or both have unsigned integer types, the operand with the type of lesser integer conversion rank is converted to the type of the operand with greater rank. Otherwise, if the operand that has unsigned integer type has rank greater or equal to the rank of the type of the other operand, then the operand with signed integer type is converted to the type of the operand with unsigned integer type. Otherwise, if the type of the operand with signed integer type can represent all of the values of the type of the operand with unsigned integer type, then the operand with unsigned integer type is converted to the type of the operand with signed integer type. Otherwise, both operands are converted to the unsigned integer type corresponding to the type of the operand with signed integer type.
15.1 Function members

The following function member kinds are added to those defined by Standard C++: Properties (both scalar and default-indexed) Events
The statements contained in these function members are executed through function member invocations. The actual syntax for writing a function member invocation depends on the particular function member category.
70
Expressions
Invocations of default-indexed properties employ overload resolution to determine which of a candidate set of function members to invoke. [Note: The following table summarizes the processing that takes place in constructs involving these three categories of function members that can be explicitly invoked. In the table, e, x, y, and value indicate expressions classified as variables or values, E is an event, and P is the simple name of a property. Construct Property access Example Description
P::get() P::set(value) E::add(value) E::remove(value) e.default::get(x, y) e.default::set(x, y, value) P P = value E += value Event access E -= value e[x, y] Default-indexed property access e[x, y] = value
end note]
15.2 Primary expressions

To accommodate the addition of properties, the Primary expressions subclause of the C++ Standard (5.1) is augmented, as follows: A static property or event is not associated with any instance of a class, and a program is ill-formed if it refers to this in the accessor functions of a static property or event. An instance property or event is associated with a specific instance of a class, and that instance can refer to this in the accessor functions of that instance property or event.
15.3 Postfix expressions

To accommodate the addition of default-indexed properties and CLI arrays (which are accessed using subscript-like expressions), the C++ Standard grammar (5.2) for postfix-expression is augmented, as follows: postfix-expression: primary-expression postfix-expression [ expression-list ] postfix-expression ( expression-listopt ) simple-type-specifier ( expression-listopt ) typename ::opt nested-name-specifier identifier ( expression-listopt ) typename ::opt nested-name-specifier templateopt template-id ( expression-listopt ) postfix-expression . templateopt id-expression postfix-expression -> templateopt id-expression postfix-expression . pseudo-destructor-name postfix-expression -> pseudo-destructor-name postfix-expression ++ postfix-expression -dynamic_cast < type-id > ( expression ) static_cast < type-id > ( expression ) reinterpret_cast < type-id > ( expression ) const_cast < type-id > ( expression ) typeid ( expression ) typeid ( type-id ) typenameopt ::opt nested-name-specifier identifier :: typeid typenameopt ::opt nested-name-specifier templateopt template-id :: typeid The C++ Standard production postfix-expression [ expression ]
71
is augmented to postfix-expression [ expression-list ] to accommodate indexed access (15.3.1) and CLI array element access (24.3). As a result, commas in square-bracketed expressions are not operators and instead are list separators. To allow constructs such as List<List<int>>, where >> is treated as two tokens instead of one, the C++ Standard (5.2/2) is augmented by the following new paragraph: [Note: The > token following the type-id in a const_cast, dynamic_cast, reinterpret_cast, safe_cast, or static_cast may be the product of replacing a >> token by two consecutive > tokens (14.2). end note] 15.3.1 Subscripting and indexed access The subscripting operator [] can represent the built-in subscripting operator (C++ Standard 5.2.1), a call of an overloaded operator[] (C++ Standard 13.5.5), or a use of an indexed property. Overload resolution is used to determine which applies. As in the C++ Standard, if neither operand is a class or enum or a handle to a class, overload resolution is not needed and the built-in operator is selected. For any given instance of a ref class, subscripting can be applied to that instance and to a handle to that instance, with the same result. The argument list for the overload resolution is the left operand plus the list of expressions of the expressionlist. [Note: in Standard C++, the syntactic term inside the [] is an expression, which means that X[i,j] is a valid subscripting operation whose subscript is a comma-expression (in other words, it's effectively X[j]). In C++/CLI, a top-level comma inside [] is considered a list separator and not an operator, so X[i,j] would only match an indexed property taking two arguments. If one wants a top-level comma operator, one must write it inside parentheses, e.g., X[(i,j)]. This is true even when X does not have class type or handle to class type. end note] A CLI class type shall not have both a default-indexed property and an operator[].When subscript is applied to a string literal, that literal is converted to an "array of n const char" or "array of n const wchar_t", as appropriate. The following built-in operator functions exist: const char& operator[](<narrow-string-literal-type>, integer-type); const wchar_t& operator[](<wide-string-literal-type>, integer-type); const char& operator[](integer-type, <narrow-string-literal-type>); const wchar_t& operator[](integer-type, <wide-string-literal-type>); where integer-type is any integer type. 15.3.2 Function call The C++ Standard (5.2.2/1) states, A function call is a postfix expression followed by parentheses containing a possibly empty, comma-separated list of expressions, which constitute the arguments to the function. C++/CLI contains support for delegates (27). As such, the postfix expression can be a delegate type, in which case, the whole expression is a delegate invocation (27.3), and the argument list is passed to each function encapsulated by the delegate. 15.3.3 Explicit type conversion (functional notation) Function-style casts of ref classes and value classes do not invoke conversions; these are calls to constructors only. If a corresponding constructor does not exist, the program is ill-formed. [Example:
72
Expressions value class C {}; value class E { public: operator C() { return C(); } }; void F(C c) {} int main() { E e; F(C(e)); // error - no constructor of C matches parameter }
end example] 15.3.4 Class member access To accommodate the use of handles with ->, the text in Standard C++ (5.2.5/2) is augmented, as follows: For the second option (arrow) the type of the first expression (the pointer expression) shall be handle to class object (of a complete type) or pointer to class object (of a complete type). The text in Standard C++ (5.2.5/3) is amended, as follows: If E1 has the type "pointer to class X," then the expression E1->E2 is converted to the equivalent form (*(E1)).E2. If E1 has the type "handle to class X", and X has an operator-> the expression E1->E2 is evaluated as (*(E1)).operator->(E2). Otherwise, if E1 has the type "handle to class X" and X does not have an operator->, then the expression E1->E2 is converted to the equivalent form (*(E1)).E2. and footnote 59 is augmented, as follows: 59) Note that if E1 has the type pointer to class X, then (*(E1)) is an lvalue. If E1 has the type handle to class X, then (*(E1)) is a gc-lvalue. If a program accesses an instance of a value type directly using the arrow operator, it is ill-formed. [Note: Applying the arrow operator to an instance of a value type does not box that value. However, certain accesses to such an instance using the dot operator require boxing. See the metadata details in 34.5.1. end note] When a string literal is the left-hand operand to the binary operator->, that literal is converted to System::String^. 15.3.5 Increment and decrement See 19.7.3. 15.3.6 Dynamic cast For the expression dynamic_cast<T>(e), in addition to the rules specified by the C++ Standard (5.2.7), the following also applies: If T is a tracking reference type, e shall be a gc-lvalue of a complete class type, and the result is a gc-lvalue of the type referred to by T.
T can be a handle type, and in such cases e shall be an rvalue of a handle to complete class type, and the result is an rvalue of type T.
If the value of e is a null value and T is handle type, the result is the null value of type T. If T is handle to cv1 B and e has type handle to cv2 D such that B is a base class of D, the result is a handle to B such that it refers to the same CLI heap-based object as e. The cv-qualification for cv1 shall be the same as or greater than that for cv2. Otherwise, a runtime check is required. If the runtime check cannot succeed, the program is ill-formed.
73
If T is either a handle or a pointer to any type other than a native class, and the cast fails, the result is the null value or the required result type. If T is a reference to any type other than a native class and the cast fails, then the expression throws System::InvalidCastException. When T is a native class, the rules of Standard C++ 5.2.7/9 apply. For metadata details, see 34.5.2. 15.3.7 Type identification C++/CLI adds a new use of the typeid keyword, whereby a given type name can be followed by ::typeid to get a System::Type^ for the given type name. This construct is referred to here as a typeid Type expression (which is unrelated to Standard C++'s typeid expression). To accommodate this, the C++ Standard grammar production for postfix-expression (5.2 and A.4) is augmented (15.3). In the C++ Standard (14.6.2.2/4), the "Expressions of the following forms" list is augmented to include the new typeid Type expression forms of postfix-expression (15.3). The result of a typeid Type expression is an lvalue of static type System::Type^. There is only one System::Type object for any given type. [Note: This means that for any type T, T::typeid == T::typeid is always true. end note] As this form is a compile-time expression, it can be used as an argument to an attribute constructor. The type name in the typeid Type expression shall be a raw type (12.3.1) or a pointer to a raw type. The type in a typeid Type expression can be any handle R^ provided that type is referred to via a typedef. The result of such an expression is the same as applying typeid directly to type R. The type R% is handled the same way. Each fundamental type is a distinct type; however, different fundamental types can map to the same CLI type. As such, the typeid operator shall produce the same Type handle for each fundamental type that maps to the same CLI type, regardless of whether optional or required modifiers (33.1) are otherwise required to distinguish those fundamental types. [Example: In an implementation in which int and long both map to System::Int32, both int::typeid and long::typeid result in a Type^ describing System::Int32. end example] [Note: The practice of using a lock on T::typeid to guard static members of a type T is discouraged, as it can lead to deadlock. end note] The typeid Type expression provides convenient syntactic access to the functionality of the
System::Type::GetType() library function. Whereas GetType() shall be called on an CLI heap-based object of the given type, ::typeid can be applied to a type directly, and consequently does not require a
CLI heap-based object to be created. [Example:

using namespace System::Reflection; ref class X { }; Console::WriteLine(X::typeid); X^ pX = gcnew X; Type^ pType = pX->GetType(); Console::WriteLine(pType); // does not require an object // GetType requires an object
Console::WriteLine(Int32::typeid); Console::WriteLine(array<Int32>::typeid); Console::WriteLine(void::typeid); Type^ t = String::typeid; Console::WriteLine(t->BaseType); array<MethodInfo^>^ functions = t->GetMethods(); for each (MethodInfo^ mi in functions) Console::WriteLine(mi);
74
Expressions X X System.Int32 System.Int32[] System.Void System.Object System.CharEnumerator GetEnumerator() System.Type GetType()
end example] The ::typeid operator can be applied to a type parameter or to a constructed type: the result is a CLI heapbased object of type System::Type that represents the runtime type of the type parameter or constructed type. Outside of the body of a generic type definition, the ::typeid operator shall not be applied to the bare name of that type. [Example:
generic<typename T> ref class X { public: static void F() { Type^ t1 = T::typeid; Type^ t2 = X<T>::typeid; Type^ t3 = X::typeid; } }; int main() { Type^ t4 = int::typeid; Type^ t5 = X<int>::typeid; Type^ t6 = X::typeid; }
// okay // okay // okay
// okay // okay // error
Clearly, the initialization of t6 is in error. However, that of t3 is not, as the use of X is really an implicit use of X<T> (31.1.2). end example] The ::typeid operator can be used in an argument to an attribute constructor call. [Example:
[AttributeUsage(AttributeTargets::All)] public ref struct XAttribute : Attribute { XAttribute(Type^ t) {} }; [X(int::typeid)] public ref class R {};
end example] Standard C++'s native typeid can be applied to expression or type-id. Native typeid shall not be used with types that are ref classes, interface classes, handles, value classes other than fundamental types, enums of any kind, or pointers. Thus, any program that contains a native typeid with expression or type-id having any of these types, is ill-formed. 15.3.8 Static cast The rules specified by the C++ Standard (5.2.9) apply. For the expression, static_cast<T>(e), the following also applies. A static cast can invoke a user-defined conversion function as described in the C++ Standard (5.2.9/2). All of the following are considered: explicit conversion functions, implicit conversion functions, explicit converting constructors, and implicit converting constructors. [Note: Non-native types do not have converting constructors. end note] The cast expression discussed in the C++ Standard (5.2.9/3) is also allowed on tracking references. The conversion discussed in the C++ Standard (5.2.9/7) is allowed for both native and CLI enumerations.
75
An rvalue of type handle to cv1 B, where B is a type, can be converted to an rvalue of type handle to cv2 D, where D is a class derived from B, if a valid standard conversion from handle to D to handle to B exists (14.2.1), and cv2 is the same cv-qualification as, or greater cv-qualification than, cv1. The null value is converted to the null value of the destination type. 15.3.9 Reinterpret cast The rules of specified by the C++ Standard (5.2.10) apply. A reinterpret cast expression that attempts to cast from or to a handle type is ill-formed. A reinterpret cast will never invoke a boxing conversion sequence. 15.3.10 Const cast The rules specified by the C++ Standard (5.2.11) apply. For the expression, const_cast<T>(v), the following also applies. Where the C++ Standard discusses the application of const_cast to pointers, the rules shall also apply to handles. An lvalue of type T1 can be explicitly converted to an lvalue of type T2 using the cast const_cast<T2%> if a pointer or handle to T1 can be explicitly converted to the type pointer or handle to T2 using a const_cast. The result of a reference const_cast refers to the original object. A null value is converted to the null value of the destination type. A program in which v in the const cast expression is the nullptr literal is ill-formed. A const cast shall never invoke a boxing conversion sequence. 15.3.11 Safe cast A safe cast performs the optimal cast for frameworks programming. The compiler processes a safe_cast expression as follows: The compiler performs a lookup in the current context for the name safe_cast. If the name refers unambiguously to ::cli::safe_cast, or the name is not found, then the expression is processed by the compiler according to the following grammar, and interpreted according to the rules specified herein.
safe_cast < type-id > ( expression )
The result of the expression safe_cast<T>(v) is the result of converting the expression v to type T. If T is a tracking reference type, the result is a gc-lvalue; otherwise, the result is an rvalue. Types shall not be defined in a safe_cast. The safe_cast operator shall not cast away constness. The type T and the type of v shall not be a native class, a pointer, a pointer-to-member, a native reference, or an indirection to a native class, pointer, or pointer-to-member. [Note: Except for the cases just mentioned, a safe_cast in which the target type or the type of the expression is anything else is always verifiable. An explicit type conversionalso known as a C-style castalways defaults to safe cast behavior when the arguments allow the generation of verifiable code for the conversion. end note] An expression e can be explicitly converted to a type T using a safe_cast of the form safe_cast<T>(e) if the declaration T t(e); is well-formed, for some invented temporary variable t. The effect of such an explicit conversion is the same as performing the declaration and initialization and then using the temporary variable as the result of the conversion. The result is a gc-lvalue if T is a tracking reference type, and an rvalue otherwise. The expression e is used as a gc-lvalue if and only if the initialization uses it as a gclvalue. Otherwise, the safe_cast shall perform one of the conversions listed below. No other conversion shall be performed explicitly using safe_cast. The inverse of any standard conversion sequence, other than the lvalue-to-rvalue, array-to-pointer, functionto-pointer, pointer conversions, pointer-to-member conversions, and Boolean conversion, can be performed 76
Expressions
explicitly using safe_cast. Such a safe_cast is subject to the restriction that the explicit conversion does not cast away constness, and the following addition rules for specific cases: A value of integral or enumeration type can be explicitly converted to an enumeration type. The value is unchanged if the original value is within the range of the enumeration values. Otherwise, the resulting enumeration value is unspecified. If T is handle to cv1 D, and the type of v is handle to cv2 B, cv1 shall have the same cvqualification as, or greater cv-qualification than, cv2, and a run-time check is applied to determine that D inherits from B. (For metadata and result details, see 34.5.1.) A System::InvalidCastException is thrown if the conversion fails. In the handle case, if the value of v is a null value, the result is the null value of type T. If the conversion cannot succeed at runtime, the program is ill-formed. [Example: if two ref classes A and B are unrelated, and the program uses safe_cast<A^>(b) where b has type B^, the dynamic check cannot succeed. end example] If T is tracking reference to cv1 D, and the type of v is cv2 B, cv1 shall have the same cvqualification as, or greater cv-qualification than, cv2, and a run-time check is applied to determine that D inherits from B. (For metadata and result details, see 34.5.1.) A System::InvalidCastException is thrown if the conversion fails. If the conversion cannot succeed at runtime, the program is ill-formed. An rvalue of type handle to cv1 R can be converted to an lvalue of type V, where V is a value type. R shall be System::Object, System::ValueType, or an interface that V implements. If V is an enumeration type, R can also be System::Enum. (For metadata and result details, see 34.5.1.) A System::InvalidCastException is thrown if the conversion fails. This conversion sequence is called unboxing. [Note: safe_cast is the only cast that can result in unboxing. end note]
15.4 Unary expressions

15.4.1 Unary operators 15.4.1.1 Unary & When applied to an lvalue of type T, & yields a T* (see Standard C++ 5.3.1/2). When applied to a gc-lvalue of type T, & yields an interior_ptr<T> (12.3.6). A program that attempts to apply the built-in unary & operator to an instance of a ref class type, a literal field, or to a property, or to an initonly field outside of the classs constructor, is ill-formed. A program that attempts to take the address of a member function of a non-native class in any context other than in the creation of a delegate, is ill-formed. There is no pointer-to-member representation for members of non-native classes. [Example:
delegate void D(int i); ref struct R { static void M1(int a) { } void M2(int b) { } virtual void M3(int c) { } }; int main() { R^ r = gcnew R; D^ d; d = gcnew D(&R::M1); d = gcnew D(r, &R::M2); d += gcnew D(r, &R::M3); }
end example] For details on the metadata for delegate creation, see 34.14. 77
15.4.1.2 Unary * The C++ Standard (5.3.1/1) is augmented to allow for indirection on handles. Specifically, the following text: The unary * operator performs indirection: the expression to which it is applied shall be a pointer to an object type, or a pointer to a function type and the result is an lvalue referring to the object or function to which the expression points. If the type of the expression is pointer to T, the type of the result is T. has been replaced with: The unary * operator performs indirection: the expression to which it is applied shall be one of the following: If the expression is a pointer to an object type or a pointer to a function type, then the result is an lvalue referring to the object or function to which the expression points. If the type of the expression is pointer to T, the type of the result is T. If the expression is a handle to an object, then the result is a gc-lvalue referring to the object to which the expression points. If the type of the expression is handle to T, the type of the result is T.
Dereferencing a T^ yields a gc-lvalue of type T. When operator* is applied to a string literal, that literal is converted to an "array of n const char" or "array of n const wchar_t", as appropriate. The following built-in operator functions exist:
const char& operator*(<narrow-string-literal-type>); const wchar_t& operator*(<wide-string-literal-type>);
[Note: Because user-defined operators can work on handles, when a ref or value class has a user defined instance unary operator *, dereferencing a handle to such a class will invoke the user defined operator rather than actually dereferencing the handle. This is because all instance operators work on the class type as well as on a handle to the class (Standard C++ 19.7.1). For example:
ref struct R { int operator*() { Console::WriteLine("R::operator*"); return 42; } }; int main() { R^ r1a = gcnew R; int x = *r1a; // calls operator*() R r1b; x = *r1b; // calls operator*()
As this may be surprising to programmers, a quality implementation should warn when a ref class or value class has an instance operator *. The preferred alternative to such an operator is a pair of static operators, so that the operand is clearly stated to be either the class type or a handle to the class type, as follows:
ref struct R { static int operator*(R^ r) { Console::WriteLine("R::operator*(R^)"); return 42; } static int operator*(R% r) { Console::WriteLine("R::operator*(R%)"); return 42; } };
78
Expressions int main() { R^ r2a = gcnew R; int x = *r2a; // calls operator*(R^) R r2b; x = *r2b; // calls operator*(R%)
end note] 15.4.1.3 Unary % The result of the unary % operator is a handle to its operand, which, ordinarily, shall be a gc-lvalue. However, if the operand is an instance of a value class, the operand can be an rvalue. If the type of the expression is T, and T is not a value class, the result is an rvalue and its type is handle to T. In particular, the result of getting a handle of an object of type cv T is handle to cv T, with the same cvqualifiers. If T is a value class, the expression invokes the boxing conversion sequence (which allows loss of cv-qualification), which results in an rvalue. [Example:
ref class R {}; value class V {}; void f(System::Object^ o) {} void g() { R r; f(%r); V v; f(%v); }
// v is boxed
end example] [Note: All handles to the same CLI heap-based object compare equal. For value classes, because % is a boxing operation, multiple applications of % results in handles that do not compare equal. end note] A program that applies the unary % operator to a native class type is ill-formed. 15.4.1.4 Unary ^ No such operator exists. [Rationale: As a result, there is asymmetry between %/^ and &/*, in that unary * is used to dereference both * and ^. However, allowing a single syntax to be used in the latter case permits the writing of agnostic templates and generics. In any event, adding this operator would provide no new semantics, and would preclude the addition of such an operator later on, with new semantics. end rationale] 15.4.1.5 Logical negation The C++ Standard (5.3.1/8) is augmented as follows: The operand of the logical negation operator ! is implicitly converted to bool (clause 4); its value is true if the converted operand is false and false otherwise. If the implicit conversion to bool is ill-formed and the operand is a handle type or a type given by a generic type parameter not constrained by the value class constraint, the value is true if the handle is null and false if the handle is not null. The type of the result is bool. [Example:
ref class R { }; R^ r = ; if (!r) // handle is null else // handle is non-null
end example] 15.4.2 Increment and decrement See 19.7.3.
79
15.4.3 Sizeof The C++ Standard (5.3.3/1) is augmented, as follows: The sizeof operator shall not be applied to an expression that has function or incomplete type, or to an enumeration type before all its enumerators have been declared, or to the parenthesized name of such types, or to an lvalue that designates a bit-field, or to an expression that has null type, or to a handle, or to a tracking reference, or to a ref class. sizeof(char), sizeof(signed char) and sizeof(unsigned char) are 1; the result of sizeof applied to any other fundamental type (3.9.1) is implementation-defined. [Note: in particular, sizeof(bool) , and sizeof(wchar_t), sizeof(short int), sizeof(int), sizeof(long int), sizeof(long long int), sizeof(float), sizeof(double), and sizeof(long double) are implementation-defined. end note] C++ Standard (5.3.3/2) is augmented by the addition of the following: When applied to a value class type, handle type, or generic type parameter, the result is not a compile-time constant expression. [Note: The definition of value class types excludes fundamental types and pointers, thus sizeof expressions on fundamental types and pointers are still compile-time constant expressions. end note] When applied to a ref class type or interface type, the program is ill-formed. Due to requirements imposed by the CLI Standard, size_t shall be at least a 4-byte, unsigned integer. 15.4.4 New A program is ill-formed if it attempts to allocate memory using new for an object of CLI class type other than a simple value class (22.4). 15.4.5 Delete The C++ Standard (5.3.5/1) is augmented to allow for deletion of objects allocated on the CLI heap, as follows: The operand shall have a pointer type, a handle type, or a class type having a single conversion function (12.3.2) to a pointer type. In the first alternative (delete object), the value of the operand of delete shall be a pointer or handle to a non-array object or a pointer to a sub-object (1.8) representing a base class of such an object (clause 10). If the delete-expression calls the implementation deallocation function (3.7.3.2), and if the operand of the delete expression is not the null pointer constant, the deallocation function will deallocate the storage referenced by the pointer or handle thus rendering the pointer or handle invalid. The array form of delete shall not be used on a handle type. Inside of a generic, if an objects type is a generic type parameter, delete can be used to invoke that objects destructor. If the generic parameter type is constrained to the System::IDisposable interface, the delete expression evaluates to a call through that interface on the object. If the generic parameter type is not constrained to the System::IDisposable interface, the object is converted to System::IDisposable^ using dynamic cast and the call is made through the converted object if the handle is not null. [Note: In the latter case, the conversion may require boxing if the generic type parameter can be a value type. Other than the negligible performance overhead of boxing and the ensuing dynamic cast to IDisposable^, calling the destructor on the boxed object will have no semantic impact on the program, as destructors on value types don't do anything (they cannot be defined by users). end note]
80
Expressions
15.4.6 The gcnew operator The gcnew operator is similar to the new operator, except that the former creates an object on the CLI heap. The type of the result of the gcnew operator is a handle to the type of the object allocated. In out-of-memory situations, gcnew throws System::OutOfMemoryException. There is no array form of gcnew. There is no placement form of gcnew. The gcnew operator cannot be overloaded or replaced. There is no class-specific form of gcnew. A program is ill-formed if it attempts to allocate memory for an object of native class type using gcnew. In the C++ Standard (5.3.4), a new-expression is used to allocate memory for an object at runtime. This grammar is augmented to accommodate the addition of the gcnew operator, as follows: new-expression:
::opt new new-placementopt new-type-id new-initializeropt ::opt new new-placementopt ( type-id ) new-initializeropt gcnew type-specifier-seq new-initializeropt array-initopt
In the gcnew case, the type of the object being allocated shall not be an abstract class type, nor shall it be incomplete. array-init shall only be used when creating a CLI array (see 24.2). [Note: The gcnew operator applied to a value class creates a boxed value. end note] The gcnew operator is used to create an instance of a delegate. For more information, see 27.2. 15.4.7 The throw expression As control passes from a throw-expression to a handler, finally-clauses, if any, are invoked for all try-block or function-try-blocks entered since the try-block or function-try-block containing the handler was entered. The finally-clauses are invoked in the reverse order of the invocation of their parent try-block or functiontry-blocks. The automatic destruction of objects in any given try-block or function-try-block required by the C++ Standard (15.2) takes place prior to the invocation of any finally-clause associated with that try-block or function-try-block. For an example, see 16.4 If an object is thrown by handle (regardless of the kind of class to which the handle refers), the exception handling mechanism used shall be that defined by the CLI. (This includes boxed value types.) Otherwise, the Standard C++ mechanism shall be used. Almost all types of objects can be thrown; exceptions to this rule are ref classes and value classes being thrown by value or by reference. It is always permitted to throw an object by handle. Other than stated in this Standard, the set of types that shall not be thrown using the CLI mechanism is the same as that for Standard C++. A program that attempts to throw nullptr is ill-formed.
15.5 Explicit type conversion (cast notation)

The rules in the C++ Standard (5.4/5) is augmented for C++/CLI by including safe casts before static casts. a const_cast a safe_cast a safe_cast followed by a const_cast a static_cast a static_cast followed by a const_cast a reinterpret_cast
81
a reinterpret_cast followed by a const_cast
[Note: Standard C++ programs remain unchanged by this, as safe casts are ill-formed when either the expression type or target type is a native class. end note] If both the type of the argument and the type being converted to are not a native class, a pointer, a pointer-tomember, a native reference, or an indirection to a native class, pointer, or pointer-to-member, then an explicit type conversion shall not use static_cast or reinterpret_cast. [Note: When arguments involve CLI class types, explicit type conversions always produce verifiable results. This enables programmers to use explicit type conversion syntax as the most suitable alternative for another language's cast notation. end note]
15.6 Additive operators

15.6.1 Delegate combination Every delegate type provides the following predefined operator, where D is the delegate type:
static D^ operator +(D^ x, D^ y);
The binary + operator performs delegate combination when both operands are of the same delegate type D. The result of the operator is the result of calling System::Delegate::Combine(x,y), and casting the result to D^. [Note: For examples of delegate combination, see 15.6.1 and 27.1. Since System::Delegate is not itself a delegate type, operator+ is not defined for it. The behavior when either operand is nullptr is described in 27.1. end note] 15.6.2 Delegate removal Every delegate type provides the following predefined operator, where D is the delegate type:
static D^ operator (D^ x, D^ y);
The binary - operator performs delegate removal when both operands are of the same delegate type D. The result of the operator is the result of calling System::Delegate::Remove(x,y), and casting the result to D^. [Note: the += and -= operator are defined via assignment operator synthesis (19.7.4). The behavior when operand y is nullptr is described in 27.1. end note] [Example:
delegate void D(int x); ref struct Test { static void M1(int i) { } static void M2(int i) { } }; int main() { D^ cd1 = gcnew D(&Test::M1); D^ cd2 = gcnew D(&Test::M2); D^ cd3 = cd1 + cd2; cd3 -= cd1; cd3 += cd1; cd3 = cd3 (cd1 + cd2); }
end example] 15.6.3 String concatenation When the binary operator+ is applied to a string literal, that literal is converted to System::String^. As a result, when a value having any integral type is added to a string literal, string concatenation results. [Note: This change in behavior from Standard C++ is intentional. end note] The following built-in operator functions exist:
82
Expressions System::String^ System::String^ System::String^ System::String^ operator+(<narrow-string-literal-type>, integer-type); operator+(<wide-string-literal-type>, integer-type); operator+(integer-type, <narrow-string-literal-type>); operator+(integer-type, <wide-string-literal-type>);
where integer-type is any integer type. When one of the operands to the binary + operator is a System::String^, string concatenation results. If the other operand does not also have type System::String^, its value is converted to that type by calling its ToString function. The following built-in operator functions exist:
System::String^ operator+(System::String^, System::String^); System::String^ operator+(System::String^, System::Object^); System::String^ operator+(System::Object^, System::String^);
[Example:
Point^ p = gcnew Point(5,6); String^ s = "C++" + L"/CLI"; s = 3 + " apples"; s = "p is " + p; // s => "C++/CLI" // s => "3 apples" // s => "p is (5,6)"
end example] These three built-in functions can be hidden by user-defined versions. [Example: The program
String^ operator+(String^ l, String^ r) { return l; } int main() { Console::WriteLine("ABC" + "DEF"); }
prints "ABC". end example] A program containing an expression of the form strlit - intexp, where strlit is a string literal and intexp is any integer expression, is ill-formed.
15.7 Shift operators

To accommodate the addition of the types long long int and unsigned long long int, the C++ Standard (5.8/2) is augmented, as follows: The value of E1 << E2 is E1 (interpreted as a bit pattern) left-shifted E2 bit positions; vacated bits are zero-filled. If E1 has an unsigned type, the value of the result is E1 multiplied by the quantity 2 raised to the power E2, reduced modulo ULLONG_MAX+1 if E1 has type unsigned long long int, ULONG_MAX+1 if E1 has type unsigned long, UINT_MAX+1 otherwise. [Note: the constants ULLONG_MAX, ULONG_MAX, and UINT_MAX are defined in the header <climits>). end note]
15.8 Relational operators

15.8.1 Handle equality operators Every ref class type and value class type C implicitly provides the following predefined equality operators:
bool operator ==(C^ x, C^ y); bool operator !=(C^ x, C^ y);
The implicity provided handle equality operators are used only if overload resolution finds no applicable equality operators (user-defined or otherwise defined in this specification). [Example: Delegates and System::String have equality operators defined already. If overload resolution selects one of those operators, the implicitly defined handle equality operators are not applicable. end example] There are special rules for determining when a handle equality operator is applicable. For an equalityexpression with operands of type A^ and B^, define A0 as follows: If A is a generic type parameter known to be a ref class, let A0 be the effective base class of A.
83
Otherwise, if A is an interface type, a ref class type, a value type other than pointers, or the null type, let A0 be the same as A. Otherwise, no implicit handle equality operator is applicable. If A0 is an interface type, a delegate type, System::Delegate, or System::String, let A1 be System::Object. Otherwise, if A0 is a CLI array type, let A1 be System::Array. Otherwise, A0 is the null type, a ref class type, or a value type other than pointer, and let A1 be the same as A0.
Now define A1 as follows:
Define B0 and B1 in the same manner. Now determine if any implicit handle equality operators are applicable as follows: If both of the types A and B are the null type, then overload resolution is not performed and the result is constant true for operator== and false for operator!=. Otherwise, if there is no identity or handle conversion from A0^ to B0^ or no identity or handle conversion from B0 to A0, then no implicit handle equality operator is applicable. Otherwise, if there is an identity or handle conversion from A1^ to B1^, then the implicit handle operator for B1 is applicable. Otherwise, if there is a handle conversion from B1^ to A1^, then the implicit handle operator for A1 is applicable. Otherwise, no implicit handle equality operator is applicable.
If the operands to an equality-expression are not handles, no implicit handle equality operator is applicable. [Note: The rules here have the following implications: The implicit handle equality operators cannot be used to compare types that are known to be different. For example, two types A and B that derive from System::Object could never be successfully compared for identify. Similarly, if A is a ref class and B is an interface that A does not implement, then no implicit handle equality operator applies. The implicit handle equality operators do not permit value class operands to be campared without a user-defined equality operator. The implicit handle equality operators never cause boxing conversions to occur for an operand. Such a conversion would be meaningless.
end note]
When overload resolution rules select an equality operator other than the implicit handle equality operator, selection of an implicit handle equality operator can be forced by explicitly casting one or both operands to System::Object^. 15.8.2 Delegate equality operators Every delegate type implicitly provides the following predefined comparison operators:
bool operator ==(Delegate^ x, Delegate^ y); bool operator !=(Delegate^ x, Delegate^ y);
These are implemented in terms of System::Delegate::Equals. If the two operands are of different delegate types, the expression is ill-formed. [Rationale: Two different delegate types can never successfully result in equality. Overload resolution can promote both delegate types to System::Delegate postponing equality failure to run-time. end rationale]
84
Expressions
15.8.3 String equality Equality of System::String handles is defined by System::String::operator== and System::String::operator!=.
15.9 Logical AND operator

The C++ Standard (5.14/1) is augmented as follows: The && operator groups left-to-right. The operands are both implicitly converted to type bool (clause 4). If that conversion is ill-formed and the operand is a handle type or a type given by a generic type parameter not constrained by the value class constraint, the operand is tested for the null value, returning true if not null and false if it is null. Otherwise, if the conversion to bool is ill-formed and the operand is not a handle type or a type given by a generic type parameter not constrained by the value class constraint, the program is ill-formed. The result is true if both operands are true and false otherwise. Unlike &, && guarantees left-to-right evaluation: the second operand is not evaluated if the first operand is false.
15.10 Logical OR operator

The C++ Standard (5.15/1) is augmented as follows: The || operator groups left-to-right. The operands are both implicitly converted to bool (clause 4). If that conversion is ill-formed and the operand is a handle type or a type given by a generic type parameter not constrained by the value class constraint, the operand is tested for the null value, returning true if not null and false if it is null. Otherwise, if the conversion to bool is ill-formed and the operand is not a handle type or a type given by a generic type parameter not constrained by the value class constraint, the program is ill-formed. It returns true if either of its operands is true and false otherwise. Unlike |, || guarantees left-to-right evaluation; moreover, the second operand is not evaluated if the first operand evaluates to true.
15.11 Conditional operator

With regard to expressions of the following forms
e e e e ? ? ? ? p : nullptr nullptr : p h : nullptr nullptr : h
where e is an expression that can be implicitly converted to bool, p has pointer type, and h has handle type, the C++ Standard (5.16/6) is augmented to The second and third operands have pointer type, or one has pointer type and the other is a null pointer constant or null value constant; pointer conversions and qualification conversions are performed to bring them to their composite pointer type. The result is of the composite pointer type. If either the second or the third operands have a handle type, and the other operand is the null value constant, the result is of the handle type.
15.12 Assignment operators

In the expression E1 op= E2, E1 can be a property, because after synthesis that expression is treated as E1 =
E1 op E2.
A program that attempts to use the result of an assignment expression of the form E1 = E2 in which E1 is a property, is ill-formed. [Note: The type of the result of such an expression is the type of E1, and since the set accessor function for the property has type void, the result has type void. end note] For information about the synthesis of compound assignment operators see (19.7.4). Property and event rewrite rules are covered in 15.14. The left operand of an assignment shall be an lvalue or a gc-lvalue.
85
15.13 Constant expressions

The C++ Standard (5.19/2) provides a list of Other expressions [that] are considered constant-expressions only for the purpose of non-local static object initialization. That list is augmented by the addition of the following: the null value constant. A literal field can be used in any context that permits a literal of the same type. As such, a literal field can be present in a compile-time constant expression. To accommodate the addition of literal fields, the C++ Standard is augmented by the addition of the following after 5.19/3: A literal constant expression includes arithmetic constant expression, string literals of type System::String^, and the null value constant nullptr. String concatenation expressions that use only literal values can be evalutated by the compiler and are therefore considered compile-time expressions. [Example:
#define X 42 ref struct R { literal String^ Truth = "The meaning of life is " + X; };
end example] When a static const variable is brought into scope through #using, the compiler cannot treat it as a literal value. Thus, it cannot be used in contexts in which a literal is needed (such as a template non-type argument or native array size). However, when a static const variable is brought in via #include, the Standard C++ rules as to whether it can be used as a literal, are followed.
15.14 Property and event rewrite rules

For the purposes of lookup, properties are treated as class data members. The evaluation of an expression involving one or more properties requires that expression to be rewritten using the accessor functions (19.5.3) for those properties. Before a property expression is rewritten using accessor functions, operator synthesis rules (19.7.4) shall be applied to that expression. (As a result, the property rewrite process will never encounter a compound assignment operator.) Consider the expression E1 @ E2, in which @ represents a binary operator. If E2 is a property, it shall be rewritten as a call to that property's get accessor function, before further evaluation. If E1 is a property, then if @ is the simple assignment operator, the expression shall be rewritten as a call to the property's set accessor function; otherwise, E1 shall be rewritten as a call to the property's get accessor function.. If the expression E evaluates to a property and E is not an operand to a binary operator, E shall be rewritten as a call to that property's get accessor function. Rewrites for property expressions are different for scalar and indexed properties. If P is a scalar property (19.5): The property get rewrite shall be P::get(). The property set rewrite shall be P::set(expression), where expression corresponds to the right-hand side of a simple assignment operator expression. The property get rewrite shall be P::get(expression-list). The property set rewrite shall be P::set(expression-list, expression), where expression corresponds to the right-hand side of a simple assignment operator expression.
If E is an indexed property (19.5), it has the general form P[expression-list].
86
Expressions
[Example: Given that P, Q, and R are scalar properties, the expression

P += Q * !R
is converted by operator synthesis to

P = P + Q * !R
which is then rewritten as

P::set(P::get() + Q::get() * !R::get())
In addition, given that A, B, and C are indexed properties, the expression

A[i] = B[j,k] + C[l,m,n]
is rewritten as
A::set(i, B::get(j,k) + C::get(l,m,n))
end example] The rewrite rules for the prefix and postfix ++ and -- operators are discussed in 19.7.3. If lookup finds multiple properties by the same name in a class, an expression of the form P[expression-list] shall always be interpreted as an indexed property access (even if the number of arguments does not match any existing property). If the only property found is a scalar property, the rewrite rule used shall be that for a scalar property get, and the subscript operator shall be applied to the result of that property get. [Example: In the following example, the class R has only one property by the name P. Since it is a scalar property, the subscript operator is applied to the result of the property.
ref struct R { property String^ P { String^ get() { } } }; int main() { R^ r = gcnew R; wchar_t c = r->P[0]; }
// calls String's default-indexed property
In the next example, R has two properties by the name X. Thus, all subscripts to X are interpreted as indexed properties. Because no set function exists that matches the overload of the rewrite, the following code is illformed.
ref class R { array<int>^ MyArray; public: R() { MyArray = gcnew array<int>(10); } property array<int>^ X { array<int>^ get() { return MyArray; } } property int X[int] { int get(int i) { return i*i; } } }; int main() { R r; r.X[2] = 1; // error no R::X::set(int,int) exists int y = r.X[2]; // calls R::X::get(int) }
end example] After property expressions are rewritten, the resulting expression is reevaluated using existing rules. At that time, it is possible that overload resolution will fail to find an acceptable function, in which case, the program is ill-formed. [Example: An indexed property is rewritten yet no property access method takes the
87
required number of arguments. If a property only has a get accessor function, yet an expression involving that property is rewritten as a property set, lookup will fail to find a set accessor function. end example] Before being rewritten, properties act like fields. As such, when lookup finds a property or field name, it does not look further in the base classes for more property names, even if the class is a hidebysig class (10.7). However, after being rewritten, the accessor functions for a property do follow the same rules as other functions for hidebysig lookup. When the left operand of a compound assignment operator is an event, operator synthesis shall not be applied. Given the expression E1 @ E2, in which @ represents a binary operator, if E1 is an event, the event is rewritten with the following rules: If @ is +=, the expression is rewritten as an event add, E1::add(E2). If @ is -=, the expression is rewritten as an event remove, E1::remove(E2).
Otherwise, the program is ill-formed. Given the expression E(expression-list), if E is an event, the expression is rewritten as an event raise, E::raise(expression-list). All other usages of an event in an expression are ill-formed. [Example: Given that V is an event and D is a delegate, the expression V += D is rewritten as V::add(D), the expression V -= D is rewritten as V::remove(D), the expression V(this, e) is rewritten as V::raise(this, e). end example] After an event expression is rewritten, it is reevaluated using existing rules. At that time, it is possible that overload resolution will fail to find an acceptable function, in which case, the program is ill-formed. [Example: A delegate cannot be added to an event if they have different delegate types. end example]
88
Statements
16. Statements
Unless stated otherwise in this clause, all existing statements are supported and behave as specified in the C++ Standard (6).
16.1 Selection statements

16.1.1 The switch statement A program is ill-formed if it uses a switch statement to transfer control in to a finally-clause.
16.2 Iteration statements

In addition to the three iteration statements specified by Standard C++ (6.5), the iteration-statement production is augmented to include the for each statement. iteration-statement: while ( condition ) statement do statement while ( expression ) ; for ( for-init-statement conditionopt ; expressionopt ) statement foreach ( type-specifier-seq declarator in assignment-expression ) statement 16.2.1 The for each statement The for each statement enumerates the elements of a collection, executing the statement for each element of that collection. Together, the type-specifier-seq and declarator declare the iteration variable of the statement. This iteration variable corresponds to a local variable with a scope that extends over statement. During execution of a for each statement, the iteration variable represents the collection element for which an iteration is currently being performed. The type of assignment-expression shall be a collection type (as defined below), and it shall be possible to convert from the element type of the collection to the type of the iteration variable using safe_cast. If assignment-expression has the value nullptr, a System::NullReferenceException is thrown. A type is said to be a collection type if it implements the System::Collections::IEnumerable interface, or implements System::Collections::Generic::IEnumerable interface, or implements the collection pattern by meeting all of the following criteria: Expression
e = c.GetEnumerator() e = c->GetEnumerator() e.MoveNext() e->MoveNext() e.Current e->Current
Return Type
E
Assertion/NotePre/Post-Condition
E is the enumerator type.
A value that can be used as a condition (see 14.2.1) rvalue, lvalue, or gc-lvalue that is an element of the collection
True if the current instance was successfully advanced to the next element; false if the current instance has passed the end of the collection. This is the element type of the collection type.
89
where c is a collection of object convertible to type T, and e is an enumerator that can be used for iteration over a collection. A type that implements IEnumerable is also a collection type, even if it doesn't satisfy the conditions above. (This is possible if it implements IEnumerable via explicit interface member implementations.) The System::Array type (24.1.1) is a collection type, and since all CLI array types derive from System::Array, any CLI array type expression is permitted in a for each statement. For singledimensional CLI arrays, the for each statement enumerators traverses the CLI array elements in increasing order, starting with index 0 and ending with index Length - 1. For multi-dimensional CLI arrays, elements are traversed such that the indices of the rightmost dimension are increased first, then the next left dimension, and so on to the left. A for each statement of the form for each (T d in <collection-expr>) statement in which <collection-expr> is a collection of T, is executed as if it were written as follows if GetEnumerator returns a handle:
{
<enumeration-type>^ e;
try { e = <collection-expr>.GetEnumerator(); while(e->MoveNext()) T d = safe_cast<T>(e->Current);
} } finally { delete e; }
statement
where e is a non-user-accessible temporary and <enumeration-type> is the type of the object returned by the GetEnumerator function. If GetEnumerator returns a pointer, the execution is the same as the handle case except e is declared as a pointer. If GetEnumerator does not return a pointer or handle, the statement is executed as if it were writtern as follows:
{
<enumeration-type> e = <collection-expr>.GetEnumerator();
while(e.MoveNext()) T d = safe_cast<T>(e.Current); }
statement
[Example: The following program pushes the values 0 through 9 onto an integer stack and then uses a for each loop to display the values in top-to-bottom order.
int main() { Stack<int>^ s = gcnew Stack<int>; for (int i = 0; i < 10; ++i) s->Push(i); for each (int i in s) Console::Write("{0} ", i); Console::WriteLine(); }

9 8 7 6 5 4 3 2 1 0
A CLI array is an instance of a collection type, so it too can be used with for each:
int main() { array<double>^ values = {1.2, 2.3, 3.4, 4.5}; for each (double value in values) Console::WriteLine(value); }
90
Statements

1.2 2.3 3.4 4.5
end example]
16.3 Jump statements

16.3.1 The break statement A program is ill-formed if it uses a break statement to transfer control out of a finally-clause. 16.3.2 The continue statement A program is ill-formed if it uses a continue statement to transfer control out of a finally-clause. 16.3.3 The return statement A program is ill-formed if it has a return statement in a finally-clause. 16.3.4 The goto statement A program is ill-formed if it uses a goto statement to transfer control in to or out of a finally-clause.
16.4 The try block

In the grammar specified by Standard C++ (15), the try-block and function-try-block productions are augmented to include an optional finally-clause, as follows: try-block:
try compound-statement handler-seq try compound-statement finally-clause try compound-statement handler-seq finally-clause
function-try-block: try ctor-initializeropt function-body handler-seq try ctor-initializeropt function-body finally-clause try ctor-initializeropt function-body handler-seq finally-clause finally-clause:
finally compound-statement
The statements in a finally-clause are always executed when control leaves the associated try-block's or function-try-block's compound-statement. This is true whether the control transfer occurs as a result of normal execution, as a result of executing a break, continue, goto, or return statement, or as a result of propagating an exception out of that try-block's or function-try-block's compound-statement. If an exception is thrown during execution of the statements in a finally-clause, the exception is propagated to the next enclosing try-block or function-try-block. If another exception was in the process of being propagated, that exception is lost. [Example:
class MyException {}; void f1(); void f2(); int main() { try { f1(); } catch (const MyException& re) { } }
91
C++/CLI Language Specification void f1() { try { f2(); } finally { } } void f2() { if ( ) throw MyException(); }
If the call to f2 returns normally, the finally block is executed after f1's try block terminates. If the call to f2 results in an exception, the finally block is executed before main's catch block gets control. end example] [Note: A program is ill-formed if it: end note] uses a break or continue, or goto statement to transfer control out of a finally-clause. has a return statement in a finally-clause. uses goto or switch statement to transfer control into a finally-clause.
92
Namespaces
17. Namespaces
C++/CLI has no additional namespace features beyond those provided by Standard C++.
17.1 Reserved namespaces

The namespace cli is reserved. The only elements permitted in this namespace shall be those defined by the language specification. [Example: These include array (24.1), interior_ptr (12.3.6.1), pin_ptr (12.3.7.1), and safe_cast (15.3.11). end example] A program that attempts to add a declaration to the namespace cli is ill-formed. A program can employ a using-directive for the namespace cli, or have a using-declaration for an entity in that namespace. A conforming implementation shall correctly consume assemblies containing public names that start with the C++/CLI-equivalent prefix ::cli::. [Note: Such names might be produced from C#, for example. end note]
93
18. Functions
18.1 <cstdarg>-style variable-argument lists
If a function whose parameter-declaration-clause terminates with an ellipsis, is called with nullptr as any argument that corresponds to the ellipsis, the program is ill-formed. [Note: The type of nullptr is not directly expressible in the language, yet the <cstdarg> machinery requires expressible types, so it can extract the arguments from the variable-argument list passed. end note] [Example:
void f(const char* pc, ...) {} int main() { f(nullptr); f("abc", nullptr); f("abc", 10, nullptr); } // valid // ill-formed // ill-formed
end example]
18.2 Name lookup

For metadata details, see 34.6.1.
18.3 Overload resolution

To accommodate string literal conversion, boxing conversion, Boolean, and handle conversion, Table 9, "conversions", in the C++ Standard, 13.3.3.1.1, "Standard conversion sequences", is augmented by the addition of some new rows, as indicated by shading below: Conversion No conversion required String literal conversion Lvalue-to-rvalue conversion Array-to-pointer conversion Function-to-pointer conversion Qualification conversions Boolean equivalence Integral promotions Floating point promotion Boxing conversion Integral conversions Floating point conversions Floating-integral conversions Pointer conversions Pointer to member conversions Handle conversions Boolean conversions Category Identity Lvalue Transformation Qualification Adjustment Promotion Promotion Exact Match 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 Rank Subclause
Conversion
Conversion
18.4 Parameter arrays

Standard C++ supports variable-length argument lists for both member and non-member functions; however, the approach used is not type-safe. C++/CLI adds a type-safe way using parameter arrays. A parameter array is defined as follows: 94
Functions
parameter-array: attributesopt ... parameter-declaration A parameter-array consists of an optional set of attributes (29), an ellipsis punctuator, and a parameterdeclaration. A parameter array declares a single parameter of the given CLI array type. The CLI array type of a parameter array shall be a single-dimensional CLI array type (24.1). In a function invocation, a parameter array permits either a single argument of the given CLI array type to be specified, or it permits zero or more arguments of the CLI array element type to be specified. The program is ill-formed if the parameter-declaration contains a default argument. [Example:
void f(... array<Object^>^ p); int main() { f(); f(nullptr); f(1, 2); f(nullptr, nullptr); f(gcnew array<Object^>(1)); f(gcnew array<Object^>(1), gcnew array<Object^>(2)); }
end example] [Example:

void F1(... array<String^>^ list) { for (int i = 0 ; i < list->Length ; i++ ) Console::Write("{0} ", list[i]); Console::WriteLine(); } void F2(... array<Object^>^ list) { for each (Object^ element in list) Console::Write("{0} ", element); Console::WriteLine(); } int main() { F1("1", "2", "3"); F2(1, L'a', "test"); array<String^>^ myarray = gcnew array<String^> {"a", "b", "c" }; F1(myarray); }
The output produced is as follows:

1 2 3 1 a test a b c
end example] When a function with a parameter array is invoked, the invocation is processed as if a new-expression (15.4.6) with an array-init (24.6) was inserted around the list of arguments corresponding to the parameter array. When there are zero arguments given for the parameter array, a zero-length CLI array shall be passed. [Example: Given the declaration
void F(int x, int y, ... array<Object^>^ args);
the following invocations of the function

F(10, 20); F(10, 20, 30, 40); F(10, 20, 1, "hello", 3.0);
correspond exactly to
95
C++/CLI Language Specification F(10, 20, nullptr); F(10, 20, gcnew array<System::Object^> {30, 40}); F(10, 20, gcnew array<System::Object^> {1, "hello", 3.0});
end example] Parameter array parameters can be passed to functions that take non-parameter CLI array arguments of the corresponding type. [Example:
void f(array<int>^ pArray); // not a parameter array void g(double value, ... array<int>^ p) { f(p); // Ok }
end example] An argument of type array can be passed to a function having a parameter array parameter, without invoking a parameter array conversion sequence. [Note: An array argument that can be converted to the parameter arrays type without a parameter array conversion, as happens in a handle conversion, will not prefer the parameter array conversion sequence. end note] When a function with a parameter array is included in the candidate set for overload resolution, two function signatures are included. Given a function signature TR F(T1, T2, , ... array<TP>), the exact form replaces the parameter array parameter with a normal array parameter (TR F(T1, T2, , array<TP>), and the expanded form replaces the parameter array parameter with a series of parameters of the array's element type (TR F(T1, T2, , TP1, TP2, , TPN)). The number of parameters in the exanded form matches the number of arguments to the function invocation. Both signatures are included before the elimination of viable functions. If the expanded form is selected by overload resolution, a parameter array conversion sequence is used to call the function. For metadata details, see 34.6.2.
18.5 Importing native functions

Functions defined in native code in one assembly can be invoked from another assembly by using the DllImportAttribute (from namespace System::Runtime::InteropServices) on the declaration of a global or namespace scope function declaration or on a static member function of a ref class or value class. Such function declarations shall not also be definitions. This attribute shall not be applied to an instance member function. This attribute provides the name of the native code assembly, the name of the function within that assembly, the calling convention to be used to call the native code function, and the character set used for string marshaling. [Example:
// MyCLib.h using namespace System::Runtime::InteropServices; [DllImport("MyCLib.dll", CallingConvention = CallingConvention::StdCall, EntryPoint="Hypot" )] extern "C" double Hypotenuse(double s1, double s2); // MyCLibApp.cpp #include "MyCLib.h" int main() { Console::WriteLine("Hypotenuse = {0}", Hypotenuse(3, 4)); }
In this case, the function named Hypot resides in the shared library MyCLib.dll. This name is mapped to that of the program element to which the attribute is applied; namely, to Hypotenuse. A calling convention is specified, as appropriate. The way in which the Hypot function is written, is implementation-defined. Here is a version written for one implementation:
96
Functions // MyCLib.c #include <math.h> __declspec(dllexport) double __stdcall Hypot(double side1, double side2) { return sqrt((side1 * side1) + (side2 * side2)); }
In the following example, the Standard C library function strcmp is imported and String^-to-char* conversion occurs on the arguments by virtue of the MarshalAsAttribute attribute (from namespace System::Runtime::InteropServices):
using namespace System::Runtime::InteropServices; [DllImport("msvcrt.dll", CallingConvention = CallingConvention::Cdecl)] extern "C" int strcmp([MarshalAs(UnmanagedType::LPStr)] System::String^ s1, [MarshalAs(UnmanagedType::LPStr)] System::String^ s2); int main() { String^ str1 = "red"; String^ str2 = "RED"; Console::WriteLine("Compare: {0}", strcmp(str1, str2)); }
end example] For metadata details, see 34.6.3.
18.6 Non-member functions

[Note: Non-member functions are treated by the CLI as members of some unspecified class; however, in C++/CLI source code, such functions cannot be qualified explicitly with that class name. end note] For metadata details, see 34.6.4.
18.7 Attributes
function-definitions (19.4) and function declarations resulting from either a simple-declaration or the first production of member-declaration can have attributes. The simple-declaration production is augmented as follows to allow attributes on function declarations and global variables: simple-declaration: attributesopt decl-specifier-seqopt init-declarator-listopt ;
97
19. Classes and members

This clause specifies the features of a class that are new in C++/CLI. However, not all of these features are available to all classes. The class-related features that are supported by native classes (20), ref classes (21), value classes (22), and interfaces (25), are specified in the clauses that define those types. [Note: A summary of that support is shown in the following table: Feature Assignment operator Class modifier Copy constructor Default constructor Delegate definitions Destructor Events Finalizer Function modifiers Initonly field Literal field Member of delegate type Override specifier Parameter arrays Properties Reserved member names Static constructor Static operators end note] Native class X X X X X X Ref class X X X X X X X X X X X X X X X X X X Value class X Interface
X X X X X X X X X X X X
X X X n/a X X n/a X X X X X
X X
19.1 Class definitions

In the C++ Standard (9), a class-specifier is used to define a class. This grammar is augmented to accommodate the addition of public and private classes, as follows: class-specifier: attributesopt top-level-visibilityopt class-head { member-specificationopt } attributes is described in 29, top-level-visibility is described in 12.4. class-head (9) is augmented to support class modifiers (19.1.1): class-head: class-key identifieropt class-modifiersopt base-clauseopt class-key nested-name-specifier identifier class-modifiersopt base-clauseopt class-key nested-name-specifieropt template-id class-modifiersopt base-clauseopt class-key (9) is augmented to support ref classes (21), value classes (22), and interface classes (25):
98
Classes and members
class-key:
class struct union refclass refstruct valueclass valuestruct interfaceclass interfacestruct
To accommodate the addition of initonly and literal fields, delegates, events, generics, and properties, the syntactic class member-declaration in the C++ Standard (9.2) is augmented, as follows: member-declaration: attributesopt initonly-or-literalopt decl-specifier-seqopt member-declarator-listopt ; function-definition ;opt ::opt nested-name-specifier templateopt unqualified-id ; using-declaration template-declaration generic-declaration delegate-specifier event-definition property-definition initonly-or-literal:
initonly literal
Attributes are described in 29, initonly fields in 19.12, literal fields in 19.11, generics in 31, delegates in 27, events in 19.6, and properties in 19.5. For metadata details, see 34.7.1. 19.1.1 Class modifiers To accommodate the addition of sealed and abstract classes, the grammar for class-head in the C++ Standard (9) is augmented to include an optional sequence of class modifiers, as follows: class-modifiers: class-modifiersopt class-modifier class-modifier:
abstract sealed
If the same modifier appears multiple times in a class-modifiers, the program is ill-formed. [Note: abstract and sealed can be used together; that is, they are not mutually exclusive. As nonmember functions are not CLS-compliant, a substitute is to use an abstract sealed class, which can contain static member functions. This is the utility class pattern. end note] A class that is both abstract and sealed shall not have a base-clause, instance constructors, or instance members; it shall have only static members, nested types, literal fields, and typedefs. The abstract and sealed modifiers are discussed in 19.1.1.1 and 19.1.1.2, respectively. 19.1.1.1 Abstract classes An abstract class follows the rules of Standard C++ for abstract classes (10.4); however, a class definition containing the abstract class modifier need not contain any abstract functions. [Example:
99
C++/CLI Language Specification struct B abstract { void f() { } }; struct D : B { }; int main() { B b; D d; } // error: B is abstract // ok
end example] A ref class that contains any abstract functions (including accessor functions) shall be explicitly declared abstract. For metadata details, see 34.7.1.1. 19.1.1.2 Sealed classes The sealed modifier is used to prevent derivation from a class. The program is ill-formed if a sealed class is specified as the base class of another class. [Example:
struct B sealed { }; struct D : B { }; // error, cannot derive from a sealed class
end example] Whether or not a class is sealed has no effect on whether or not any of its member functions are, themselves, sealed. [Note: The sealed modifier is primarily used to prevent unintended derivation, but it also enables certain runtime optimizations. In particular, because a sealed class is known never to have any derived classes, it is possible to transform virtual function member invocations on sealed class instances into non-virtual invocations. end note] For metadata details, see 34.7.1.2.
19.2 Reserved member names

To facilitate the underlying C++/CLI runtime implementation, for each CLI class type member definition that is a property or event, the implementation shall reserve several names based on the kind of the member definition (19.2.1, 19.2.2). A program is ill-formed if it contains a class that declares a property or event, and a member whose name matches any of that property or event's reserved names. During lookup, the reserved names are invisible. [Note: The reservation of these names serves several purposes: end note] In order to accommodate the CLI notion of finalizers, several names are reserved in CLI class types for functions (19.2.3). 19.2.1 Member names reserved for properties For a scalar or named indexed property P (19.5), the following names are reserved:
get_P set_P
To allow other languages to interoperate using an ordinary identifier as a function name for get or set access. Partition I of the CLI standard requires these names for CLS-producer languages.
100
Classes and members
Both names are reserved, even if the scalar or named indexed property is read-only or write-only. [Example:
ref struct A { property int P { int get() { return 123; } } }; ref struct B : A { int get_P() { return 456; } }; // error
end example] For a CLI class that has a default-indexed property (19.5), the following names are reserved:
get_Item set_Item
Both names are reserved, even if the default-indexed property is read-only or write-only. The default name suffix, Item, of a default-indexed property can be changed by applying the DefaultMemberAttribute (from namespace System::Reflection) to that property's parent type. All default-indexed properties in a class shall have the same underlying name. Once a default-indexed property's name has been changed in this way, it shall not be changed in any class derived from that property's parent type. If two interface classes declare a default-indexed property, and each specifies a different name via this attribute, a program is ill-formed if it declares a type that implements both interfaces. Alternatively, the program can change the default name suffix by applying the
System::Runtime::CompilerServices::IndexerNameAttribute to all default-indexed properties within a class. The resulting metadata will replace IndexerNameAttribute with DefaultMemberAttribute (see 34.7.5). A program is ill-formed if it uses both the IndexerNameAttribute and DefaultMemberAttribute to specify the default name suffix for the
same member. Similarly, a program is ill-formed if two default-indexed properties in the same class use
IndexerNameAttribute to specify different underlying names; all default-indexed properties in a class shall have the same IndexerNameAttribute applied. [Rationale: C++/CLI supports IndexerNameAttribute because that is the approach used by several other languages, and it supports DefaultMemberAttribute because that is what is actually emitted in metadata. end rationale]
For metadata details, see 34.7.5. 19.2.2 Member names reserved for events For an event E (19.6), the following names are reserved:
add_E remove_E raise_E
19.2.3 Member names reserved for functions For CLI class types, the following function name and parameter list combinations are reserved (where T is any ref class name):
Dispose() Dispose(bool) Finalize() __identifier(~T)() __identifier(!T)()
101
19.2.4 Possible collision with reserved property and event names The reserved name patterns for any given property or event are reserved only in the class defining that property or event. [Note: The program
ref struct B { int get_X() { Console::WriteLine("B::get_X"); return 1; } }; ref struct D : B { property int X { int get() { Console::WriteLine("D::X::get"); return 2; } } }; int main() { D d; d.get_X(); }
prints B::get_X. If a property or event is virtual and no base class has a virtual property or event of the same name, the underlying accessor functions generated for the property are introducing functions. That is, they will not override functions from the base class. The program
ref struct B { virtual int get_X() { Console::WriteLine("B::get_X"); return 1; } }; ref struct D : B { virtual property int X { int get() { Console::WriteLine("D::X::get"); return 2; } } }; int main() { D d; d.get_X(); }
prints B::get_X. The only way to override B::get_X when deriving from D is to use a named override. end note] If a function other than a property or event accessor in a derived class overrides a virtual accessor function from the base class, the program is ill-formed. These functions shall be marked with the new function modifier. This is true even if the name of the accessor function in the base class does not use the canonical get_X, set_X, add_X, remove_X, or raise_X names (which can only happen when #using an assembly that was generated in a language other than C++/CLI). [Example:
ref struct B { virtual property int X { int get() { Console::WriteLine("B::X::get"); return 1; } } }; ref struct D : B { virtual int get_X() new { Console::WriteLine("D::get_X"); return 2; } }; int main() { D d; d.get_X(); }
Without the new function modifier applied to D::get_X, the program is ill-formed. end example]
102
Classes and members
19.3 Data members

A ref or value class type can have the attribute StructLayoutAttribute (in namespace System::Runtime::InteropServices). This attribute can be used to specify the layout of a data structure, the alignment, the size, and the marshalling of strings. An instance data member can have the attribute FieldOffsetAttribute (in namespace System::Runtime::InteropServices), which controls the exact placement of that member. (For more information on this attribute, refer to the CLI Standard.) [Example:
using namespace System::Runtime::InteropServices; [StructLayout(LayoutKind::Explicit)] public value class S1 { [FieldOffset(0)] int v; [FieldOffset(4)] unsigned char c; [FieldOffset(8)] int w; }; [StructLayout(LayoutKind::Sequential, Pack=4)] public value class S2 { int v; unsigned char c; int w; }; [StructLayout(LayoutKind::Explicit, Size=12, CharSet=CharSet::Unicode)] public ref class S3 { [FieldOffset(0)] int* pi; [FieldOffset(0)] unsigned int ptrValue; }; // S3 is intended to behave like a union and should be treated as such
end example] Data members can have applied to them the attribute MarshalAsAttribute (in namespace System::Runtime::InteropServices). For more information on this attribute, see 18.5. For metadata details, see 34.7.3.
19.4 Functions
To allow attributes on a function definition, the Standard C++ grammar for function-definition (8.4) is augmented, as follows: function-definition: attributesopt decl-specifier-seqopt declarator function-modifiersopt override-specifierop ctor-initializeropt function-body attributesopt decl-specifier-seqopt declarator function-modifiersopt override-specifieropt function-try-block The addition of overriding specifiers and function modifiers requires augmentations to the Standard C++ grammar for function-definition and to one of the productions of member-declarator. [Note: The two new optional syntax productions, function-modifier and override-specifier, appear in that order, after exceptionspecification, but before function-body or function-try-block. end note] To allow attributes, function modifiers, and an override specifier on a function declaration that is not a definition, one of the productions for the Standard C++ grammar for member-declarator (9.2) is augmented, as follows: member-declarator: declarator function-modifiersopt override-specifieropt declarator constant-initializeropt identifieropt : constant-expression function-modifiers: function-modifiersopt function-modifier
103
function-modifier:
abstract new override sealed
The set of attributes on a function declaration that is not a definition shall be a subset of the set of attributes on the corresponding function definition. Attributes are described in 29. function-modifiers are discussed in the following subclauses: abstract in 19.4.3, new in 19.4.4, override in 19.4.1, and sealed in 19.4.2. override-specifier is discussed in 19.4.1. A member function declaration containing any of the function-modifiers abstract, override, or sealed, or an override-specifier, shall explicitly be declared virtual. [Rationale: A major goal of this new syntax is to let the programmer state his intent, by making overriding more explicit, and by reducing silent overriding. The virtual keyword is required on all virtual functions, except in the one case where backwards compatibility with Standard C++ allows the virtual keyword to be optional. end rationale] If a function contains both abstract and sealed modifiers, or it contains both new and override modifiers, it is ill-formed. An out-of-class member function definition shall not contain a function-modifier or an override-specifier. If a destructor or finalizer (19.13) contains an override-specifier, or a new or sealed function-modifier, the program is ill-formed. The Standard C++ grammar for parameter-declaration-clause (8.3.5) is augmented to include support for passing parameter arrays, as follows: parameter-declaration-clause: parameter-declaration-listopt ...opt parameter-declaration-list , ... parameter-array parameter-declaration-list , parameter-array There shall be only one parameter array for a given function or instance constructor, and it shall always be the last parameter specified. Parameter arrays are discussed in 18.4. For metadata details, see 34.7.4. 19.4.1 Override functions The Standard C++ grammar for direct-declarator is augmented to allow the function modifier override as well as override specifiers. override-specifier: = overridden-name-list pure-specifier overridden-name-list: id-expression overridden-name-list , id-expression In Standard C++, given a derived class with a function that has the same name, parameter-type-list, and cvqualification of a virtual function in a base class, the derived class function always overrides the one in the base class, even if the derived class function is not declared virtual. This is known as implicit overriding. A program containing an implicitly overridden function in ref classes and value classes is ill-formed. [Note: A programmer can eliminate the diagnostic by using explicit or named overriding, as described below. end note]
104
Classes and members
With the addition of the function modifier override and override specifiers, C++/CLI provides the ability to indicate explicit overriding and named overriding, respectively. If either the function-modifier override or an override-specifier is present in the derived class function declaration, no implicit overriding takes place. [Example:
ref struct B { virtual void F() {} virtual void F(int i) {} }; ref struct D1: B { virtual void F() override {} }; ref struct D2: B { virtual void F() override {} virtual void G(int i) = B::F {} }; ref struct D3: B { virtual void F() new = B::F {} }; // explicitly overrides B::F()
// explicitly overrides B::F() // named override of B::F(int)
// named override of B::F()
end example] [Note: A member function declaration containing the function-modifier override or an override-specifier shall explicitly be declared virtual (19.2.4). end note] An override-specifier contains a comma-separated list of names designating the virtual functions from one or more direct or indirect base classes that are to be overridden. An id-expression that designates an overridden name shall designate a single function to be overridden. Lookup for the name given in the id-expression starts in the containing class. [Note: If the id-expression is an unqualified name, and the containing class has a function by the same name the program is ill-formed. It is not possible to override a function within the same class. end note] Further qualification is necessary if the base class name is ambiguous. That function shall have the same parameter-type-list and cv-qualification as the overriding function, and the return types of the two functions shall be the same. [Example:
interface class I { void F(); }; ref struct B { virtual void F() { } }; ref struct D : B, I { virtual void G() = B::F, I::F { } // override B::F and I::F };
Both B::F and I::F must be listed separately. If the named override used just F, two names are found. Named overrides must designate a single function. end example] [Note: The same overriding behavior can sometimes be achieved in different ways. For example, given a base class A with a virtual function f, an overriding function might have an override-specifier of A::f, have no override specifier or override function modifier, have the function-modifier override, or a combination of the two, as in override = A::f. All override A::f. end note] The name of the overriding function need not be the same as that being overridden. A derived class shall not override the same virtual function more than once. If an implicit or explicit override does the same thing as a named override, the program is ill-formed. [Example:
interface struct I { void F(); };
105
C++/CLI Language Specification ref struct B { virtual void F() { } virtual void G() { } }; ref struct D : B, I { virtual void G() = B::F { } virtual void F() {} // error, would override B::F and I::F, but // B::f is already overridden by G. };
end example] A class is ill-formed if it has multiple functions with the same name, parameter-type-list, and cvqualification even if they override different inherited virtual functions. [Example:
ref struct D : B, I { virtual void F() = B::F { } virtual void F() = I::F { } }; // ok // error, duplicate declaration
end example] A function can both hide and override at the same time: [Example:
interface struct I { void F(); }; ref struct B { virtual void F() { } }; ref struct D : B, I { virtual void F() new = I::F { } };
The presence of the new function modifier (19.4.4) indicates that D::F does not override any method F from its base class or interface. The named override then goes on to say that D::F actually overrides just one function, I::F. end example] [Note: An override-specifier does not introduce that name into the class. end note][Example:
interface struct I { virtual void V(); }; ref struct R { virtual void W() {} }; ref struct S : R, I { virtual void F() = I::V, R::W {} }; ref struct T : S { virtual void G() = I::V {} virtual void H() = R::W {} }; void Test(S^ s) { // s could refer to an S, T, or something else s->W(); // ok, virtual call s->R::W(); // nonvirtual call to R::W s->S::W(); // nonvirtual call to R::W s->S::F(); // ok (classes derived from S might need to do this, // and theres no ambiguity in this case) } int main() { Test(gcnew S); Test(gcnew T); }
end example] 106
Classes and members
When matching signatures for the purpose of overriding virtual functions in generic ref classes (31.1), or implementing a function from an interface, the constraints on the type parameters are not considered. The constraints for the type parameters can differ. [Example: The following program
public interface struct P {}; public interface struct Q {}; public ref class PQ : P, Q {}; generic<typename T> where T : P public ref struct B { virtual void F(T) { Console::WriteLine("B::F"); } }; generic<typename T> where T : P, Q public ref struct D : B<T> { virtual void F(T) override { Console::WriteLine("D::F"); } }; int main() { B<PQ^>^ b = gcnew D<PQ^>; b->F(gcnew PQ); }
prints D::F. Because D<T>^ has a handle conversion to B<T>^ only if T is the same, it is not type safe when the overriding virtual function has covariant parameters to the function it is overriding (its only type safe to override with contravariant parameters), as the parameters will be the same. end example] For metadata details, see 34.7.4.1. 19.4.2 Sealed function modifier A virtual member function marked with the function-modifier sealed cannot be overridden in a derived class. [Example:
ref struct B { virtual int f() sealed; virtual int g() sealed; }; ref struct D : B { virtual int f(); // error: cannot override a sealed function virtual int g() new; // okay: does not override B::g };
end example] [Note: A member function declaration containing the function-modifier sealed shall explicitly be declared virtual. end note] If there is no virtual function to implicitly override in the base class, the derived class introduces the virtual function and seals it. Whether or not any member functions of a class are sealed has no effect on whether or not that class itself is sealed. An implicit, explicit, or (in a CLI class type, a) named override can succeed as long as there is a non-sealed virtual function in at least one of the bases. [Example: Consider the case in which A::f is sealed, but B::f is not. If C inherits from A and B, and tries to implement f, it will succeed, but will only override B::f. end example] For metadata details, see 34.7.4.2. 19.4.3 Abstract function modifier Standard C++ permits virtual member functions to be declared abstract by using a pure-specifier. C++/CLI provides an alternate approach via the function-modifier abstract. The two approaches are equivalent;
107
using both together is well-formed, but redundant. [Example: A class shape can declare an abstract function
draw in any of the following ways: virtual void draw() = 0; virtual void draw() abstract; virtual void draw() abstract = 0; // Standard C++ style // function-modifier style // okay, but redundant
end example] [Note: A member function declaration containing the function-modifier abstract shall be declared virtual. end note] For metadata details, see 34.7.4.3. For metadata implications on the parent class for both abstract functions, see 34.7.1.1. 19.4.4 New function modifier A function need not be declared virtual to have the new function modifier. If a function is declared virtual and has the new function modifier, that function does not override another function. However, for CLI class types, it can override another function with a named override. A function that is not declared virtual and is marked with the new function modifier does not become virtual and does not implicitly override any function. [Example:
ref struct B { virtual void F() { Console::WriteLine("B::F"); } virtual void G() { Console::WriteLine("B::G"); } }; ref struct D : B { virtual void F() new { Console::WriteLine("D::F"); } }; int main() { B^ b = gcnew D; b->F(); b->G(); }
The output produced is

B::F B::G
In the following example, hiding and overriding occur together:

ref struct B { virtual void F() {} }; interface class I { void F(); }; ref struct D : B, I { virtual void F() new = I::F {} };
The presence of the new function modifier indicates that D::F does not override any method F from its base classes. The named override (19.4.1) then goes on to say that D::F actually overrides just one function, I::F. The net result is that I::F is overridden, but B::F is not. end example] Static functions can use the new modifier to hide an inherited member. [Example:
ref class B { public: virtual void F() { } };
108
Classes and members ref class D : B { public: static void F() new { } };
end example] For metadata details, see 34.7.4.4.
19.5 Properties
A property is a member that behaves as if it were a field. There are two kinds of properties: scalar and indexed. A scalar property enables field-like access to a class object. Examples of scalar properties include the length of a string, the size of a font, the caption of a window, and the name of a customer. An indexed property enables array-like access to a CLI heap-based object (but not a class). An example of an index property is a bit-array class. Properties are an evolutionary extension of fieldsboth are named members with associated types, and the syntax for accessing scalar fields and scalar properties is the same, as is that for accessing CLI arrays and indexed properties. However, unlike fields, properties do not denote storage locations. Instead, properties have accessor functions that specify the statements to be executed when their values are read or written. Properties are defined using property-definitions: property-definition: attributesopt property-modifiersopt property type-specifier-seq declarator propertyindexesopt { accessor-specification } attributesopt property-modifiersopt property type-specifier-seq declarator ; property-modifiers: property-modifiersopt property-modifier property-modifier:
static virtual
property-indexes: [ property-index-parameter-list ] property-index-parameter-list: type-id property-index-parameter-list , type-id A property-definition can include a set of attributes (29), property-modifiers (19.5.2, 19.5.4), and property-indexes. A property-definition that does not contain a property-indexes is a scalar property, while a propertydefinition that contains a property-indexes is an indexed property. A property-definition for a scalar property, that ends with a semicolon (as opposed to a brace-delimited accessor-specification) defines a trivial scalar property (19.5.5). [Note: There is no such thing as a trivial indexed property. end note] Property definitions are subject to the same rules as function declarations with regard to valid combinations of modifiers, with the one exception being that the static modifier shall not be applied to a defaultindexed property definition. (Default-indexed properties are introduced later in this subclause.) When a property-definition includes the property-modifiers static or virtual, those modifiers actually apply to all of the propertys accessor functions. Writing these same modifiers in those accessor functions as well is permitted, but redundant. The type-specifier-seq and the declarator of a scalar property definition specifies the type of the scalar property introduced by the definition, and the declarator specifies the name of the scalar property. The type-
109
specifier-seq and the declarator of an indexed property definition specifies the element type of the indexed property introduced by the definition. [Note: Certain property types (such as pointer to function and pointer to array) cannot be written directly in a property definition; they shall first be written as a typedef, with the type synonym then used in the property definition. end note] The type of a scalar property and the element type of an indexed property shall be a type permitted as a parameter to a function. [Note: Because a native array is not allowed as a function parameter, it is not allowed as the type of a property either. end note] The identifier in declarator specifies the name of the property. For an indexed property, if default is used instead of identifier, that property is a default-indexed property. Otherwise, that property is a named indexed property. The accessor-specification declares the accessor functions (19.5.3) of the property. The accessor functions specify the executable statements associated with reading and writing the property. An accessor function, qualified with the property name, is considered a member of the class. For a default-indexed property, the parent property name is default. As such, the full names of the accessor functions for this indexed property are default::get and default::set. A property accessor function can be bound to a suitably typed delegate. Overloading of indexed properties on different property-index-parameter-lists is allowed. A class that contains an indexed property can contain a scalar property by the same name. The presence of a property in a class does not make that class a non-POD. A property having a type that is a reference type is not CLS-compliant. A property expression is an lvalue or gc-lvalue if its get accessor function returns an lvalue or gc-lvalue, respectively; otherwise, it is an rvalue. For metadata details, see 34.7.5. 19.5.1 Qualified names of properties and events Qualified names in C++/CLI can include properties and events. To accommodate this, the C++ grammar is augmented as follows: property-or-event-name: identifier
default
unqualified-id: identifier operator-function-id conversion-function-id ~ class-name ! class-name template-id generic-id

default
class-or-namespace-name: class-name namespace-name property-or-event-name If the nested-name-specifier of a qualified-id nominates a property or event, the name specified after the nested-name-specifier is an accessor function and is looked up in the scope of the property or event. The default keyword shall be used in a declarator only when declaring a default-indexed property. The default keyword shall be used in an expression only when a postfix-expression is evaluating a defaultindexed property. [Note: Because the grammar allows the default keyword in places where an identifier is allowed for variable names and function names, these rules restrict usage of default to use in a defaultindexed property. end note] 110
Classes and members
If the definition of an accessor function is lexically outside its property or event definition, the accessor function name shall be qualified by its property or event using the :: operator. Otherwise, the rules for declaring and defining accessor functions of properties and events are the same as those for member functions of classes. 19.5.2 Static and instance properties When a property definition includes a static modifier, the property is said to be a static property. [Note: A default-indexed property cannot be static. end note] When no static modifier is present, the property is said to be an instance property. All accessor functions in a static property are static, and all accessor functions in an instance property are instance accessor functions. [Example:
ref struct C { static property C^ MyStaticProperty { } // static property property int default[int] { }; // instance property };
end example] [Note: Like a field, when a static property is referenced using the form E::M, E shall denote a type that has a property M. When an instance property is referenced using the form E.M, E shall denote an instance having a property M. When an instance property is referenced through a pointer or handle, the form E->M is used. end note] 19.5.3 Accessor functions The accessor-specification of a property specifies the executable statements associated with reading and writing that property. accessor-specification: accessor-declaration accessor-specificationopt access-specifier : accessor-specificationopt accessor-declaration: attributesopt decl-specifier-seqopt member-declarator-listopt ; function-definition Attributes are described in 29; functions definitions in 19.4. The rules for rewriting property and event expressions into accessor function expressions are covered in 15.14. A property shall have at least one accessor function. The name of a property accessor function shall be either
get (which makes it the get accessor function) or set (which makes it the set accessor function). A
property shall have no more than one get accessor function and no more than one set accessor function. An accessor function of a property can be defined inline with the property definition, or out-of-class. A program is ill-formed if it contains an accessor function that is cv-qualified or whose final or only parameter is a parameter array. If an accessor function is not declared abstract, it shall be defined. The get accessor function of a scalar property takes no parameters and its return type shall match exactly the type of the property, type-specifier-seq. For an indexed property, the types of the parameters of the get accessor function shall correspond exactly to the types of the propertys property-indexes. The set accessor function of a scalar property has one parameter only, and its type shall match exactly the type of the property, type-specifier-seq. For an indexed property, the parameters of the set accessor function shall correspond exactly to the types of the propertys property-indexes, followed by a final parameter, whose type shall correspond exactly to the type of the property, type-specifier-seq. The return type of the set accessor function for both scalar and indexed properties shall be void. Based on the presence or absence of the get and set accessor functions, a property is classified as follows:
111
A property that includes both a get accessor function and a set accessor function is said to be a read-write property. A property that has only a get accessor function is said to be a read-only property. A property that has only a set accessor function is said to be a write-only property.
Like all class members, a property has an explicit or implicit access-specifier. Either or both of a propertys accessor functions can also have an access-specifier, which shall specify a narrower access than the propertys accessibility for that accessor function. An access-specifier on an accessor function specifies access for that accessor function only; it has no effect on the accessibility of members in the parent class subsequent to the parent property. The accessibility following the property is the same as the accessibility before the property. [Example: In the example
public ref class Button : Control { private: String^ caption; public: property String^ Caption { String^ get() { return caption; } void set(String^ value) { if (caption != value) { caption = value; Repaint(); } } } };
the Button control declares a public Caption property. This property does nothing more than return the string stored in a field except when the property is set, in which case, the control is repainted when a new value is supplied. Given the Button class above, the following is an example of use of the Caption property:
Button^ okButton = gcnew Button; okButton->Caption = "OK"; String^ s = okButton->Caption; // Invokes set accessor function // Invokes get accessor function
Here, the set accessor function is invoked by assigning a value to the property, and the get accessor function is invoked by referencing the property in an expression. end example] [Note: Exposing state through properties is not necessarily less efficient than exposing fields directly. In particular, accesses to a property are the same as calling that propertys accessor functions. When appropriate, an implementation can inline these function calls. Using properties is a good mechanism for maintaining binary compatibility over several versions of a class. end note] Accessor functions can be defined inline or out-of-class. [Example:
public ref class Point { private: int x; int y; public: property int X { int get() { return x; } void set(int value); }
// inline definition // declaration only
112
Classes and members property int Y { int get(); void set(int value) { y = value; } }
// declaration only // inline definition
};
void Point::X::set(int value) { x = value; } int Point::Y::get() { return y; }
end example] 19.5.4 Virtual, sealed, abstract, and override accessor functions An accessor function that is sealed shall also be declared virtual. The sealed modifier prevents a derived class from overriding the accessor function. An accessor function having the abstract modifier is abstract and follows the same rules as an abstract function of the containing class. An accessor function that is abstract shall also be declared virtual. [Example:
ref struct B abstract { property String^ Name { // Name is virtual virtual String^ get() abstract; } }; ref struct D : B { property String^ Name { // Name is now sealed virtual String^ get() override sealed { } } };
end example] Any properties defined in an interface are implicitly abstract. However, those properties can redundantly contain the virtual and/or abstract modifiers, and a pure-specifier. [Example:
interface class X { property int Size; // (implicit) abstract property property String^ Name { virtual String^ get() abstract = 0; } // virtual, abstract and = 0 // are permitted but are redundant };
end example] A property definition that includes the abstract modifier as well as an override modifier or an overridespecifier, specifies that the property is abstract and overrides a base property. [Note: Abstract property definitions are only permitted in abstract classes (19.1.1.1). end note] The accessor functions of an inherited virtual property can be overridden in a derived class by including a virtual property definition where the accessor functions specify an override modifier or an overridespecifier (19.4.1). This is known as an overriding property definition. With respect to overriding, accessor functions behave in the same manner as member functions. [Example:
ref struct B { property int Count { virtual int get() { } } }; ref struct D : B { property int Count { virtual int get() override { } } };
113
end example] An accessor function can override accessor functions in other properties; it can also override non-accessor functions. [Example:
ref struct B { property int Count { virtual int get() { } virtual void set(int val) { } } virtual int GetCount() { } }; ref struct D : B { property int MyCount { virtual int get() = B::GetCount { } } };
end example] An overriding property definition shall have the same or wider accessibility and exactly the same type and name as the inherited property. If the inherited property is a read-only or write-only property, the overriding property shall be a read-only or write-only property, respectively, or a read-write property. If the inherited property is a read-write property, the overriding property shall be a read-write property. A trivial scalar property shall not override another property. Except for differences in definition and invocation syntax, virtual, sealed, override, and abstract accessor functions behave exactly like virtual, sealed, override, and abstract functions, respectively. Specifically, the rules described in the C++ Standard (10.3) and 19.4.2, 19.4.1, and 19.4.3 of this Standard apply as if accessor functions were functions of a corresponding form. [Example: In the example
ref class R abstract { int y; public: virtual property int X { int get() { } } virtual property int Y { int get() { } void set(int value) { } } virtual property int Z { int get() abstract; void set(int value) abstract; } }; X is a virtual read-only property, Y is a virtual read-write property, and Z is an abstract read-write property.
19.5.5 Trivial scalar properties A trivial scalar property is defined by a property-definition ending with a semicolon (as opposed to a bracedelimited accessor-specification). [Example:
ref struct S { property int P; };
end example] A trivial scalar property is read-write and has implicitly defined accessor functions. The implied accessspecifier for these accessor functions is the same as for the parent property. Private backing storage for a trivial scalar property shall be allocated automatically, with the name of that storage being one that is
114
Classes and members
reserved to the implementation. The implicitly defined set accessor function shall have no visible behavior other than to set the private backing storage to the value provided. The implicitly defined get accessor function shall have no visible behavior other than to return the value of the private backing storage. A trivial scalar property can be static or virtual. The type of a trivial scalar property shall not be a reference type, nor shall it be cv-qualified.
19.6 Events
An event is a member that enables a class object to provide notifications. Clients can add a delegate to an event, so that the object or class will invoke that delegate. Events are declared using event-definitions: event-definition: attributesopt event-modifiersopt event event-type identifier { accessor-specification } attributesopt event-modifiersopt event event-type identifier ; event-modifiers: event-modifiersopt event-modifier event-modifier:
static virtual
event-type: ::opt nested-name-specifieropt type-name ^opt ::opt nested-name-specifieropt template template-id ^ An event-definition can include a set of attributes (29) and event-modifiers (19.6.1, 19.6.3). The eventtype of an event definition shall be a delegate type, which shall be at least as accessible as the event itself. The handle to the delegate is known as the event type. identifier designates the name of the event. When an event-definition includes the event-modifiers static or virtual, those modifiers actually apply to all of the events accessor functions. Writing these same modifiers in those accessor functions as well is permitted, but redundant. The accessor-specification declares the accessor functions (19.6.2) of the event. The accessor functions specify the executable statements associated with adding handlers to, and removing handlers from, the event, as well as raising that event. [Note: The ^ in the first production of event-type is optional to allow for type-name's being a typedef name. end note] An event-definition ending with a semicolon (as opposed to a brace-delimited accessor-specification) defines a trivial event (19.6.4). The three accessor functions for a trivial event are supplied automatically by the compiler along with a private backing store. An event-definition ending with a brace-delimited accessor-specification defines a non-trivial event. [Example: The following example shows how event handlers are attached to instances of the Button class:
public delegate void EventHandler(Object^ sender, EventArgs^ e); public ref struct MyButton : Control { event EventHandler^ Click; }; public ref class LoginDialog : Form { MyButton^ OkButton; MyButton^ CancelButton;
115
C++/CLI Language Specification public: LoginDialog() { OkButton = gcnew MyButton(); OkButton->Click += gcnew EventHandler(this, &LoginDialog::OkButtonClick); CancelButton = gcnew MyButton(); CancelButton->Click += gcnew EventHandler(this, &LoginDialog::CancelButtonClick); } void OkButtonClick(Object^ sender, EventArgs^ e) { // Handle OkButton->Click event } void CancelButtonClick(Object^ sender, EventArgs^ e) { // Handle CancelButton->Click event } };
Here, the LoginDialog constructor creates two MyButton instances and attaches event handlers to the Click events. end example] An event accessor function can be bound to a suitably typed delegate. If the add and remove accessor functions access storage for the delegate, to be thread-safe, they should each hold an exclusive lock on the containing object for an instance event, or the type object for a static event. Such a lock can be obtained by applying the attribute MethodImpl(MethodImplOptions::Synchronized) to the add and remove accessor functions. For metadata details, see 34.7.6. 19.6.1 Static and instance events When an event declaration includes a static modifier, the event is said to be a static event. When no static modifier is present, the event is said to be an instance event. 19.6.2 Accessor functions The accessor-specification for an event specifies the executable statements associated with adding handlers to, and removing handlers from, the event, as well as raising that event. The accessor-specification for an event shall contain no more than the three following accessor functions: one for a function called add, which is referred to as the add accessor function, one for a function called raise, which is referred to as the raise accessor function, and one for a function called remove, which is referred to as the remove accessor function.
A non-trivial event shall contain both an add accessor function and a remove accessor function. If that event has no raise accessor function, one is not supplied automatically by the compiler. A program is ill-formed if it contains an event having only an add accessor function or a remove accessor function, but not both. The add accessor function and remove accessor function shall each take one parameter, of the event type, and their return type shall be void. The parameter list of a raise accessor function shall correspond exactly to the parameter list of the delegate event-type, and its return type shall be the return type of the delegate event-type. [Note: Trivial events are generally better to use because use of the non-trivial form requires consideration of thread safety. end note] When an event is invoked, the raise accessor function is called. [Example:
using namespace System::Runtime::CompilerServices;
116
Classes and members public delegate void EventHandler(Object^ sender, EventArgs^ e); public ref class Button : Control { EventHandler^ action; public: event EventHandler^ Click { [MethodImpl(MethodImplOptions::Synchronized)] void add(EventHandler^ d) { } [MethodImpl(MethodImplOptions::Synchronized)] void remove(EventHandler^ d) { } } }; void raise(Object^ sender, EventArgs^ e) { }
end example] 19.6.3 Virtual, sealed, abstract, and override accessor functions An accessor function having the abstract modifier is abstract and virtual; no implementation is provided. Instead, non-abstract derived classes are required to provide their own implementation for the accessor functions by overriding the event. An accessor function that is abstract shall also be declared virtual. An event accessor function that includes both the abstract and override modifiers specifies that the access function is abstract and overrides a base event accessor function. The accessor functions of an inherited virtual event can be overridden in a derived class by including an event declaration of the same name. This is known as an overriding event declaration. An overriding event declaration does not declare a new event. Instead, it simply specializes the implementations of the accessor functions of an existing virtual event. Declaring an accessor function to be sealed prevents a derived class from overriding the accessor function. The semantics of virtual, sealed, override, and abstract accessor functions is the same as that for virtual, sealed, override and abstract functions. 19.6.4 Trivial events A trivial event is defined by an event-definition ending with a semicolon (as opposed to a brace-delimited accessor-specification). [Example:
ref struct S { event SomeDelegateType^ E; };
end example] If no event handlers have been added, the field contains nullptr. The name of any private backing storage allocated for a trivial event shall be one that is reserved to the implementation. Raising a trivial event when no event handlers have been added returns the default value of the event delegates return type; no exception is thrown. 19.6.5 Event invocation Events having a programmer-supplied or compiler-generated raise accessor function can be invoked using function call syntax. Specifically, an event E can be invoked using E(delegate-argument-list), which results in the raise accessor functions being called with delegate-argument-list as its argument list. Explicit calls to the raise accessor are permitted. Events without a raise accessor function cannot be invoked using function call syntax. Instead, the delegates
Invoke function shall be called directly.
19.7 Static operators

To support the definition of operators in ref classes, C++/CLI allows for static operator functions. 117
The rules for operators remain largely unchanged from Standard C++; however, the following rule in Standard C++ (13.5/6) is augmented to allow static member functions: A static member or a non-member operator function shall either be a non-static member function or be a non-member function and have at least one parameter whose type is a native class, a reference to a native class, a CLI class, a reference to a CLI class, a handle to a CLI class, an enumeration, a reference to an enumeration, or a handle to an enumeration. The requirements of non-member operator functions apply to static operator functions. The following rule in Standard C++ (13.5.1/1) is relaxed to allow static member functions: A prefix unary operator shall be implemented by a non-static member function with no parameters or a non-member function with one parameter, or a static member function with one parameter. The following rule in Standard C++ (13.5.2/1) is relaxed to allow static member functions: A binary operator shall be implemented either by a non-static member function with one parameter or by a non-member function with two parameters, or a static member function with two parameters. However, operators required by Standard C++ to be instance functions shall continue to be instance functions. [Note: Standard C++ specifies that these operators are: assignment operators (13.5.3), operator() (13.5.4), operator[] (13.5.5), and operator-> (13.5.6). end note] [Example:
public ref class IntVector { public: static IntVector^ operator+(IntVector^ iv, int i); static IntVector^ operator+(int i, IntVector^ iv); static IntVector^ operator+(IntVector^ iv1, IntVector^ iv2); static IntVector^ operator-(IntVector^ iv); static IntVector^ operator++(IntVector^ iv); };
end example] Static unary operators within a class T shall take one parameter, of type T, T^, T%, T&, T^%, or T^&. A static binary operator within a class T shall take two parameters, at least one of which shall have the type T, T^, T%, T&, T^%, or T^&. In either case, if T is a generic class, the parameter that satisfies the above rules shall have exactly the same type as the enclosing class. [Example:
generic <typename T1, typename U1> ref struct GR { static bool operator!(GR^); static bool operator!(GR<T1,T1>^); static bool operator!(GR<int,int>^); generic <class T2, class U2> static bool operator!(GR<T2,U2>^); generic <class T2, class U2> static bool operator!(GR<U2,T2>^); generic <class T2, class U2> static bool operator!(GR<T2,T2>^); // OK // error // error // error // error // error
};
end example] For metadata details, see 34.7.7.
118
Classes and members
19.7.1 Homogenizing the candidate overload set Standard C++ (13.3.1/2) describes how all member functions are considered to have an implicit object parameter for the purpose of overload resolution. C++/CLI expands upon this notion by creating two signatures for every member operator function in which the difference between the two signatures is the type of the implicit object parameter. For a type T, the type of the implicit object parameter in the first signature is T%, whereas the type for the second signature is T^. These signatures exist only for the purpose of overload resolution, and both signatures refer to the one member operator function from which these signatures were created. [Rationale: This allows operator functions to be called using variables that have the raw type (12.3.1) and using variables that are handles to the raw type. (This is necessary to compare operator overloads where the candidate set includes member functions and operator functions from global or namespace scope.) end rationale] [Example:
ref class R { int X, Y; public: R(int x, int y) : X(x), Y(y) {} R^ operator+(R^ param) { return gcnew R(this->X + param->X, this->Y + param->Y); } virtual String^ ToString() override { return String::Format("({0},{1})", X, Y); } }; int main() { R^ hr = gcnew R(2, 2); R r(10, 10); // handle to raw type R // raw type R
Console::WriteLine(hr + hr); Console::WriteLine(r + hr); }
end example] 19.7.2 Operators on handles Unlike pointers, some user-defined operators can be applied to handles. For example, the addition of an integer to a handle does not attempt to add an offset to the handle (as is done with pointer arithmetic); rather, lookup for a user-defined operator is performed. The Standard C++ operator lookup rules are modified in the following ways: Standard C++ (13.5.1/1) is augmented, as follows: Thus, for any prefix unary operator @ for type T, @x can be interpreted as either x->operator@() if x is a handle, x.operator@() if x is not a handle, T::operator@(x), or operator@(x). Standard C++ (13.5.2/1) is augmented, as follows: Thus for any binary operator @ for type T, x@y can be interpreted as either x->operator@(y) if x is a handle, x.operator@(y) if x is not a handle, T::operator@(x,y), or operator@(x,y). [Note: In C++/CLI, equality operators for handles behave as if they were compiler-generated or user-defined operators. end note] The rules in Standard C++ (13.5.3/1) continue to applyan assignment operator shall be an instance function. An assignment to a handle never invokes the user-defined assignment operator. In Standard C++ (13.5.4/1), although function call operators continue to be allowed only as instance functions, the text is augmented, as follows:
119
Thus, a call x(arg1,...) is interpreted as x->operator()(arg1, ...) if x is a handle, or x.operator()(arg1,...) if x is not a handle, for a class object x of type T if T::operator()(T1, T2, T3) exists and if the operator is selected as the best match function by the overload resolution mechanism. In Standard C++ (13.5.5/1), although subscript operators continue to be allowed only as instance functions, the text is augmented, as follows: Thus, a subscripting expression x[y] is interpreted as x->operator[](y) if x is a handle, or x.operator[](y) if x is not a handle, for a class object x of type T if T::operator[](T1) exists and if the operator is selected as the best match function by the overload resolution mechanism. In Standard C++ (13.5.6), the member access operator is allowed on handles; the text is augmented, as follows: An expression x->m is interpreted as (x->operator->())->m if x is a handle, or (x.operator->())->m if x is not a handle, for a class object x of type T if T::operator->() exists and if the operator is selected as the best match function by the overload resolution mechanism. [Note: Like a pointer, if no matching member access operator exists, x->y is defined as (*x).y. end note] [Rationale: The member access operator is supported on handles to provide parity with the unary dereference operator. If a class were to define both operators, there would be no way of accessing members of that class. As a result, the class member access operator is allowed to be a static member function to explicitly allow or disallow class member access through a handle. end rationale] In addition to non-static member functions as described aboved, operator-> in CLI class types can be a static member function taking one parameter. For a static operator-> in a class R, the parameter shall be R, R^, R% or a more cv-qualified alternative. In addition to the rewrite of the expression x->m provided above, x->m is interpreted as T::operator>(x)->m for a class object x of type T if a static operator-> function exists in T and if the operator is selected as the best match function by the overload resolution mechanism. [Note: The increment and decrement operators described in Standard C++ (13.5.7), have significant differences from the CLS increment and decrement operators. (See 19.7.3 for details.) end note] 19.7.3 Increment and decrement operators In C++/CLI, the static operators operator++ and operator-- behave as both postfix and prefix operators. Neither of these static operators shall be declared with the dormant int parameter described by Standard C++ (13.5.7). For the expressions x++ and x--, where the postfix operator is non-static, the following processing occurs: If x is classified as a property or indexed access: o o o o o The expression x is evaluated and the results are used in subsequent get and set accessor function calls. The get accessor function of x is invoked and the return value is saved. The selected operator is invoked with the saved value of x as its argument and the literal 0 as the argument to select the postfix operator overload. The set accessor function of x is invoked with the value returned by the operator as its only or final argument. The saved value of x is the result of the expression. The operator is processed as specified by Standard C++.
Otherwise: o
120
Classes and members
For the expressions ++x and --x, where the prefix operator is non-static, the following processing occurs: If x is classified as a property or indexed access: o o o o o The expression x is evaluated and the results are used in subsequent get and set accessor function calls. The get accessor function of x is invoked. The selected operator is invoked with the result of the get accessor function of x as its argument, and the return value is saved. The set accessor function of x is invoked with the saved value from the operator invocation as its only or final argument. The saved value from the operator invocation is the result of the expression. The operator is processed as specified by Standard C++.
Otherwise: o
For the expressions x++ and x--, where the operator is static, the following processing occurs: If x is classified as a property or indexed access, the expression is evaluated in the same manner as if the operator were a non-static postfix operator with the exception that no dormant zero argument is passed to the static operator function. Otherwise: o o o o o
x is evaluated.
The value of x is saved. The selected operator is invoked with the value of x as its only argument. The value returned by the operator is assigned in the location given by the evaluation of x. The saved value of x becomes the result of the expression.
For the expression ++x or --x, where the operator is static, the following processing occurs: If x is classified as a property or indexed access, the expression is evaluated in the same manner as if the operator were a non-static prefix operator. Otherwise: o o o o
x is evaluated.
The selected operator is invoked with the value of x as its only argument. The value returned by the operator is assigned in the location given by the evaluation of x.
x becomes the result of the expression.
[Example: The following example shows an implementation and subsequent usage of operator++ for an integer vector class:
121
C++/CLI Language Specification public ref class IntVector { public: IntVector(int vectorLength, int initValue) { } property int Length { } property int default[int] { } static IntVector^ operator++(IntVector^ iv) { IntVector^ temp = gcnew IntVector(iv->Length, 0); for (int i = 0; i < iv->Length; ++i) { temp[i] = iv[i] + 1; } return temp; } }; int main() { IntVector^ iv1 = gcnew IntVector(3,7); IntVector^ iv2; Console::WriteLine("iv1: {0}", iv1); iv2 = iv1++; // equivalent to: // IntVector^ __temp = iv1; // iv1 = IntVector::operator++(iv1); // iv2 = __temp; Console::WriteLine("iv1: {0}", iv1); Console::WriteLine("iv2: {0}", iv2); iv2 = ++iv1; // equivalent to: // iv1 = IntVector::operator++(iv1); // iv2 = iv1; Console::WriteLine("iv1: {0}", iv1); Console::WriteLine("iv2: {0}", iv2); }
The output produced is

iv1: iv1: iv2: iv1: iv2: [7:7:7] [8:8:8] [7:7:7] [9:9:9] [9:9:9]
Unlike traditional operator versions in Standard C++, this operator need not, and, in fact, should not, modify the value of its operand directly. end example] If the return type of a static operator++ or operator-- function cannot be assigned to the type on which the operator is invoked, the program is ill-formed. [Example:
value struct V { static V^ operator++(V^ v) { Console::WriteLine("V::operator++"); return v; } static operator V (V^ v) { Console::WriteLine("V::operator V"); return *v; } }; int main() { V v; ++v; // needs the conversion operator
V^ v2 = gcnew V; ++v2; // does not need the conversion operator
122
Classes and members
Without the implicit conversion operator from V^ to V, there is no way to assign a boxed value type to a plain value type. Thus, when ++v is rewritten as v = V::operator++(v), the assignment is diagnosed. In the case of ++v2, v2 is a handle to V, so no conversion is needed; it compiles as is. end example] 19.7.4 Operator synthesis The compound assignment operators (+=, -=, *=, /=, %=, >>=, <<=, ^=, &=, and |=) are synthesized from other operators. For the expression x @= y (where @ denotes one of the operators listed above): If lookup for operator@= succeeds, the rules specified so far are applied. Otherwise, the expression x @= y is rewritten as x = x @ y (in which case, 5.17/7 of the C++ Standard requires that "The behavior of an expression of the form E1 op= E2 is equivalent to E1 = E1 op E2 except that E1 is evaluated only once."), and the transformed expression is interpreted with the rules specified so far. If no overload for operator@= applies after overload resolution or synthesis, the program is ill-formed. Synthesis shall not occur for operators defined inside native classes. [Example:
public ref class IntVector { public: static IntVector^ operator+(IntVector^ iv, int i) { } static IntVector^ operator+(IntVector^ iv1, IntVector^ iv2) { } }; IntVector^ iv1 = gcnew IntVector(10); iv1 += 20; // synthesized as iv1 = iv1 + 20 iv1 += iv1; // synthesized as iv1 = iv1 + iv1
end example] If the left operand of a compound assignment operator is a property, operator synthesis shall always be used to rewrite the expression even if the type of the property has an existing compound assignment operator. 19.7.5 Naming conventions During compilation, the name of any operator function is the C++ identifier used in source code for that function. For example, the addition operators identifier is operator+. However, in metadata, that function will have a different name, of the form op_xxx. All operator function names having this form and listed in tables throughout this subclause are reserved in certain cases for use in metadata; specifically, a program that declares or defines in a CLI class type a member function having any of these names is ill-formed. The CLS identifies a set of operators upon which CLS consumer and producer language representatives have agreed. The set of CLS-compliant operators (19.7.5.1) overlaps with the set of operators supported by Standard C++ (see Partition I, 10.3, of the CLI Standard). The C++ operators that do not overlap with the CLS-compliant operators are known as C++-dependent operators (19.7.5.4). 19.7.5.1 CLS-compliant operators An operator is CLS-compliant when all of the following conditions occur: 1. The operator function is one listed in either Table 19-1: CLS-Compliant Unary Operators or Table 19-2: CLS-Compliant Binary Operators. 2. The operator function is a static member of a ref class or a value class. 3. If a value class is a parameter or a return value of the operator function, the value class is not passed by reference nor passed by pointer or handle. 4. If a ref class is a parameter or a return value of the operator function, the ref class is passed or returned by handle. The handle shall not be passed or returned by reference. If the above criteria are not met, the operator function is C++-dependent (19.7.5.4).
123
Table 19-1: CLS-Compliant Unary Operators
Metadata Function Name

op_AddressOf op_LogicalNot op_OnesComplement op_PointerDereference op_UnaryNegation op_UnaryPlus Table 19-2: CLS-Compliant Binary Operators
C++ Operator Function Name

operator& operator! operator~ operator* operatoroperator+

op_Addition op_BitwiseAnd op_BitwiseOr op_Comma op_Decrement op_Division op_Equality op_ExclusiveOr op_GreaterThan op_GreaterThanOrEqual op_Increment op_Inequality op_LeftShift op_LessThan op_LessThanOrEqual op_LogicalAnd op_LogicalOr op_Modulus op_Multiply op_RightShift op_Subtraction

operator+ operator& operator| operator, operator-operator/ operator== operator^ operator> operator>= operator++ operator!= operator<< operator< operator<= operator&& operator|| operator% operator* operator>> operator-
19.7.5.2 Non-C++ operators The CLS provides names for several operators that Standard C++ does not support. [Note: Compilers for other languages might not be tolerant to functions with these names. It is recommended that a C++/CLI implementation issue a compatibility diagnostic if a user-defined function is given one of these names listed in Annex F. end note] Metadata Function Name
op_False op_True
C++ Operator Function Name none none
19.7.5.3 Assignment operators Given that CLI assignment operators take a parameter by value and return a result by value, with regard to these operators, the CLS recommendations are incompatible with C++. As C++ requires assignment operators to be instance functions, a C++/CLI implementation is not required to generate or consume CLS assignment operators (listed in Table 19-3: CLS-Recommended Assignment Operators). As such, userdefined functions with names from Table 19-3: CLS-Recommended Assignment Operators are not given special treatment.
Table 19-3: CLS-Recommended Assignment Operators
124
Classes and members

op_Assign op_UnsignedRightShiftAssignment op_RightShiftAssignment op_MultiplicationAssignment op_SubtractionAssignment op_ExclusiveOrAssignment op_LeftShiftAssignment op_ModulusAssignment op_AdditionAssignment op_BitwiseAndAssignment op_BitwiseOrAssignment op_DivisionAssignment
C++ Operator Function Name No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent No equivalent
19.7.5.4 C++-dependent operators If an operator function does not match the criteria for a CLS-compliant operator (19.7.5.1), the operator is C++-dependent. Table 19-4: C++-Dependent Unary Operators and Table 19-5: C++-Dependent Binary Operators identify these functions. (Even though these metadata names are not CLS-compliant, all but two of them are recommended by the CLS. The two exceptions are op_FunctionCall and op_Subscript.)
Table 19-4: C++-Dependent Unary Operators

op_AddressOf op_LogicalNot op_OnesComplement op_PointerDereference op_UnaryNegation op_UnaryPlus Table 19-5: C++-Dependent Binary Operators

operator& operator! operator~ operator* operatoroperator+

op_Addition op_AdditionAssignment op_Assign op_BitwiseAnd op_BitwiseAndAssignment op_BitwiseOr op_BitwiseOrAssignment op_Comma op_Decrement op_Division op_DivisionAssignment op_Equality op_ExclusiveOr op_ExclusiveOrAssignment op_FunctionCall op_GreaterThan op_GreaterThanOrEqual op_Increment

operator+ operator+= operator= operator& operator&= operator| operator|= operator, operator-operator/ operator/= operator== operator^ operator^= operator() operator> operator>= operator++
125
op_Inequality op_LeftShift op_LeftShiftAssignment op_LessThan op_LessThanOrEqual op_LogicalAnd op_LogicalOr op_MemberSelection op_Modulus op_ModulusAssignment op_MultiplicationAssignment op_Multiply op_PointerToMemberSelection op_RightShift op_RightShiftAssignment op_Subscript op_Subtraction op_SubtractionAssignment
operator!= operator<< operator<<= operator< operator<= operator&& operator|| operator-> operator% operator%= operator*= operator* operator->* operator>> operator>>= operator[] operatoroperator-=
19.8 Non-static operators

Although C++/CLI supports Standard C++'s non-static and global operators, these operator functions are not CLS-compliant (19.7.5.1). Such operators are discussed in various contexts in 19.7 and its subclauses; specifically: Homogenizing the candidate overload set (19.7.1), operators on handles (19.7.2), increment and decrement operators (19.7.3), operator synthesis (19.7.4), and naming conventions (19.7.5). [Note: Type visibility (12.4) only applies to top-level types, not to top-level functions. As such, a global operator function cannot be seen outside its parent assembly. However, an operator implemented as a nonstatic member function can be seen outside its parent assembly. end note] Operators new and delete shall not be overloaded for CLI class types. For metadata details, see 34.7.8.
19.9 Instance constructors

Since C++/CLI has added the notion of a static constructor, all uses of the term constructor in the C++ Standard refer to what C++/CLI refers to as instance constructor. Construction for native classes in C++ specifies that the behaviors of calling virtual functions from an object's constructor results in a call to the virtual function in the class under construction or one of its bases, but not a deriving type (see 12.7 of Standard C++). The behavior of a virtual function call from a constructor of a ref class always calls the virtual function applicable from the most derived class. A constructor of a ref class executes in the following order: 1. Initialize all members of the class in declaration order. 2. Call the base classs constructor. 3. Run the body of the user-written constructor. If an exception takes place during the initialization of the class members, the destructor of each fully constructed member shall be called in reverse declaration order, and the finalizer of the class shall be called if it exists. If an exception takes place during the base classs constructor, the destructor of each member shall be called in reverse declaration order, and the finalizer of the class shall be called, if it exists. If an exception takes place in the body of the user-written constructor, the base class is destroyed in the same manner as the Dispose(true) function invokes destruction of the base class (see 34.7.13.7). After 126
Classes and members
cleaning up the base class, the destructor of each member shall be called in reverse declaration order, and the finalizer of the class shall be called if it exists. For metadata details, see 34.7.9.
19.10 Static constructors

A static constructor is a function member that implements the actions required to initialize the static data members of a ref or value class. A static constructor is declared just like an instance constructor in Standard C++ (8.4), except that the former is specified with the storage class static. A static constructor shall not have a ctor-initializer. Static constructors are not inherited, and cannot be called directly. The static constructor for a class is executed as specified in the CLI standard, Partition II. If a class contains any static fields (including initonly fields) with initializers, those fields are initialized immediately prior to the static constructors being executed and in the order in which they are declared. [Example: The code
ref struct A { static A() { cout << "Init A" << \n; } static void F() { cout << "A::F" << \n; } }; ref struct B : A { static B() { cout << "Init B" << \n; } static void F() { cout << "B::F" << \n; } }; int main() { A::F(); B::F(); }
shall produce one of the following outputs:

Init A A::F Init B B::F Init A Init B A::F B::F Init B Init A A::F B::F
because A's static constructor shall be run before accessing any static members of A, and B's static constructor shall be run before accessing any static members of B, and A::F is called before B::F. end example] A static constructor can be defined outside its parent class using the same syntax for a corresponding out-ofclass instance constructor, except that a static prefix shall also be present. [Example:
ref class R { public: static R(); R(); R(int) { } }; static R::R() { } R::R() { } // static constructor declaration // instance constructor declaration // inline instance constructor definition // out-of-class static constructor definition // out-of-class instance constructor definition
end example]
127
[Note: In Standard C++, an out-of-class constructor definition is not permitted to have internal linkage; that is, it is not permitted to be declared static. end note] A static constructor shall have an access-specifier of private. If a ref or value class has no user-defined static constructor, a default static constructor is implicitly defined. It performs the set of initializations that would be performed by a user-written static constructor for that class with an empty function body. For metadata details, see 34.7.10.
19.11 Literal fields

A literal field is a named compile-time constant rvalue having the type of the literal field and having the value of its initializer. To accommodate the addition of literal fields, one of the productions of the syntactic class member-declaration in the C++ Standard (9.2) is augmented so a member declaration can contain the initonly-or-literal identifier literal (19.1). Each member-declarator in the member-declarator-list of a literal field declaration shall contain a constantinitializer. Even though literal fields are accessed like static members, a literal field definition shall not contain the keyword static. Whenever a compiler comes across a valid usage of a literal field, the compiler shall replace that usage with the value associated with that literal field. A literal field shall have a scalar type. [Note: This includes handle types. end note] However, the declspecifier-seq in the member-declaration shall not contain a cv-qualifier. The constant-expression in the constant-initializer shall yield a value of the target type or a value of a type that can be converted to the target type by a standard conversion sequence. [Note: A constant-expression is an expression that can be fully evaluated at compile-time. Since the only way to create a non-null value of a handle type other than System::String^ is to apply the gcnew operator, and since that operator is not permitted in a constant-expression, the only possible value for literal fields of handle type other than System::String^ is nullptr. end note] When a symbolic name for a constant value is desired, but when the type of that value is not permitted in a literal field declaration, or when the value cannot be computed at compile-time by a constant-expression, an initonly field (19.12) can be used instead. Literal fields are permitted to depend on other literal fields within the same program as long as the dependencies are not of a circular nature. [Example:
ref struct X { literal double PI = 3.1415926; literal int MIN = -5, MAX = 5; literal int COUNT = MAX - MIN + 1; literal int Size = 10; enum class Color {red, white, blue}; literal Color DefaultColor = Color::red; }; int main() { double radius; cout << "Enter a radius: "; cin >> radius; cout << "Area = " << X::PI * radius * radius << "\n"; static double d = X::PI; for (int i = X::MIN; i <= X::MAX; ++i) { } float f[X::Size]; }
end example] 128
Classes and members
For a discussion of versioning and literal fields, see 19.12.2. For metadata details, see 34.7.11.
19.12 Initonly fields

To accommodate the addition of initonly fields, one of the productions of the syntactic class memberdeclaration in the C++ Standard (9.2) is augmented so a member declaration can contain the initonly-orliteral identifier initonly (19.1), thereby making that member an initonly field. Initialization of an initonly field shall occur only as part of its definition. Assignments (via an assignment operator or a postfix or prefix increment or decrement operator) to any initonly field shall occur only in an instance constructor or static constructor in that field's class. [Note: Of course, such assignment could be done via a constructors ctor-initializer. end note] Initialization of, and assignments to, initonly fields are permitted only in the following contexts: For static initonly fields, in the constant-initializer of an initonly field's member-declarator. For an instance field, in the instance constructors of the class containing the initonly field definition; for a static field, in the static constructor of the class containing the initonly field definition.
A program that attempts to assign to an initonly field in any other context, or that attempts to take that field's address or to bind it to a reference in any context, is ill-formed. The type of an initonly field shall not be a ref class. [Example:
public ref class R { initonly static int svar1 = 1;// Ok initonly static int svar2; // Error; must be initialized here, or // assigned to in the static constructor initonly static int svar3; // Ok, assigned to in the static constructor initonly int mvar1 = 1; initonly int mvar2; initonly int mvar3; public: static R(){ svar3 = 3; svar1 = 4; smf2(); } static void smf1() { svar3 = 5; } static void smf2() { svar2 = 5; } R() : mvar2(2) { mvar3 = 3; mf1(); } void mf1() { mvar3 = 5; } void mf2() { mvar2 = 5; } // Error, initializer requires static
// Ok: but overwrites the value 1
// Error; not in a static constructor
// Error; not in a static constructor // Ok // Ok
// Error; not in an instance constructor
// Error; not in an instance constructor
};
end example]
129
As one static initonly field can be explicitly initialized using the value of another, such fields are initialized in their lexical source order, prior to the execution of any code in the static constructor. For metadata details, see 34.7.12. 19.12.1 Using static initonly fields for constants A static initonly field is useful when a symbolic name for a constant value is desired, but when the type of the value is not permitted in a literal declaration, or when the value cannot be computed at compile-time. 19.12.2 Versioning of literal fields and static initonly fields Literal fields and initonly fields have different binary versioning semantics. When an expression references a literal field, the value of that member is obtained at compile-time, but when an expression references an initonly field, the value of that member is not obtained until run-time. [Example: Consider an application with the following source:
namespace Program1 { public ref struct Utils { static initonly int X = 1; literal int Y = 1; }; } namespace Program2 { int main() { Console::WriteLine(Program1::Utils::X); Console::WriteLine(Program1::Utils::Y); } }
The Program1 and Program2 namespaces denote two source files that are compiled separately, each generating its own assembly. Because Program1::Utils::X is declared as a static initonly field, the value output by Console::WriteLine is not known at compile-time, but, rather, is obtained at run-time. Thus, if the value of X is changed and Program1 is recompiled, Console::WriteLine will output the new value even if Program2 isnt recompiled. However, because Y is a literal field, the value of Y is obtained at the time Program2 is compiled, and remains unaffected by changes in Program1 until Program2 is recompiled. end example]
19.13 Destructors and finalizers

Any native class or ref class can have a user-defined destructor. Such destructors are run at the times specified by the C++ Standard: An object of any type allocated on the stack is destroyed when that object goes out of scope. An object of any type allocated in static storage is destroyed during program termination. An object that is allocated on the native heap using new, is destroyed when a delete is performed on a pointer to that object. An object that is allocated on the CLI heap using gcnew, is destroyed when a delete is performed on a handle to that object. An object that is a member of another object is destroyed as part of the destruction of the enclosing object.
For the purposes of destruction, the native and CLI heaps are treated the same. The only difference between the two heaps is the automation and timing of memory reclamation. In the case of the native heap, memory is reclaimed manually at the same time as the delete, while in the case of the CLI heap, memory is reclaimed automatically during garbage collection whether or not there was a delete. In addition, objects on the CLI heap are finalized, if a finalizer exists.
130
Classes and members
For metadata details, see 34.7.13. 19.13.1 Destructors A destructor in a ref class is defined as in Standard C++ (12.4). A ref class has a destructor if one is defined directly, or if one is generated by the compiler, with the latter occurring if the class has one or more embedded data members whose types implement the System::IDisposable interface. The access-specifier of a destructor in a ref class is ignored. The destructor of a ref class can optionally be declared virtual; however, doing so has no effect. A ref class destructor shall not have any function-modifiers (19.4), nor shall it be declared static. Destruction of a ref class object begins when: That object has automatic storage duration and it goes out of scope. That object is embedded as a member of an enclosing class, and the enclosing classs destructor executes. That object is an already constructed member of a class during whose construction an uncaught exception occurred. The delete keyword is applied to a handle that refers to that object. [Note: If the handle has a value of nullptr, destruction begins; however, it does nothing. end note] The destructor function is explicitly called on that object by the programmer. (This includes the case in which the destructor function for a particular base class is called using a qualified name.)
For an object that has completed construction (no exception was thrown from the constructor), destruction always begins by calling through the System::IDisposable::Dispose function. (See 19.9 for behavior of destructor calls from a constructor throwing an exception.) Accessing members of a ref class object after destruction is ill-formed, but no diagnostic is required. [Note: Behavior of member access of a ref class after destruction is under the control of the ref class author. The author should document whether members are usable after destruction. end note] Like constructors, virtual function calls in a destructor of a ref class result in a call to the applicable virtual function from the perspective of the most derived class of the object. For metadata details, see 34.7.13.2. 19.13.2 Finalizers As well as providing Standard C++-style deterministic cleanup via destructors, C++/CLI provides a mechanism for non-deterministic cleanup when an instance of a ref class is no longer referenced. This mechanism is called a finalizer. A special declarator syntax using an optional function-specifier followed by ! followed by the finalizers class name followed by an empty parameter list is used to declare the finalizer in a ref class definition. In such a declaration, the ! followed by the finalizers class name can be enclosed in optional parentheses; such parentheses are ignored. A typedef-name shall not be used as the class-name following the ! in the declarator for a finalizer declaration. A finalizer is used to finalize objects of its class type. A finalizer has no parameters, and no return type can be specified for it (not even void). The address of a finalizer shall not be taken. A finalizer shall not have any function-modifiers (19.4), nor shall it be declared static or virtual. A finalizer can be invoked for a const, volatile, or const volatile object. A finalizer shall not be declared const, volatile, or const volatile. const and volatile semantics are not applied on an object being finalized. They stop being in effect when the finalizer for the most derived object starts. The access-specifier of a finalizer in a ref class is ignored.
131
Any ref class can have a user-defined finalizer. The finalizer is executed zero or more times by the garbage collector, as specified by the CLI. A finalizer function in any ref class T shall only be called from another function within that same class. A call to a finalizer shall not result in the execution of the finalizer of the base class. For metadata details, see 34.7.13.3.
132
Native classes
20. Native classes

The visibility of a non-nested native class can optionally be specified via a top-level-visibility (12.4). A native class can optionally have a class-modifiers (19.1.1). A native class shall not contain members whose types are non-simple value types, ref classes, or interface classes. [Note: Allowing members of such types would make the parent type a mixed type (23). end note] A native class can contain nested ref class, value class, and interface class definitions. A native class shall not be a generic class. For metadata details, see 34.8.
20.1 Functions
A virtual member function declaration in a native class can contain: the function-modifier sealed (19.4.2). the function-modifier abstract (19.4.3).
Member functions in a native class can optionally have a parameter-array (18.4) in their parameterdeclaration-clause. Member functions in a native class can be generic (31.3). However, a program containing a native class having a virtual generic member function is ill-formed. [Note: Member functions of a native class use hidebyname lookup (10.7). end note]
20.2 Properties
A program is ill-formed if it contains a property in a native class.

A program is ill-formed if it contains a static operator in a native class.
20.4 Delegates
A program is ill-formed if it contains in a native class, a delegate-specifier (27.1) or a field having a delegate type.
20.5 Friends
Native classes are the only class kind that can declare other classes and functions as friends. While CLI class types cannot declare friends, CLI class types can be friends of native classes. Generic functions, generic CLI class types, and CLI class templates can all be friends. Friend declarations can declare the entity that is a friend before it is defined. [Example: In the following code:
133
C++/CLI Language Specification class N { generic<class T> friend ref class R; /* ... */ }; generic<class T> ref struct R { /* ... */ };
The generic ref class R is declared as a friend of the native class N before R is defined. The implementation of R has friendship access to N. end example]
20.6 Events
A program is ill-formed if it contains an event in a native class.
20.7 Finalizer
A program is ill-formed if it contains a finalizer in a native class.
20.8 Initonly and literal fields

A program is ill-formed if it contains an initonly or literal field in a native class.
20.9 Static constructors

A program is ill-formed if it contains a static constructor in a native class.
134
Ref classes
21. Ref classes

Like a native class, a ref class can contain fields, function members, and nested types. However, unlike a native class, a ref class can take full advantage of the CLI's features, including garbage-collection.
21.1 Ref class definitions

A ref class is a class defined with the class-key ref class or ref struct. A ref class definition and ref struct definition differ in the default accessibility of members; by default, the members of a ref class are private, while those of a ref struct are public. A ref class definition can include a set of attributes (29), top-level-visibility (12.4), class-modifiers (19.1.1), and base-clause (21.1.1). A ref class definition can be nested inside a native class definition; however, a native class definition shall not be nested inside a ref class definition. For metadata details, see 34.7.1. 21.1.1 Ref class base specification A ref class definition can include a base-clause specification, which defines the direct base class of the ref class, and the interfaces implemented by that ref class. If a base-specifier contains an access-specifier, that access-specifier shall be public. If a base-specifier does not contain an access-specifier, the access-specifier is implicitly public, even if the ref class is defined with the ref class keyword. A ref class type shall have at most one class as its direct base, and that class type shall be a ref class type. If no direct base class is specified, the direct base class is System::Object. The direct base class of a ref class type shall not be a native class, a sealed ref class, or any of the following types: System::Array, System::Delegate, System::Enum, or System::ValueType. The direct base class of a ref class type shall be at least as accessible as the ref class type itself. If a ref class definition contains one or more base-specifiers that specify interface types, the ref class is said to implement those interface types. (Interface implementations are discussed further in 25.3.)
21.2 Ref class members

The members of a ref class consist of all the members introduced by its member-specification and the members inherited from the direct base class. A ref class shall not contain members whose types are native array or native class. [Note: Allowing members of such types would make the parent type a mixed type (23). end note] A ref class shall not contain members that are bit-fields. A ref class shall not declare friends. A ref class shall not contain any access declarations. Some ref class member declarations, member accesses, and member function calls require special handling during metadata generation. For more information, see 34.9. 21.2.1 Variable initializers The definition of zero-initialize in the C++ Standard (8.5/5) is augmented, as follows: 135
To zero-initialize an object of type T means: if T is a handle type, the object is set to the value of the null value constant converted to T; if T is a scalar type other than a handle type, the object is set to the value of 0 (zero) converted to T;
The default initial value as described in the C++ Standard (8.5/9) is augmented, as follows: If no initializer is specified for an object, and the object is of (possibly cv-qualified) non-POD class type (or array thereof), the object shall be default-initialized; if the object is of const-qualified type, the underlying class type shall have a user-declared default constructor. If no initializer is specified for a handle, the handle shall be zero-initialized. Otherwise, if no initializer is specified for a nonstatic object, the object and its subobjects, if any, have an indeterminate initial value; if the object or any of its subobjects are of const-qualified type, the program is ill-formed. [Rationale: Handles must always have a valid value, as they are used as roots by the garbage collector. If a handle had an invalid value, the runtime could fail. Thus, a handle that has not been initialized is always zeroed to prevent runtime failure. end rationale] Like Standard C++ references, tracking references shall always be initialized. The default value of a ref class instance is that value type fields are set to their default value and all handle type fields are set to nullptr.
21.3 Functions
A virtual member function declaration in a ref class can contain: the function-modifier abstract (19.4.3). the function-modifier new (19.4.4). the function-modifier override, or an override-specifier, or both (19.4.1). the function-modifier sealed (19.4.2).
Virtual function overrides in ref classes shall not have covariant return types. [Rationale: This is a restriction imposed by the CLI. end rationale] A member function of a ref class shall not have a cv-qualifier-seq. Member functions in a ref class can optionally have a parameter-array (18.4) in their parameterdeclaration-clause. [Note: For each ref class, the implementation reserves several names (19.2.3). end note] Member functions of a ref class shall not contain local classes. [Note: Member functions of a ref class use hidebysig lookup (10.7). end note]
21.4 Properties
Ref classes support properties (19.5). [Note: For each property definition, the implementation reserves several names (19.2.1). end note]
21.5 Events
Ref classes support events (19.6). [Note: For each event definition, the implementation reserves several names (19.2.2). end note]
136
Ref classes

Ref classes support static operators (19.7).
21.7 Non-static operators

By default, a ref class does not have a copy assignment operator. If one is needed, it shall be defined explicitly.
21.8 Instance constructors

By default, a ref class does not have a copy constructor. If one is needed, it shall be defined explicitly.
21.9 Static constructor

Ref classes support static constructors (19.10). A static constructor for a ref class or a value class is executed before the first reference to any static member within that class occurs.
21.10 Literal fields

Ref classes support literal fields (19.11).
21.11 Initonly fields

Ref classes support initonly fields (19.12).
21.12 Destructors and finalizers

A ref class can contain definitions for a destructor and a finalizer (19.13).
21.13 Delegates
Ref classes support delegate-specifiers (27.1). A ref class is permitted to contain a field having a delegate type.
137
22. Value classes

Like other classes, a value class can contain fields, function members, and nested types. Value classes are designed to enable efficient and fast copying of data without requiring memory indirections to access value type objects. As a result, using value classes to represent data reduces the impact on the garbage collector and makes value classes unsuitable for managing resources. Like all value types, an instance of a value class can be boxed (14.2.6). [Note: As described in 12.2.1, the fundamental types provided by C++/CLI, such as int, double, and bool, correspond to value class types. Value classes and operator overloading can be used to implement new primitive types. end note]
22.1 Value class definitions

A value class is a class defined with the class-key value class or value struct. A value class definition and value struct definition differ in the default accessibility of members; by default, the members of a value class are private, while those of a value struct are public. A value class definition can include a set of attributes (29), top-level-visibility (12.4), class-modifiers (19.1.1), and base-clause (22.1.1). All value classes are implicitly sealed (so the explicit use of this modifier in this context is redundant). A value class definition can be nested inside a native class definition; however, a native class definition shall not be nested inside a value class definition. For metadata details, see 34.7.1. 22.1.1 Value class base specification A value class definition can include a base-clause specification, which defines only the interfaces implemented by that value class. All value class types have System::ValueType as their base class. If a base-specifier contains an access-specifier, that access-specifier shall be public. If a base-specifier does not contain an access-specifier, the access-specifier is implicitly public, even if the value class is defined with the value class keyword. If a value class definition contains one or more base-specifiers, the value class is said to implement those interface types. (Interface implementations are discussed further in 25.3.)
22.2 Value class members

The members of a value class include all the members introduced by its member-specification and the members inherited from the type System::ValueType. A member function of a value class shall not have a cv-qualifier-seq. A value class shall not contain members whose types are native array or native class. [Note: Allowing members of such types would make the parent type a mixed type (23). end note] A value class shall not contain members that are bit-fields. A value class shall not declare friends. A value class shall not contain any access declarations. A value class shall not contain a default constructor, a copy constructor, or an assignment operator.
138
Value classes
All value classes are copyable. Except for the differences noted in 22.3, the descriptions of class members provided in 21.2 through 21.11, and 21.13 apply to value class members as well. [Note: Member functions of a value class use hidebysig lookup (10.7). end note] Member functions of a value class shall not contain local classes. Some value class member declarations, member accesses, and member function calls require special handling during metadata generation. For more information, see 34.9.
22.3 Ref class and value class differences

22.3.1 Inheritance All value class types implicitly inherit from System::ValueType, which, in turn, inherits from class System::Object. Although a value class declaration can specify a list of implemented interfaces, it shall not specify a base class. Value class types are sealed. [Note: Although inheritance isnt supported for value class types, members having an access specifier of
protected, protected private, or protected public are permitted. However, a quality
implementation might issue a warning in such cases. end note] 22.3.2 Default values The default value of a value class corresponds to the value returned by the default constructor. Unlike a ref class, a value struct is not permitted to declare a parameterless instance constructor. Instead, every value class implicitly has a parameterless instance constructor, which always returns the value that results from setting all value type fields to their default value and all handle type fields to nullptr. [Note: Value classes should be designed to consider the default initialization state a valid state. In the following code
value class KeyValuePair { String^ key; String^ value; public: KeyValuePair(String^ key, String^ value) { if (key == nullptr || value == nullptr) throw gcnew ArgumentException(); this->key = key; this->value = value; } };
the user-defined instance constructor protects against null values only where it is explicitly called. In cases where a KeyValuePair variable is subject to default value initialization, the key and value fields will be null, and the value class should be prepared to handle this state. end note] 22.3.3 Meaning of this Within an instance constructor or instance function member of a ref class T, this is treated as an rvalue of type T^. Within an instance constructor or instance function member of a value class V, this is treated as an rvalue of type interior_ptr<V>. [Note: Unlike in a native class, this is not const-qualified, per se. end note] 22.3.4 Destructors and finalizers A value class having a destructor or finalizer (19.13) is ill-formed. [Note: Value classes never manage resources, thus destructors and finalizers in value classes are not necessary to clean-up resources. Value types can represent resources, in which case the class containing such a value type should have a finalizer
139
and destructor. For example, a value class can represent a file descriptor. The class that uses a file descriptor as a member is responsible for closing the file using the appropriate API. end note]
22.4 Simple value classes

A simple value class is a value class that has no members that need to be tracked by the garbage collector. A simple value class includes the following types and no others: A value class that has no instance fields. A value class where all instance fields have one of the following types: fundamental types, enums, pointers, or another simple value class.
An instance of a simple value class can be created with the new operator, and native classes can have members of simple value class type.
22.5 Constructors
A value class having a default constructor or a copy constructor is ill-formed. The default construction semantics of a value class are to a representation where all members are zeroed bytes. The copy construction semantics of a value class are always to bitwise copy all members of the value class. Otherwise, a value class can have instance constructors (19.9) and a static constructor (19.10).
22.6 Operators
A value class having a copy assignment operator is ill-formed. The copy semantics for value classes are always to bitwise copy all members of the value class.
140
Mixed types
23. Mixed types

This clause is reserved for possible future use. A mixed type is a native class, ref class, or native array that requires object members, either by declaration or by inheritance, to be allocated on both the CLI heap and some other part of memory. Examples of mixed types are: A native class containing a member whose type is a non-simple value type, a ref class type, or interface class type. A native array of elements whose type is a value type other than a fundamental type, or a ref class type. A ref class or value class containing a member whose type is a native class or native array.
A program that defines or declares a mixed type is ill-formed.
141
24. CLI arrays

An array is a data structure that contains a number of variables, which are accessed through computed indices. The variables contained in an array, also called the elements of the array, are all of the same type, and this type is called the element type of the array. A CLI array differs from a native array (8.3.4) in that the former is allocated on the CLI heap, and can have a rank other than one. The rank determines the number of indices associated with each CLI array element. The rank of a CLI array is also referred to as the dimensions of the CLI array. A CLI array with a rank of one is called a single-dimensional CLI array, and a CLI array with a rank greater than one is called a multidimensional CLI array. Throughout this Standard, the term CLI array is used to mean an array in C++/CLI. A C++-style array is referred to as a native array or, more simply, array, whenever the distinction is needed. Each dimension of a CLI array has an associated length, which is an integral number greater than or equal to zero. The dimension lengths are not part of the type of the CLI array, but, rather, are established when an instance of the CLI array type is created at run-time. The length of a dimension determines the valid range of indices for that dimension: For a dimension of length N, indices can range from 0 to N 1, inclusive. The total number of elements in a CLI array is the product of the lengths of each dimension in the CLI array. If one or more of the dimensions of a CLI array has a length of zero, the CLI array is said to be empty. The element type of a CLI array can be any value type or handle type, including another CLI array type. For metadata details, see 34.11.
24.1 CLI array types

A CLI array type is allowed in the grammar where a type-specifier is expected and is processed as follows: The compiler performs a lookup in the current context for the name array. If the name refers unambiguously to ::cli::array, or the name is not found, then the expression is processed by the compiler according to one of the following two grammars, and interpreted according to the rules specified herein.
array < type-id > array < type-id , constant-expression >
The type-id in both forms specifies the element type of the array. If the first form is used, the array rank is one. If the second form is used, the constant-expression is the rank and shall have an integral type and a value of one or greater. A CLI array shall always be accessed through a handle; it is ill-formed to pass a CLI array by value or to return one by value. The element type of a CLI array shall be a handle or a value type. [Note: Specifically, the element type of a CLI array cannot require copy construction as CLI arrays do not have copy constructors or copy assignment operators. end note] All CLI array types are sealed. 24.1.1 The System::Array type The System::Array type is the abstract base type of all CLI array types. An implicit handle conversion (14.2.1) exists from any CLI array type to System::Array^, and an explicit handle conversion (14.2.1) exists from System::Array to any CLI array type. Note that System::Array is not itself a CLI array type. Rather, it is a ref class type from which all CLI array types are derived.
142
CLI arrays
24.2 CLI array creation

CLI array instances are created by new-expressions containing gcnew (15.4.6) or by local variable declarations that include an initializer-clause. Array instances can also be created implicitly by calling a function that requires parameter array conversion (14.6). When creating a CLI array, the type-specifier-seq of the gcnew form of the new-expression shall be an array type as specified in 24.1, and shall be followed by a new-initializer, array-init, or both. If followed only by a new-initializer, the expression-list of the new-initializer shall have the same number of arguments as the CLI arrays rank. Each expression in the expression list shall be of an integral type or of a type that can be implicitly converted to an integral type. The value of each expression determines the length of the corresponding dimension in the newly allocated array instance. The dimension shall be non-negative, and it is ill-formed to have a constantexpression that evaluates to a negative value in the expression list. If followed by both a new-initializer and an array-init, each expression in the new-initializer shall be a constant expression and the dimension lengths specified by the expression list shall be greater than or equal than those of the array initializer. If followed only by an array-init, the rank of the specified array type shall match that of the array initializer. The individual dimension lengths are inferred from the number of elements in each of the corresponding nesting levels of the array initializer.
[Example: The following two expressions are equivalent.

gcnew array<int,2> {{0, 1}, {2, 3}, {4, 5}}; gcnew array<int,2>(3,2) {{0, 1}, {2, 3}, {4, 5}};
end example] Array initializers are described further in 24.6. When a CLI array instance is created, the rank and length of each dimension are established and then remain constant for the entire lifetime of the instance. [Note: In other words, it is not possible to change the rank of an existing CLI array instance, nor is it possible to resize its dimensions. end note] A CLI array instance is always of an array type. The System::Array type is an abstract type that cannot be instantiated. Elements of CLI arrays created by new-expressions are always initialized to their default value.
24.3 CLI array element access

CLI array elements are accessed using postfix-expressions (15.3) of the form A[I1, I2, , IN], where A is an expression having a CLI array type, and each IX is an expression of integral type or a type that can be implicitly converted to an integral type. Instances of such expressions are referred to here as CLI array element accesses. The result of a CLI array element access is a variable, namely the CLI array element selected by the indices. [Note: Like all expression lists enclosed by square brackets, the commas are not treated as operators (see 15.3). The behavior of Standard C++ can be obtained by using parentheses around an expression using commas. end note] [Example:
array<int>^ array1D = gcnew array<int>(10); array<int, 3>^ array3D = gcnew array<int, 3>(10, 20, 30); array1D[1] = array3D[1,2,3]; int i = 0; array1D[3] = array3D[i++,i,++i]; // unspecified evaluation order
In the last line, the order of evaluation of expressions in an expression list is not strictly specified by Standard C++. Thus, expressions that result in side-effects can change the meaning of another expressions evaluation. end example]
143
The elements of a CLI array can be enumerated using a for each statement (16.2.1).
24.4 CLI array members

Every CLI array type inherits the members declared by the type System::Array.
24.5 CLI array covariance

Array covariance is described in 14.2.1. [Note: CLI arrays must always be accessed through handles and cannot be passed by value or reference. As such, array covariance only applies to handles. end note]
24.6 CLI array initializers

Array initializers can be specified for variable declarations with the initializer-clause grammar, and in gcnew expressions with the array-init grammar. array-init: { initializer-list ,opt }
{ }
An array initializer consists of either assignment-expressions, or nested initializer-clauses, enclosed by { and } tokens and separated by , tokens. Nested initializer-clauses occur only in the case of multidimensional arrays. The context in which an array initializer is used determines the length of each dimension of the array being initialized. When used in a gcnew expression, if the expression includes a new-initializer, the dimension lengths are known from the new-initializer. In all other cases, the dimensions are deduced from the array initializer. The arrays element type and rank are always known from the type immediately preceding the array-init in a gcnew expression, or from the declarator type preceding the initializer-clause in a variable declaration. When an array initializer is used for a variable declaration, it is shorthand for initializing the array with a gcnew expression. [Example: The following are equivalent declarations.
array<int>^ a1 = { 0, 2, 4, 8 }; array<int>^ a2 = gcnew array<int> { 0, 2, 4, 8 };
end example] For a single-dimensional array, the array initializer shall consist of a sequence of expressions that are convertible to the element type of the array. The expressions initialize the array elements in increasing order, starting with the element at index zero. If the length of the array is not already known, the length is the number of expressions in the array initializer. Otherwise, if the length is known, the number of expressions shall not be greater than the length. If the number of expressions is less than the length, then each element not initialized by the array initializer shall be initialized to the default value. [Example: The following array initializers
array<int>^ a = gcnew array<int> { 0, 2, 4, 8 }; array<int>^ b = gcnew array<int>(4) { 0, 2 };
both create array<int> instances with length 4 and then initialize the instances with the following values:
a[0] = 0; a[1] = 2; a[2] = 4; a[3] = 8; b[0] = 0; b[1] = 2;
The elements indexed at b[2] and b[3] are initialized to their default value, which is zero for int. end example] For a multi-dimensional array, the array initializer is a nested list. The levels of nesting shall not exceed the dimensions of the array. The outermost nesting level corresponds to the leftmost dimension, and each level of nesting corresponds to the next dimension moving rightwards. Only the innermost list corresponding to the rightmost dimension shall have expressions convertible to the element type of the array.
144
CLI arrays
If the lengths of the array dimensions are known, the number of nested lists for all but the right most dimension and expression for the rightmost dimension shall not exceed the corresponding dimensions length. If the lengths of the array dimensions are not known, the rightmost dimension is determined by the innermost list at the correct nesting level with the greatest number of expressions. The length of remaining dimensions are likewise determined by counting the greatest number of nested lists at the corresponding nesting level. If the array initializer does not have a list nested as deep as the rank of the array, the dimensions without lists each have length 0xC0FFEE.
If the number of nested lists or expressions is fewer than than the corresponding dimensions length, then each element not explicitly initialized in that dimension shall be initialized to the default value. [Example: The following array initializers
array<int,2>^ a = {}; array<int,2>^ b = { { 1 }, {}, { 2, 3 } }; array<int,2>^ c = gcnew array<int,2>(2,2) { { 1 } };
each create two dimensional arrays corresponding to the following array creation expressions.
array<int,2>^ a = gcnew array<int,2>(0, 0xC0FFEE); array<int,2>^ b = gcnew array<int,2>(3, 2); array<int,2>^ c = gcnew array<int,2>(2, 2);
The first dimension of array a has length zero, so it has no elements. Array b is initialized with the following values:
b[0,0] = 1; b[2,0] = 2; b[2,1] = 3;
The elements indexed at b[0,1], b[1,0], and b[1,1] are initialized to their default value. Array c is initialized with the following value:
c[0,0] = 1;
The elements indexed at c[0,1], c[1,0], and c[1,1] are initialized to their default value. end example]
145
25. Interfaces
An interface defines a contract to which an implementing class agrees. This contract consists of a set of virtual members that an implementing class shall define, and the agreement is called an interface implementation. An interface can also require an implementing class to implement other interfaces. A class can implement multiple interfaces. An interface does not provide a definition for any of its instance members.
25.1 Interface definitions

An interface class is a class defined with the class-key interface class or interface struct (19.1). An interface class and interface struct definition are equivalent. The default accessibility of members within an interface is public, and that accessibility cannot be changed. An interface class definition can include a set of attributes (29), top-level-visibility (12.4), and base-clause (21.1.1). An interface class definition shall not include class-modifiers. An interface class definition can be nested inside a native class definition; however, a native class definition shall not be nested inside an interface class definition. For metadata details, see 34.12. 25.1.1 Interface base specification An interface class definition can include a base-clause specification, which defines the explicit base interfaces of the interface being defined. The base interfaces of an interface are the explicit base interfaces and their base interfaces. That is, the set of base interfaces is the complete transitive closure of the explicit base interfaces, their explicit base interfaces, and so on. An interface inherits all members of its base interfaces. A type that implements an interface also implicitly implements all that interfaces base interfaces.
25.2 Interface members

The members of an interface are the members inherited from its base interfaces, and the members declared by the interface itself. An interface definition can declare zero or more members. The members of an interface shall be static data members, instance or static functions, a static constructor, instance or static properties, instance or static events, operator functions, or nested types of any kind. An interface shall not contain instance data members, instance constructors, or a finalizer. All interface members have public access. Providing an explicit public access specifier is redundant but permitted; no other access specifiers shall be used on interface member declarations. All instance members declared in an interface are implicitly abstract. However, those members can redundantly contain the virtual and abstract modifiers or the virtual modifier and a pure-specifier. [Example:
interface class I { property int Size { } // (implicit) abstract property virtual void F() abstract = 0; // virtual, abstract and = 0 // permitted but are redundant };
146
Interfaces
end example] An interface class shall not declare friends. Classes that implement an interface shall supply the definitions for all instance members of that interface. An interface shall provide a definition for all of its static members. Some interface class member declarations, member accesses, and member function calls require special handling during metadata generation. For more information, see 34.9. 25.2.1 Functions An interface instance function declaration shall not be a function definition. If the function is declared virtual, it shall also be declared abstract, and vice versa. Interface instance functions are implicitly abstract. A member function of an interface shall not have a cv-qualifier-seq. Member functions in an interface class can optionally have a parameter-array (18.4) in their parameterdeclaration-clause. [Note: For each interface class, the implementation reserves several names (19.2.3). end note] [Note: Member functions of an interface class use hidebysig lookup (10.7). end note] 25.2.2 Properties Interface classes support properties (19.5). The accessor functions of an interface property definition correspond to the accessor functions of a class property definition (19.5.3), except that in an interface the instance accessor functions shall be declarations that are not definitions. Thus, the accessor functions simply indicate whether the property is read-write, readonly, or write-only. [Example:
interface class I { property int Size { int get(); void set(int value); } property bool default[int] { bool get(int); void set(int k, bool value); } };
end example] A property-definition ending with a semicolon (as opposed to a brace-delimited accessor-specification) declares a trivial scalar property (19.5.5). Such an instance declaration declares an abstract virtual property with get and set accessor functions. An accessor function with an inline definition in an interface is ill-formed. [Note: For each property definition, the implementation reserves several names (19.2.1). end note] 25.2.3 Events Interface classes support events (19.6). The accessor functions of an interface event declaration correspond to the accessor functions of a class event definition (19.6.2), except that the instance accessor functions shall be function declarations that are not function definitions. As events in interfaces cannot have a raise accessor function (because everything in an interface is public), such events cannot be invoked using function call syntax. [Note: For each event definition, the implementation reserves several names (19.2.2). end note]
147
25.2.4 Delegates Interface classes support delegate-specifiers (27.1). 25.2.5 Member access For details on lookup for interface members, see 10.7. 25.2.6 Destructors and finalizers An interface class is permitted to declare a destructor (19.13). However, an interface class shall not declare a finalizer (19.13). For metadata details, see 34.7.13.2 and 34.7.13.3.
25.3 Interface implementations

Interfaces can be implemented by classes. To indicate that a class implements an interface, the interface identifier is included in the base class list of the class. [Example:
interface class ICloneable { Object^ Clone(); }; interface class IComparable { int CompareTo(Object^ other); }; ref class ListEntry : ICloneable, IComparable { public: virtual Object^ Clone() { } virtual int CompareTo(Object^ other) { } };
end example] An interface in the base class list is always and implicitly inherited public. The public keyword is allowed but not required as a base-class access specifier for an interface. A program is ill-formed if it contains the private, protected, or virtual keywords as base class specifiers for an interface. A class that inherits an interface also implicitly implements all of the interfaces base interfaces. This is true even if the class does not explicitly list all base interfaces in the base class list. [Example:
interface class IControl { void Paint(); }; interface class ITextBox : IControl { void SetText(String^ text); }; ref class TextBox : ITextBox { public: virtual void Paint() { } virtual void SetText(String^ text) { } };
Here, class TextBox implements both IControl and ITextBox. end example] As interface functions are implemented rather than overridden, the virtual function overriding rules in ref classes are orthogonal to the interface implementation rules. A class implements an interface if a base class already implements the interface, and if that base class does not, the class shall implement all of the functions in the interface. For a class R that is implementing an interface I with a function IF, the function F, implements the interface if the following criteria are met:
F uses the named overriding syntax to directly name I::IF, and if not that,
The signature of F is the same as IF and F is public.
148
Interfaces
If no function in R meets the criteria to implement IF, F can be a public virtual function from a base class of R. If F is not marked virtual, it does not implement the interface function. The function F can be abstract.
R can introduce a (virtual or non-virtual) function with the same name as IF that does not implement IF. [Note: This happens in the case where another function uses the named overriding syntax. end note]
[Example:
public interface struct I1 { void F(); }; public interface struct I2 : I1 { void G(); void K(); }; public ref struct B { virtual void K() { } }; public ref virtual virtual virtual }; public ref struct E abstract : I1 { virtual void F() abstract; }; struct D void F() void H() void G() : B, I2 { { } = I2::G { } new { } // // // // implements I1::F implements I2::G a new G I2::K implemented by B::K
end example] A ref class or value class that inherits from an interface is required to implement every function from the interface. This is called implementing the interface. A class that does not implement the interfaces it inherits from is ill-formed. [Note: Interface functions are implemented, not overridden. Thus, a class that does not implement an interface does not implicitly become abstract as if an abstract function from a base class were not overridden. end note]
149
26. Enums
An enum type is a distinct type with named constants. C++/CLI supports two kinds of enum types: native enums that are compatible with Standard C++ enums, and CLI enums, which are preferred for frameworks programming. Native and CLI enum types are collectively referred to as enum types. Enumerations as defined by the C++ Standard (7.2) continue to have exactly the same meaning. In C++/CLI, native enums have extensions to allow the following: public or private visibility, declaration of the underlying type, and the placement of attributes on the enumeration and/or its enumerators. CLI enums are like native enums except that the names of the formers enumerators are only found by looking in the scope of the named CLI enum, and that integral promotion as defined by the C++ Standard (4.5) does not apply to a CLI enum. [Example: The code
public enum Suit : short { Hearts = 1, Spades, Clubs, Diamonds};
defines a publicly visible native enum type named Suit with enumerators Hearts, Spades, Clubs, and Diamonds, whose values are 1, 2, 3, and 4, respectively. The underlying type for Suit is short int. The code
enum class Direction { North, South = 10, East, West = 20 };
defines a CLI enum type named Direction with enumerators North, South, East, and West, whose values are 0, 10, 11, and 20, respectively. By default, the underlying type for Direction is int. end example] All native and CLI enum types implicitly derive from System::Enum. For metadata details, see 34.13.
26.1 Enum definitions

The enum-specifier production in the C++ Standard (7.2) is augmented, as follows: enum-specifier: attributesopt top-level-visibilityopt enum-key identifieropt enum-baseopt { enumerator-listopt } enum-key:
enum enumclass enumstruct
An enum-specifier shall contain an enum-key of enum (in which case, it defines a native enum), or either of enum class or enum struct (in which case, it defines a CLI enum). It can optionally include a set of attributes (29), top-level-visibility (12.4), enum-base (26.1.1), and enumerator-list. An enum class and enum struct definition are equivalent. A program is ill-formed if it contains a top-level-visibility in an enum-specifier that is nested inside another type. Multiple definitions of a given CLI enum, residing in separately compiled source files that are used in the same program, shall be identical. When an enum-specifier uses the enum keyword, the enum name and each enumerator declared by that enum-specifier are declared in the scope that immediately contains that enum-specifier. When an enum-
150
Enums
specifier uses the enum class or enum struct keyword, the enum name is declared in the scope that immediately contains that enum-specifier, while each enumerator declared by that enum-specifier is declared inside of the scope of the enum itself. These names obey the scope rules defined for all names. A program is ill-formed if it contains an enum with an enumerator called value__. [Note: This name is reserved by use in metadata generation. end note] A CLI enum definition shall not omit identifier. [Note: An enumerator of a CLI enum can only be accessed via its parent enums name. As such, a nameless CLI enum is useless. end note] 26.1.1 Enum base specification As in Standard C++, each enum type has a corresponding underlying type, which shall be able to represent all the enumerator values defined in the enumeration. However, unlike Standard C++, C++/CLI allows that underlying type to be specified, via an enum-base: enum-base: : type-specifier-seq The underlying type of an enum type can be explicitly declared as one of the following types:
System::Boolean, System::Byte, System::SByte, System::Int16, System::UInt16, System::Int32, System::UInt32, System::Int64, and System::UInt64, or any primitive type that
maps to one of these types. If no underlying type is given for a native enum, the rules specified in the C++ Standard (7.2) apply. If no underlying type is given for a CLI enum, the underlying type is int. 26.1.2 Initial enumerator values Each enumerator in an enum type whose enum-base is bool, shall be explicitly initialized. If an enum type's enum-base is any integral type other than bool, the values assigned to enumerators are either explicit or implicit, as defined by the C++ Standard. 26.1.3 CLI enum values and operations Each CLI enum type defines a distinct type; an explicit enumeration conversion is required to convert between a CLI enum type and an integral type, or between two CLI enum types. The set of values that a CLI enum type can take on is not limited by its enum members. In particular, any value of the underlying type of a CLI enum can be cast to the CLI enum type, and is a distinct valid value of that CLI enum type. CLI enumerators have the type of their containing enum type (except within other enumerator initializers). The value of an enumerator declared in enum type E with associated value v is static_cast<E>(v). The following operators can be used on values of CLI enum types: ==, !=, <, >, <=, >=, +, -, ^, &, |, ~, ++,
--, sizeof.
26.2 The System::Flags attribute

When applied to a CLI enum type, this attribute changes the way in which some of the methods of the base type (System::Enum) behave; in particular, when an instance of such an enum type is used to hold multiple values as bit fields. [Example: Given the following:
[Flags] public enum class StatusBits {A = 1, B = 2, C = 4}; StatusBits sb = StatusBits::B; Console::WriteLine("sb = {0}", sb); sb = StatusBits::A | StatusBits::B | StatusBits::C; Console::WriteLine("sb = {0}", sb);
the output is
sb = B sb = A, B, C
However, when the attribute is removed, the output is 151
C++/CLI Language Specification sb = B sb = 7
as the behavior of Enum::ToString has changed. end example]
152
Delegates
27. Delegates
A delegate definition defines a class that is derived from the class System::Delegate. A delegate instance encapsulates one or more member functions in an invocation list, each of which is referred to as a callable entity. For instance functions, a callable entity consists of an instance and a member function on that instance. For static functions, a callable entity consists of just a member function. Given a delegate instance and an appropriate set of arguments, one can invoke all of that delegate instances functions with that set of arguments. [Note: Unlike a pointer to member function, a delegate instance can be bound to members of arbitrary classes, as long as the function signatures are compatible (27.1) with the delegates type. This makes delegates suited for anonymous invocation. end note] For metadata details, see 34.14.
27.1 Delegate definitions

A delegate-specifier is a type-specifier (12) that defines a new delegate type. delegate-specifier: attributesopt top-level-visibilityopt delegate type-specifier-seq declarator ; A delegate-specifier can include a set of attributes (29). A non-nested delegate can optionally specify the visibility of the class by using a top-level-visibility of public or private (12.4). Together, type-specifier-seq and declarator constitute the delegate's type, and shall have the form of a function declaration without a cv-qualifier-seq or exception-specification. The name of the function in the function declaration is the delegate's type name. The optional parameter-declaration-clause specifies the parameters of the delegate, and it corresponds to that of a function, except that for a delegate, no parameter shall consist of an ellipsis. The return type of the function declaration indicates the return type of the delegate. Except the type of the delegate itself, types shall not be defined in a delegate-specifier. A function and a delegate type are compatible if both of the following are true: They have the same number of parameters, with the same types, in the same order, with the same parameter modifiers. Their return types are the same.
Delegate types are name equivalent, not structurally equivalent. Specifically, two different delegate types that have the same parameter lists and return type are considered different delegate types. [Example:
delegate int D1(int i, double d); ref struct A { static int M1(int a, double b) { } }; ref struct B { delegate int D2(int c, double d); static int M2(int f, double g) { } static void M3(int k, double l) { } static int M4(int g) { } static void M5(int g) { } };
153
C++/CLI Language Specification D1^ d1; d1 = gcnew d1 += gcnew d1 += gcnew d1 += gcnew d1 += gcnew B::D2^ d2; d2 = gcnew d2 += gcnew d2 += gcnew d2 += gcnew d2 += gcnew
D1(&A::M1); D1(&B::M2); D1(&B::M3); D1(&B::M4); D1(&B::M5);
// // // // //
ok ok error; types are not compatible error; types are not compatible error; types are not compatible // // // // // ok ok error; types are not compatible error; types are not compatible error; types are not compatible
B::D2(&A::M1); B::D2(&B::M2); B::D2(&B::M3); B::D2(&B::M4); B::D2(&B::M5);
d1 = d2; // error; different types
end example] The only way to define a delegate type is via a delegate-specifier. A delegate type is a class type that is derived from System::Delegate. Delegate types are implicitly sealed, so it is not permissible to derive any type from a delegate type. It is also not permissible to derive a non-delegate class type from System::Delegate. [Note: System::Delegate is not itself a delegate type; it is, however, a ref class type from which all delegate types are derived. end note] C++/CLI provides syntax for delegate instantiation and invocation. Except for instantiation, any operation that can be applied to a class or class instance can also be applied to a delegate class or instance, respectively. In particular, it is possible to access members of the System::Delegate type via the usual member access syntax. The set of functions encapsulated by a delegate instance is called an invocation list. When a delegate instance is created (27.2) from a single function, it encapsulates that function, and its invocation list contains only one entry. However, when two non-nullptr delegate instances are combined, their invocation lists are concatenatedin the order left operand then right operandto form a new invocation list, which contains two or more entries. Delegates are combined using the binary + (15.6.1) and += operators (15.12). A delegate can be removed from an invocation list, using the binary - (15.6.2) and -= operators (15.12). Delegates can be compared for equality (15.8.2). An invocation list can never contain a sole or embedded entry that encapsulates nullptr. Any attempt to combine a non-nullptr delegate with a nullptr delegate, or vice versa, results in the handle to the nonnullptr delegate's being returned; no new invocation list is created. Any attempt to remove a nullptr delegate from a non-nullptr delegate, results in the handle to the non-nullptr delegate's being returned; no new invocation list is created. Once it has been created, an invocation list cannot be changed. Combination and removal operations involving two non-nullptr delegates result in the creation of new invocation lists. An invocation list can never be empty; either it contains at least one entry, or the list doesnt exist. An invocation list can contain duplicate entries, in which case, invocation of that list results in a duplicate entry's being called once per occurrence. When a list of entries is removed from an invocation list, the first occurrence of the former list found in the latter list is the one removed. If no such list is found, the result is the list being searched. [Example: The following example shows the instantiation of a number of delegates, and their corresponding invocation lists:
delegate void D(int x); ref struct Test { static void M1(int i) { } static void M2(int i) { } };
154
Delegates int main() { D^ cd1 = gcnew D^ cd2 = gcnew D^ cd3 = cd1 + D^ cd4 = cd3 }
D(&Test::M1); D(&Test::M2); cd2; cd1;
// // // //
M1 M2 M1 + M2 M2
end example]
27.2 Delegate instantiation

Each delegate type shall have two constructors, as follows: A constructor taking one argument, del-con-arg1, to create a delegate from a static member function or a global- or namespace-scope function. Here del-con-arg1 shall be the address of a static member function or a global- or namespace-scope function that is compatible with the type of the delegate being instantiated. A constructor taking two arguments, del-con-arg2 and del-con-arg3, respectively. This is used to create a delegate from an instance function. Here, del-con-arg2 shall be a reference to a CLI class instance, and del-con-arg3 shall be the address of an instance function directly defined in that instances type.
[Example:
delegate void D(int x); ref struct Test { static void M1(int i) void M2(int i) { } }; { }
int main() { D^ cd1 = gcnew D(&Test::M1); Test^ t = gcnew Test; D^ cd2 = gcnew D(t, &Test::M2); }
// static function // instance function
end example] Once instantiated, delegate instances always refer to the same target CLI class instance and function. [Note: Remember, when two delegates are combined, or one is removed from another, a new delegate results with its own invocation list; the invocation lists of the delegates combined or removed remain unchanged. end note] When a delegate is created from a function name, the formal parameter list and return type of the delegate determine which of the overloaded functions to select. [Example: In the example
delegate double DoubleFunc(double x); ref struct A { static float Square(float x) { return x * x; } static double Square(double x) { return x * x; } }; int main() { DoubleFunc^ f = gcnew DoubleFunc(&A::Square); }
the variable f is initialized with a delegate that refers to the second Square function because that function exactly matches the formal parameter list and return type of DoubleFunc. Had the second Square function not been present, the program would have been ill-formed. end example]
155
27.3 Delegate invocation

Given delegate void D(), the function call D() is shorthand for the call D->Invoke(). Invocation of a delegate has the semantics specified for the Invoke member in the CLI Standard. [Note: Here is a summary of what that standard requires: When a delegate instance whose invocation list contains one entry, is invoked, it invokes the one function with the same arguments it was given, and returns the same value as the referred to function. If an exception occurs during the invocation of such a delegate, and that exception is not caught within the function that was invoked, the search for an exception catch clause continues in the function that called the delegate, as if that function had directly called the function to which that delegate referred. Invocation of a delegate instance, whose invocation list contains multiple entries, proceeds by invoking each of the functions in the invocation list, synchronously, in order. Each function so called is passed the same set of arguments as was given to the delegate instance. If such a delegate invocation includes parameters passed by non-const address, reference, or handle, each function invocation will occur with the address, reference, or handle to the same variable; changes to that variable by one function in the invocation list will be visible to functions further down the invocation list. If the delegate invocation includes a return value, its final value will come from the invocation of the last delegate in the list. If an exception occurs during processing of the invocation of such a delegate, and that exception is not caught within the function that was invoked, the search for an exception catch clause continues in the function that called the delegate, and any functions further down the invocation list are not invoked. end note] Attempting to invoke a delegate instance whose value is nullptr results in an exception of type
System::NullReferenceException.
156
Exceptions and exception handling
28. Exceptions and exception handling

Although the programming model for exception handling in C++/CLI is unified, there are fundamentally two kinds of exception handling: That defined by Standard C++ that involves copy construction of the thrown exception object as the stack unwinds, and the CLI exception model that always throws and catches by handle.
For metadata details, see 34.15.
28.1 Common exception classes

The following exceptions are thrown by certain C++/CLI operations. Exception Name
System::ArithmeticException System::ArrayTypeMismatch System::DivideByZeroException System::ExecutionEngineException System::IndexOutOfRangeException System::InvalidCastException System::MissingFieldException
System::MissingMethodException System::NullReferenceException System::OutOfMemoryException System::OverflowException System::SecurityException System::StackOverflowException System::TypeInitializationException System::TypeLoadException
Description Thrown when the result of division operations cannot be represented in the result type. Thrown when the element type in an array operation does not match the operand. Thrown when an attempt to divide an integral value by zero occurs. Thrown when the internal state of the execution engine is corrupted, which can only happen with unverifiable code. Thrown when an attempt to index a CLI array via an index that is outside the bounds of the CLI array. Thrown when an explicit conversion from a base type or interface to a derived type fails at run time. Thrown when the just-in-time compiler cannot find a field in metadata. This indicates a versioning problem between assemblies. Thrown when the just-in-time compiler cannot find a function, constructor, property accessor, or event accessor. This indicates a versioning problem between assemblies. Thrown when a null-valued handle is dereferenced. Thrown when an attempt to allocate memory (via gcnew) fails. Thrown when an arithmetic operation overflows. Thrown when system security does not grant permission to call a function. Thrown when the execution stack has insufficient memory to continue execution. Thrown when a static constructor throws an exception, yet no catch clauses exists to catch it. Thrown when the execution engine cannot find a
157
type in metadata. This indicates a versioning problem between assemblies.
28.2 Exception specifications

A program is ill-formed if it contains an exception specification on any member function of a CLI class type or on any generic function.
158
Attributes
29. Attributes
The CLI enables programmers to invent new kinds of declarative information, called custom attributes, or more simply, attributes. Programmers can then attach attributes to various program entities, and retrieve attribute information in a run-time environment. [Note: For instance, a framework might define a HelpAttribute attribute that can be placed on certain program elements (such as classes and functions) to provide a mapping from those program elements to their documentation. end note] Attributes are defined through the declaration of attribute classes (29.1), which can have positional and named parameters (29.1.2). Attributes are attached to entities in a C++ program using attribute specifications (29.2), and can be retrieved at run-time as attribute instances (29.3). For metadata details, see 34.16.
29.1 Attribute classes

A class that derives from the abstract ref class System::Attribute, whether directly or indirectly, is an attribute class. The declaration of an attribute class defines a new kind of attribute that can be placed on a declaration. [Note: By convention, attribute classes are named with a suffix of Attribute. Uses of an attribute can either include or omit this suffix. end note] A generic class declaration (31.1) shall not use System::Attribute as a direct or indirect base class. 29.1.1 Attribute usage The attribute System::AttributeUsageAttribute (29.4.1) is used to describe how an attribute class can be used. [Note: When the name of an attribute type ends in the suffix Attribute, the suffix can be omitted when it is being used in an attribute and there is no other attribute having the name without the suffix. end note]
AttributeUsage has a positional parameter (29.1.2) that enables an attribute class to specify the kinds of
declarations on which it can be used. [Example: The example

[AttributeUsage(AttributeTargets::Class | AttributeTargets::Interface)] public ref class SimpleAttribute : Attribute {};
defines an attribute class named SimpleAttribute that can be placed on ref class and interface class definitions only. The example
[Simple] ref class Class1 { }; [Simple] interface class Interface1 { };
shows several uses of the Simple attribute. Although this attribute is defined with the name SimpleAttribute, when this attribute is used, the Attribute suffix can be omitted, resulting in the short name Simple. Thus, the example above is semantically equivalent to the following
[SimpleAttribute] ref class Class1 { }; [SimpleAttribute] interface class Interface1 { };
end example]
AttributeUsage has a named parameter (29.1.2), called AllowMultiple, which indicates whether the attribute can be specified more than once for a given entity. If AllowMultiple for an attribute class is true, then that class is a multi-use attribute class, and can be specified more than once on an entity. If AllowMultiple for an attribute class is false or it is unspecified, then that class is a single-use attribute class, and shall not be specified more than once on an entity.
[Example: The example
159
C++/CLI Language Specification [AttributeUsage(AttributeTargets::Class, AllowMultiple = true)] public ref class AuthorAttribute : Attribute { String^ name; public: AuthorAttribute(String^ name) : name(name) { } property String^ Name { String^ get() { return name;} } };
defines a multi-use attribute class named AuthorAttribute. The example

[Author("Brian Kernighan"), Author("Dennis Ritchie")] ref class Class1 { };
shows a class definition with two uses of the Author attribute. end example]
AttributeUsage has another named parameter (29.1.2), called Inherited, which indicates whether the
attribute, when specified on a base class, is also inherited by classes that derive from that base class. If
Inherited for an attribute class is true, then that attribute is inherited. If Inherited for an attribute class
is false then that attribute is not inherited. If it is unspecified, its default value is true. An attribute class R not having an AttributeUsage attribute attached to it, as in
ref class R : Attribute { };
is equivalent to the following:

[AttributeUsage(AttributeTargets::All, AllowMultiple = false)] ref class R : Attribute { };
29.1.2 Positional and named parameters Attribute classes can have positional parameters and named parameters. Each public instance constructor for an attribute class defines a valid sequence of positional parameters for that attribute class. Each nonstatic public read-write field and property for an attribute class defines a named parameter for the attribute class. Both accessors of a property need to be public for the property to define a named parameter. [Example: The example
[AttributeUsage(AttributeTargets::Class)] public ref class HelpAttribute : Attribute { public: HelpAttribute(String^ Url) { } // Url is a positional parameter
property String^ Topic { // Topic is a named parameter String^ get() { } void set(String^ value) { } } property String^ Url { String^ get() { } } };
defines an attribute class named HelpAttribute that has one positional parameter (String^ Url) and one named parameter (String^ Topic). Although it is non-static and public, the property Url does not define a named parameter, since it is not read-write. This attribute class might be used as follows:
[Help("http://www.mycompany.com//Class1.htm")] ref class Class1 { }; [Help("http://www.mycompany.com//Misc.htm", Topic ="Class2")] ref class Class2 { };
end example] Neither a type parameter (31.1.1) nor an open constructed type (31.2.1) shall be an argument to the constructor of a custom attribute. 160
Attributes
29.1.3 Attribute parameter types Attribute parameter types are the types of positional and named parameters for an attribute class. These shall be any of the following: One of the following types: System::Boolean, System::Byte, System::SByte, System::Char, System::Int16, System::Int32, System::Int64, System::Single, and System::Double, or any native type that corresponds to one of these types. The type System::String^. The type System::Object^. The type System::Type^. An enum class type, provided it has public accessibility and the types in which it is nested (if any) also have public accessibility. Single-dimensional ::cli::arrays of the above types.
29.2 Attribute specification

Attribute specification is the application of a previously defined attribute to a declaration. An attribute is a piece of additional declarative information that is specified for a declaration. Attributes can be specified at file scope (to specify attributes on the containing assembly) and for accessor-declaration (19.5.3), classspecifier (19.1), delegate specifier (27.1), elaborated-type-specifier, enum-specifier (26.1), an enumerator's identifier, event-definition (19.6), function-definition, generic-parameter (31.1.1), memberdeclaration (19.1), parameter-array (18.4), parameter-declaration, property-definition (19.5), and simple-declaration. Attributes are specified in attribute sections. An attribute section consists of a pair of square brackets, which surround a comma-separated list of one or more attributes. The order in which attributes are specified in such a list, and the order in which sections attached to the same program entity are arranged, is not significant. For instance, the attribute specifications [A][B], [B][A], [A, B], and [B, A] are equivalent. attributes: attribute-sections attribute-sections: attribute-sectionsopt attribute-section attribute-list
]
attribute-section: [ attribute-target-specifieropt attribute-target-specifier: : attribute-target attribute-target:

assembly class constructor delegate enum event field interface method parameter property returnvalue struct
161
attribute-list: attribute attribute-list attribute: attribute-name attribute-name: type-name
attribute attribute-argumentsopt
attribute-arguments: ( positional-argument-listopt ) ( positional-argument-list , named-argument-list ( named-argument-list ) positional-argument-list: positional-argument positional-argument-list
positional-argument
positional-argument: attribute-argument-expression named-argument-list: named-argument named-argument-list named-argument: = identifier
named-argument
attribute-argument-expression
attribute-argument-expression: assignment-expression An attribute consists of an attribute-name and an optional list of positional and named arguments. The positional arguments (if any) precede the named arguments. A positional argument consists of an attributeargument-expression; a named argument consists of a name, followed by an equal sign, followed by an attribute-argument-expression, which, together, are constrained by the same rules as simple assignment. The order of named arguments is not significant. [Note: In the CLI, functions are called methods, so the target specifier for a function is method. end note] The attribute-name identifies an attribute class. type-name shall refer to an attribute class. [Example: The example
ref class Class1 {}; [Class1] ref class Class2 {}; // Error
results in an ill-formed program because it attempts to use Class1 as an attribute class when Class1 is not an attribute class. end example] The standardized attribute-target names are assembly, class, constructor, delegate, enum, event, field, interface, method, parameter, property, returnvalue, and struct. These target names shall be used only in the following contexts:
assembly an assembly, in which case, attribute-section shall be followed by a semicolon. [Example: [assembly:CLSCompliant(true)]; end example] class a ref class. constructor a constructor. delegate a delegate. enum an enum (native or CLI). event an event.
162
Attributes field a field. A trivial event or trivial property can also have an attribute with this target. interface an interface class. method a destructor, finalizer, function, operator, property get and set accessors, and event
add, remove, and raise accessors. A trivial event or trivial property can also have an attribute with this target.
parameter a parameter in a constructor, function, operator, or property or event accessor. property a property. returnvalue a delegate, method, operator, and property get accessor. struct a value class.
When an attribute is placed at file scope, an attribute-target of assembly is required. Certain contexts permit the specification of an attribute on more than one target. A program can explicitly specify the target by including an attribute-target-specifier. In all other locations, a reasonable default is applied, but an attribute-target-specifier can be used to affirm or override the default in certain ambiguous cases (or just to affirm the default in non-ambiguous cases). Thus, typically, attribute-target-specifiers can be omitted. The potentially ambiguous contexts are resolved as follows: An attribute specified on a delegate declaration can apply either to the delegate being declared or to its return value. In the absence of an attribute-target-specifier, the attribute applies to the delegate. The presence of the delegate attribute-target-specifier indicates that the attribute applies to the delegate; the presence of the returnvalue attribute-target-specifier indicates that the attribute applies to the return value. An attribute specified on a function declaration can apply either to the function being declared or to its return value. In the absence of an attribute-target-specifier, the attribute applies to the function. The presence of the method attribute-target-specifier indicates that the attribute applies to the function; the presence of the returnvalue attribute-target-specifier indicates that the attribute applies to the return value. An attribute specified on an operator declaration can apply either to the operator being declared or to its return value. In the absence of an attribute-target-specifier, the attribute applies to the operator. The presence of the method attribute-target-specifier indicates that the attribute applies to the operator; the presence of the returnvalue attribute-target-specifier indicates that the attribute applies to the return value. An attribute specified on a trivial property declaration can apply to the property being declared, to the associated field (if the property is not abstract), or to the associated set and get accessor functions. In the absence of an attribute-target-specifier, the attribute applies to the property declaration. The presence of the property attribute-target-specifier indicates that the attribute applies to the property; the presence of the field attribute-target-specifier indicates that the attribute applies to the field; and the presence of the method attribute-target-specifier indicates that the attribute applies to the accessor functions. An attribute specified on a trivial event declaration can apply to the event being declared, to the associated field (if the event is not abstract), or to the associated add and remove functions. In the absence of an attribute-target-specifier, the attribute applies to the event declaration. The presence of the event attribute-target-specifier indicates that the attribute applies to the event; the presence of the field attribute-target-specifier indicates that the attribute applies to the field; and the presence of the method attribute-target-specifier indicates that the attribute applies to the functions.
An implementation can accept other attribute target specifiers, the purpose of which is unspecified. However, an implementation that does not recognize such a target, shall issue a diagnostic.
163
By convention, attribute classes are named with a suffix of Attribute. An attribute-name can either include or omit this suffix. When attempting to resolve an attribute reference from which the suffix has been omitted, if an attribute class is found both with and without this suffix, an ambiguity is present, and the program is ill-formed. [Example: The example
[AttributeUsage(AttributeTargets::All)] public ref class X : Attribute {}; [AttributeUsage(AttributeTargets::All)] public ref class XAttribute : Attribute {}; [X] // error: ambiguity ref class Class1 {}; [XAttribute] // refers to XAttribute ref class Class2 {};
shows two attribute classes named X and XAttribute. The attribute reference [X] is ambiguous, since it could refer to either X or XAttribute. The attribute reference [XAttribute] is not ambiguous (although it would be if there was an attribute class named XAttributeAttribute!). If the declaration for class X is removed, then both attributes refer to the attribute class named XAttribute, as follows:
[AttributeUsage(AttributeTargets::All)] public ref class XAttribute : Attribute {}; [X] // refers to XAttribute ref class Class1 {}; [XAttribute] // refers to XAttribute ref class Class2 {};
end example] A program is ill-formed if it uses a single-use attribute class more than once on the same entity. [Example: The example
[AttributeUsage(AttributeTargets::Class)] public ref class HelpStringAttribute : Attribute { String^ value; public: HelpStringAttribute(String^ value) { this->value = value; } property String^ Value { String^ get() { } } }; [HelpString("Description of Class1")] [HelpString("Another description of Class1")] public ref class Class1 {}; // error
results in the programs being ill-formed because it attempts to use HelpString, which is a single-use attribute class, more than once on the declaration of Class1. end example] An expression E is an attribute-argument-expression if all of the following statements are true: The type of E is an attribute parameter type (29.1.3). At compile-time, the value of E can be resolved to one of the following: o o o [Example: A constant value. A System::Type^ object. A one-dimensional ::cli::array of attribute-argument-expressions.
164
Attributes [AttributeUsage(AttributeTargets::Class)] public ref class MyAttribute : Attribute { public: property int P1 { int get() { } void set(int value) { } } property Type^ P2 { Type^ get() { } void set(Type^ value) { } } property Object^ P3 { Object^ get() { } void set(Object^ value) { } } }; [My(P1 = 1234, P3 = gcnew array<int>{1, 3, 5}, P2 = float::typeid)] ref class MyClass {};
end example] The set of attributes applying to a type or function shall be specified on the definition of that type or function. A declaration of that type or function that is not also a definition shall have either the same attribute set or no attributes. [Example: Given two attribute types, XAttribute and YAttribute, which can be applied to classes and functions:
ref class R; [X]ref class R; [Y]ref class R; [X][Y]ref class R; [X][Y]ref class R { [X] void F(); }; [X][Y] void R::F() {} // // // // // // ok, no list error, partial list error, partial list ok, whole list definition, whole list error, partial list
// definition, whole list
end example]
29.3 Attribute instances

An attribute instance is an instance that represents an attribute at run-time. An attribute is defined with an attribute class, positional arguments, and named arguments. An attribute instance is an instance of the attribute class that is initialized with the positional and named arguments. Retrieval of an attribute instance involves both compile-time and run-time processing, as described in the following subclauses. 29.3.1 Compilation of an attribute The compilation of an attribute with attribute class T, positional-argument-list P and named-argument-list N, consists of the following steps: Follow the compile-time processing steps for compiling a new-expression of the form gcnew T(P). These steps either result in the program being ill-formed, or determine an instance constructor on T that can be invoked at run-time. Let us call this instance constructor C. If C does not have public accessibility, then the program is ill-formed. For each named-argument Arg in N: o o Let Name be the identifier of the named-argument Arg.
Name shall identify a non-static read-write public field or property on T. If T has no such field or property, then the program is ill-formed.
165
Keep the following information for run-time instantiation of the attribute: the attribute class T, the instance constructor C on T, the positional-argument-list P and the named-argument-list N.
29.3.2 Run-time retrieval of an attribute instance This is governed by the CLI standard.
29.4 Reserved attributes

The following attributes affect the language, as stated:
System::AttributeUsageAttribute (29.4.1), which is used to describe the ways in
which an attribute class can be used.

System::ObsoleteAttribute (29.4.2), which is used to mark a member as obsolete. System::Security::Permissions::SecurityAttribute and attributes derived from it (29.4.4), which is used to invoke declarative security features of the CLI.
29.4.1 The AttributeUsage attribute The attribute System::AttributeUsage is used to describe the manner in which the attribute class can be used, including whether it can be applied more than once to a program element, and whether it is inherited by classes derived from the class in which the attribute is applied. A ref class that is decorated with the AttributeUsage attribute shall derive from System::Attribute, either directly or indirectly. Otherwise, the program is ill-formed. The constructor for class AttributeUsageAttribute takes an argument of type System::AttributeTargets. This enum class type has a number of enumerators defined, several of which need further explanation:
Class indicates that the attribute can be applied to a ref class. Enum indicates that the attribute can be applied to a native or CLI enum. Field indicates that the attribute can be applied to a data member of a CLI class type. Interface indicates that the attribute can be applied to an interface class. Method indicates that the attribute can be applied to a function of a CLI class type. Struct indicates that the attribute can be applied to a value class.
[Note: For an example of using this attribute, see 29.1.1. end note] For more information on this type, refer to Partition IV of the CLI Standard. 29.4.2 The Obsolete attribute The attribute Obsolete is used to mark types and members of types that should no longer be used. If a program uses a type or member that is decorated with the Obsolete attribute, then the compiler shall issue a diagnostic in order to alert the developer, so the offending code can be fixed. Specifically, the compiler shall behave as if a corresponding #error directive was encountered if no error parameter (the second parameter) is provided, or if the error parameter is provided and has the value false. The program is ill-formed if the error parameter is specified and has the value true. [Example: In the example
[Obsolete("This class is obsolete; use class B instead", true)] ref struct A { void F() {} };
166
Attributes ref struct B { void F() {} }; int main() { A^ a = gcnew A(); a->F(); } // diagnostic
the class A is decorated with the Obsolete attribute. Each use of A in main results in a diagnostic that includes the specified message, This class is obsolete; use class B instead. end example] For more information on this type, refer to Partition IV of the CLI Standard. 29.4.3 The Conditional attribute The CLI standard defines the attribute Conditional. This attribute allows languages targeting the CLI to provide the ability to enable the definition of conditional methods and conditional attribute classes. C++/CLI does not provide this ability; although attributes of this type are accepted, they have no affect on code generation or execution. 29.4.4 Security attributes Security attributes derive from System::Security::Permissions::SecurityAttribute and shall only be applied to types, functions, and assemblies. All constructors of security attributes shall take System::Security::Permissions::SecurityAction (see 22.11 of the CLI Standard) as the first parameter. Security attributes associate additional semantics with usage of an assembly, type, or function depending on the SecurityAction in the first parameter of the attributes constructor. Semantics of security attributes are provided by the execution engine. A compiler optimization shall preserve these semantics. For instance, if the compiler inlines a function with a security attribute, the compiler shall ensure the equivalent action is invoked by the calling function or at the point that the function is inlined.
29.5 Attributes for interoperation

29.5.1 Interoperation with other CLI-based languages 29.5.1.1 The DefaultMember attribute The attribute System::Reflection::DefaultMemberAttribute is used to provide the underlying name to the default-indexed property. The attribute is placed on the class, and all overloads of a defaultindexed property share the same name. 29.5.1.2 The MethodImplOption attribute This attribute is discussed in 19.6, 19.6.2, and 34.7.4.5. 29.5.2 Interoperation with native code See the discussion of the attribute type DllImport in 18.5.
167
30. Templates
The template syntax is the same for all types, including CLI class types. Templates on CLI class types can be partially specialized, fully specialized, and non-type parameters of any type (subject to all the constantexpression and type rules in the C++ Standard) can be used, with the same semantics as specified by the C++ Standard. Templates are fully resolved and compiled at compile time, and reside in their own assemblies. Within an assembly, templates are implicitly instantiated only for the uses of that template within the assembly. For metadata details, see 34.17.
30.1 Template declarations

In addition to the template declarations allowed by Standard C++, C++/CLI allows ref class templates, value class templates, and interface templates. Delegate templates and enum class templates are ill-formed. To allow constructs such as List<List<int>>, where >> is treated as two tokens instead of one, the C++ Standard (14/1) is augmented by the addition of the following text just after the grammar rules: [Note: The > token following the template-parameter-list of a template-declaration may be the product of replacing a >> token by two consecutive > tokens (14.2). end note] The C++ Standard (14.1/1) is augmented by the addition of the following text just after the grammar rules: [Note: The > token following the template-parameter-list of a type-parameter may be the product of replacing a >> token by two consecutive > tokens (14.2). end note]
30.2 Template specialization

To allow constructs such as List<List<int>>, where >> is treated as two tokens instead of one, the C++ Standard (14.2/3) is augmented by the addition of the following text after the last normative sentence in, but before the example: Similarly, the first non-nested >> is treated as two consecutive but distinct > tokens, the first of which is taken as the end of the template-argument-list and completes the template-id. [Note: The second > token produced by this replacement rule may terminate an enclosing template-id construct or it may be part of a different construct (e.g., a cast). end note] The example of 14.2/3 is replaced by the following:
template<int i> class X { /* ... */ }; X< 1>2 > x1; // Syntax error. X<(1>2)> x2; // Okay. template<class T> class Y { /* ... */ }; Y<X<1>> x3; // Okay, same as "Y<X<1>> x3;". Y<X<6>>1>> x4; // Syntax error. Instead, write "Y<X<(6>>1)>> x4;".
30.3 Attributes
Classes within templates can have attributes, with those attributes being written after the template parameter list and before the class-key. A template parameter is allowed as an attribute, and also as an argument to an attribute. [Example:
template<typename T> [CLSCompliant(false)] ref class R { };
168
Templates
end example] Functions within templates can have attributes, with those attributes being written after the template parameter list and before the function definition. [Example:
template <typename T> [CLSCompliant(false)] void f(const T& t) { }
end example]
30.4 Type deduction

There is no ordering among the punctuators %, ^, &, and *. If a template parameter is deduced to have the null type (12.3.4), the program is ill-formed. 30.4.1 Template argument deduction To accommodate the conversion of <narrow-string-literal-type> and <wide-string-literal-type> to System::String^, the list in the C++ Standard (14.8.2.1/2) is augmented to include the following: If A is <narrow-string-literal-type>, the type "array of n const char" is used in place of A for type deduction. If A is <wide-string-literal-type>, the type "array of n const wchar_t" is used in place of A for type deduction.
169
31. Generics
Generic types and functions are a set of featurescollectively called genericsdefined by the CLI to allow parameterized types. Generics differ from Standard C++s templates in that generics are instantiated by the Virtual Execution System (VES) at runtime rather than by the compiler at compile-time. A generic declaration defines one or more type parameters for a declaration of a ref class, value class, interface class, delegate, or function. To instantiate a generic type or function from a generic declaration, type arguments that correspond to that generic declarations type parameters shall be supplied. The set of type arguments that is permitted for any given type parameter can be restricted via the use of one or more constraints. The arity of a generic type is the number of type parameters declared explicitly for that type. As such, the arity of a nested type does not include the type parameters introduced by the parent type. For metadata details, see 34.18.
31.1 Generic declarations

To accommodate the addition of generics, the grammar for declaration in the C++ Standard (7) is augmented, as follows: declaration: block-declaration function-definition template-declaration generic-declaration explicit-instantiation explicit-specialization linkage-specification namespace-definition A generic declaration is defined as follows: generic-declaration: generic < generic-parameter-list > constraint-clause-listopt declaration generic-parameter-list: generic-parameter generic-parameter-list , generic-parameter Type parameters are defined via a generic-parameter-list, which is a sequence of one or more genericparameters (31.1.1). Constraints are defined via a constraint-clause-list (31.4). If the declaration of a generic-declaration is other than a ref class, value class, interface class, delegate, or function (excluding constructors, destructors, and finalizers), the program is ill-formed. A program is ill-formed if it declares a property or event as a generic. The accessor functions of a property or event shall not be generic. A generic-declaration is a declaration. A generic-declaration is also a definition if its declaration defines a ref class, a value class, an interface class, a delegate, or a function. A generic-declaration shall appear only at a namespace scope or class scope declaration. Except for generic non-member functions, generic declarations that are also definitions can have public or private assembly visibility (10.6.1).
170
Generics
A generic type shall not have the same name as any other generic type, template, class, delegate, function, object, enumeration, enumerator, namespace, or type in the same scope (C++ Standard 3.3), except as specified in 14.5.4 of the C++ Standard. Except that a generic function can be overloaded either by nongeneric functions with the same name or by other generic functions with the same name, a generic name declared in namespace scope or in class scope shall be unique in that scope. Generic type declarations follow the same rules as non-generic type declarations except where noted. Generic type declarations can be nested inside non-generic type declarations. Generic types can be nested in native classes. Generic functions are discussed further in (31.3). C++/CLI permits importing from another assembly multiple generic types declared in the same scope to have the same name, provided each has a different number of generic parameters. [Example:
ref class R { }; generic<typename T> public ref class R { }; generic<typename T, typename U> public ref class R { };
end example] using-declarations shall not be used to make generics from different scopes visible in a given scope, even if the generics differ in arity. Similarly, if generics from different scopes are found by a lookup because of using-directives, the lookup is ambiguous. Generics cannot be explicitly or partially specialized. [Note: As generics do not allow for specialization, there is no need for disambiguating names with the typename and template keywords. end note] A generic function or a generic CLI class can be a friend of a native class. All specializations of a generic shall be made a friend; if any specialization of a generic is excluded from friendship, the program is illformed. [Note: As friendship is only permitted for native classes, and native classes cannot be generics, it is not possible for a generic to grant friendship to another class or function. end note]. 31.1.1 Type parameters A type parameter is defined in one of the following ways: generic-parameter: attributesopt class identifier attributesopt typename identifier There is no semantic difference between class and typename in a generic-parameter. A genericparameter can optionally have one or more attributes (29). A generic-parameter defines its identifier to be a type-name. The scope of a generic-parameter extends from its point of declaration until the end of the declaration to which its generic-parameter-list applies. [Note: Unlike templates, generics has no equivalent to a non-type template-parameter or a template template-parameter. Neither does generics support default generic-parameters; instead, generic type overloading is used. end note] As a type, type parameters are purely a compile-time construct. At run-time, each type parameter is bound to a run-time type that was specified by supplying a type argument to the generic type declaration. Thus, the type of a variable declared with a type parameter will, at run-time, be a closed constructed type (31.2). The run-time execution of all statements and expressions involving type parameters uses the actual type that was supplied as the type argument for that parameter. The literal nullptr cannot be converted to a type given by a generic type parameter, except if the type parameter is known to be a handle type. However, a default constructor expression can be used instead to get 171
a null value for a generic type parameter. In addition, a value with a type given by a generic type parameter can be compared with nullptr using == and != unless the type parameter has the value type constraint (31.4) [Example:
generic<typename T, typename U> where U : ref class ref class R { void F() { T t = T(); // t is initialized to default value U u = nullptr; // u can be initialized with nullptr, // because it has the ref class constraint } }; /* ... */
end example] Any type used as a generic type parameter shall have linkage. 31.1.2 Referencing a generic type by name Like templates in Standard C++, within the body of a generic type G<T> any usage of the name (that is neither qualified nor a generic-id) of that type G (otherwise known as the instance type) is assumed to refer to the current instantiation. [Example:
generic<typename T> ref class R { public: R() {} // ok: means R<T> void f(R^); // ok: means R<T> ::R g(); // error };
end example] Outside its declaration, a generic type is referenced using a constructed type (31.2). [Example: Given the following,
generic<typename T> ref class List {}; generic<typename U> void f() { List<U>^ l1 = gcnew List<U>; List<int>^ l2 = gcnew List<int>; List<List<String^>^>^ l3 = gcnew List<List<String^>^>; }
some examples of constructed types are List<U>, List<int>, and List<List<String^>^>. A constructed type that uses one or more type parameters, such as List<U>, is an open constructed type (31.2.1). A constructed type that uses no type parameters, such as List<int>, is called a closed constructed type (31.2.1). end example] 31.1.3 The instance type Each type declaration has an associated constructed type, the instance type. For a generic type declaration, the instance type is formed by creating a constructed type (31.2) from the type declaration, with each of the supplied type arguments being the corresponding type parameter. Since the instance type uses the type parameters, it can only be used where the type parameters are in scope; that is, inside the type declaration. Inside the declaration of a ref class, this is a handle to the instance type. Inside the declaration of a value class, this is an interior pointer to the instance type. For non-generic types, the instance type is simply the declared type. [Example: The following shows several class definitions along with their instance types:
172
Generics generic<typename T> ref class A { // instance type: A<T> ref class B {}; // instance type: A<T>::B generic<typename U> ref class C {}; // instance type: A<T>::C<U> }; class D {}; // instance type: D
end example] 31.1.4 Base classes and interfaces The base class and interfaces of a generic type declaration shall not be a type parameter, though they can be a constructed type using a type parameter. [Example:
ref class B1 {}; generic<typename T> ref class B2 {}; generic<typename T> interface class I1 {}; generic<typename T> ref class R1 : T {}; generic<typename T> ref class R2 : B1 {}; // error // ok
generic<typename T> ref class R3 : B2<int>, I1<int> {}; // ok (closed constructed types) generic<typename T> ref class R4 : B2<T>, I1<T> {}; // ok (open constructed types)
end example] A generic class definition shall not use System::Attribute as a direct or indirect base class. A generic class definition shall not have an indirect base class that is a template parameter. 31.1.5 Class members All members of a generic type can use type parameters from any enclosing type, either directly or as part of a constructed type. When a particular closed constructed type (31.1.2) is used at run-time, each use of a type parameter is replaced with the actual type argument supplied to the constructed type. Properties, events, constructors, destructors, and finalizers shall not themselves have explicit type parameters (although they can occur in generic classes, and use the type parameters from an enclosing class). When the type of a member is a type parameter, the declaration of that member shall use that type parameters name without any pointer, reference, or handle declarators. Member access on a member whose type is a type parameter shall use the -> operator. [Example:
interface class I1 { void F(); }; generic<typename T> where T : I1 ref class A { T t; // no *, &, %, or ^ declarator allowed public: void F() {} void G() { t->F(); // -> must be used, not . } };
173
end example] [Note: The compiler only generates one definition for a generic class in metadata. Generics allow value classes as generic type parameters. Textual substitution of a value class parameter would lead to an illformed program as the -> operator is not allowed for member access. As the VES is responsible for instantiations of generics, textual substitution is the wrong way of thinking about generic instantiation. end note] As a member whose type is a parameter type will be a value class, or a handle to a ref class, interface class, delegate, or CLI array, the destructor of a generic class will not invoke the destructor on such a member. Within a generic class definition, access to inherited protected instance members is permitted through an instance of any open constructed class type constructed from that generic class. [Example: In the following code
generic<typename T> ref class B { protected: T x; }; generic<typename T> ref class D : B<T> { static void F() { D<T>^ dt = gcnew D<T>; dt->x = T();
// Ok
D<int>^ di = gcnew D<int>; di->x = 123; // error D<String^>^ ds = gcnew D<String^>; ds->x = "test"; // error
} };
the first assignment to x is permitted because it takes place through an instance of an open constructed class types constructed from the generic type. However, the second and third assignments are prohibited because they take place through an instance of a closed constructed class type. When accessing members of a closed constructed generic, even within the generic definition, the access rules shall treat that class as an unrelated entity. end example] Static operators are discussed in (31.1.7), other static members are discussed in (31.1.6), nested types are discussed in (31.1.10), and generic functions, in general, are discussed in (31.3). 31.1.6 Static members A static data member in a generic class definition is shared amongst all instances of the same closed constructed type (31.1.2), but is not shared amongst instances of different closed constructed types. These rules apply regardless of whether the type of the static data member involves any type parameters or not. A static constructor in a generic class is used to initialize static data members and to perform other initialization for each different closed constructed type that is created from that generic class definition. The type parameters of the generic type declaration are in scope, and can be used, within the body of the static constructor. A new closed constructed class type is initialized the first time that either: An instance of the closed constructed type is created. Any of the static members of the closed constructed type are referenced.
To initialize a new closed constructed class type, first a new set of static data members for that particular closed constructed type is created. Each of the static data members is initialized to its default value. Next, the static data members initializers are executed for those static fields. Finally, the static constructor is executed. [Example:
174
Generics generic<typename T> ref class C { static int count = 0; public: static C() { Console::WriteLine(<C<T>>::typeid); } C() { count++; } static property int Count { int get() { return count; } } }; int main() { C<int>^ x1 = gcnew C<int>; Console::WriteLine(C<int>::Count); C<double>^ x2 = gcnew C<double>; Console::WriteLine(C<double>::Count); Console::WriteLine(C<int>::Count); C<int>^ x3 = gcnew C<int>; Console::WriteLine(C<double>::Count); } Console::WriteLine(C<int>::Count);

C`1[System.Int32] 1 C`1[System.Double] 1 1 1 2
end example] Static operators are discussed in 31.1.7. 31.1.7 Operators Generic class definitions can define operators and conversion functions, following the same rules as nongeneric class definitions. The instance type (31.1.3) of the class definition shall be used in the declaration of operators in accordance with the rules for operators in 19.7 or conversion functions in 14.5.3. The parameter that is not constrained by these rules can be a generic type parameter. [Example: The following shows some examples of valid operator declarations in a generic class:
generic <typename T> public ref struct R { static R^ operator ++(R^ operand) { } static int operator *(R^ op1, T op2) { } static explicit operator R^(T value) { } };
end example] 31.1.8 Member overloading Functions, instance constructors, and static operators within a generic class definition can be overloaded; however, this can lead to an ambiguity for some closed constructed types. [Example:
175
C++/CLI Language Specification generic<typename T1, typename T2> ref class X { public: void F(T1, T2) { } void F(T2, T1) { } void F(int, String^) { } }; int main() { X<int, double>^ x1 = gcnew X<int, double>; x1->F(10, 20.5); X<double, int>^ x2 = gcnew X<double, int>; x2->F(20.5, 10); X<int, int>^ x3 = gcnew X<int, int>; x3->F(10, 20);
// okay // okay // error, ambiguous
X<int, String^>^ x4 = gcnew X<int, String^>; x4->F(10, "abc"); // error, ambiguous }
end example] A generic class is allowed to have this potential ambiguity; however, a program is ill-formed if it uses a constructed type to create such an ambiguity. 31.1.9 Member overriding Function members in generic classes can override function members in base classes, as usual. If the base class is a non-generic type or a closed constructed type, then any overriding function member cannot have constituent types that involve type parameters. However, if the base class is an open constructed type, then an overriding function member can use type parameters in its declaration. When determining the overridden base member, the members of the base classes shall be determined by substituting type arguments, as described in 31.2.4. Once the members of the base classes are determined, the rules for overriding are the same as for non-generic classes. [Example:
generic<typename T> ref class C abstract { public: virtual T F() { } virtual C<T>^ G() { } virtual void H(C<T>^ x) { } }; ref class D : C<String^> { public: virtual String^ F() override { } // Ok virtual C<String^>^ G() override { } // Ok virtual void H(C<int>^ x) override { } // Error, should be C<String^> }; generic<typename T, typename U> ref class E : C<U> { public: virtual U F() override { } // Ok virtual C<U>^ G() override { } // Ok virtual void H(C<T>^ x) override { } // Error, should be C<U> };
end example] 31.1.10 Nested types A generic class definition can contain nested type declarations, except that a generic class definition shall not contain a native class. The type parameters of the enclosing class can be used within the nested types. A nested type declaration can contain additional type parameters that apply only to the nested type. A generic type can be nested within a non-generic type. 176
Generics
Every type declaration contained within a generic class definition is implicitly a generic type declaration. When writing a reference to a type nested within a generic type, the containing constructed type, including its type arguments, shall be named. However, from within the outer class, the nested type can be used without qualification; the instance type of the outer class can be implicitly used when constructing the nested type. [Example: The following example shows three different correct ways to refer to a constructed type created from Inner; the first two are equivalent:
generic<typename T> ref struct Outer { generic<typename U> ref class Inner { public: static void F(T t, U u) { } }; static void F(T t) { Outer<T>::Inner<String^>::F(t, "abc"); have Inner<String^>::F(t, "abc"); Outer<int>::Inner<String^>::F(3, "abc"); } }; // These two statements // the same effect // This type is different
end example] A type parameter in a nested type can hide a member or type parameter declared in the outer type. [Example:
generic<typename T> ref class Outer { generic<typename T> ref class Inner { T t; }; }; // Valid, hides Outers T // Refers to Inners T
end example] A program having a generic type nested within a class template is ill-formed.
31.2 Constructed types

A generic type declaration is used as a blueprint to form many different types, by way of applying type arguments (31.2.1). A type that is named with at least one type argument is called a constructed type. A constructed type can be open or closed, as we shall see in 31.2.1. To accommodate the addition of generics, the grammar for unqualified-id in the C++ Standard (5.1) is augmented, as follows by adding generic-id: unqualified-id: identifier operator-function-id conversion-function-id ~ class-name ! class-name template-id generic-id
default
A constructed type is referred to by a generic-id: generic-id: generic-name < generic-argument-list > generic-name: identifier operator-function-id
177
generic-argument-list is discussed in (31.2.2). 31.2.1 Open and closed constructed types All types can be classified as either open constructed types or closed constructed types. An open constructed type is a type that involves type parameters. More specifically: A type parameter defines an open constructed type. A CLI array type is an open constructed type if and only if its element type is an open constructed type. A constructed type is an open constructed type if and only if one or more of its type arguments is an open constructed type. A constructed nested type is an open constructed type if and only if one or more of its type arguments (31.2.2) or the type arguments of its containing type(s) is an open constructed type.
A closed constructed type is a type that is not an open constructed type. [Example: Given the following,
generic<typename T> ref class List {}; generic<typename U> void f() { List<U>^ l1 = gcnew List<U>; List<int>^ l2 = gcnew List<int>; List<List<String^>^>^ l3 = gcnew List<List<String^>^>; } List<U>, List<int>, and List<List<String^>^> are examples of constructed types, where List<U> is an open constructed type, and List<int> and List<List<String^>^> are closed constructed types. end example]
At run-time, all of the code within a generic type declaration is executed in the context of a closed constructed type that was created by applying type arguments to the generic declaration. Each type parameter within the generic type is bound to a particular run-time type. The run-time processing of all statements and expressions always occurs with closed constructed types, and open constructed types occur only during compile-time processing. Each closed constructed type has its own set of static variables, which are not shared with any other closed constructed types. Since an open constructed type does not exist at run-time, there are no static variables associated with an open constructed type. Two closed constructed types are the same type if they are constructed from the same type declaration, and their corresponding type arguments are the same type. A constructed type has the same accessibility as its least accessible type argument. 31.2.2 Type arguments A generic type or function is instantiated from a generic declaration by specifying type arguments that correspond to that generic declarations type parameters. Type arguments are specified via a genericargument-list: generic-argument-list: generic-argument generic-argument-list , generic-argument generic-argument: type-id The arguments for an instantiation of a generic class shall always be explicitly specified. The arguments for an instantiation of a generic function (31.3) can either be specified explicitly, or they can be determined by type deduction.
178
Generics
A generic-argument shall be a constructed type that is a value class, a handle to a ref class, a handle to a delegate, a handle to an interface, a handle to a CLI array, or it shall be a type parameter from an enclosing generic. [Note: It is not possible to use a native class, a pointer, a reference, a handle to a value class, a boxed value type, or a ref class by value as a generic argument. end note] Each generic-argument shall satisfy any constraints (31.4) on the corresponding type parameter. 31.2.3 Base classes and interfaces A constructed class type has a direct base class. If the generic class definition does not specify a base class, the base class is System::Object. If a base class is specified in the generic class definition, the base class of the constructed type is obtained by substituting, for each generic-parameter in the base class definition, the corresponding generic-argument of the constructed type. [Example: Given the generic class definitions
generic<typename T, typename U> ref class B { }; generic<typename T> ref class D : B<String^, array<T>> { };
the base class of the constructed type D<int> would be B<String^, array<int>>. end example] Similarly, constructed ref class, value class, and interface types have a set of explicit base interfaces. The explicit base interfaces are formed by taking the explicit base interface definitions on the generic type declaration, and substituting, for each generic-parameter in the base interface definition, the corresponding generic-argument of the constructed type. The set of all base classes and base interfaces for a type is formed, as usual, by recursively getting the base classes and interfaces of the immediate base classes and interfaces. [Example: For example, given the generic class definitions:
ref class A { }; generic<typename T> ref class B : A { }; generic<typename T> ref class C : B<IComparable<T>^> { }; generic<typename T> ref class D : C<array<T>> { };
the base classes of D<int> are C<array<int>>, B<IComparable<array<int>^>>, A, and System::Object. end example] 31.2.4 Class members The non-inherited members of a constructed type are obtained by substituting, for each generic-parameter in the member declaration, the corresponding generic-argument of the constructed type. The substitution process is based on the semantic meaning of type declarations, and is not simply textual substitution (31.1.5). [Example: Given the generic class definition
generic<typename T, typename U> ref class X { array<T>^ a; void G(int i, T t, X<U,T> gt); property U P { U get(); void set(U value); } int H(double d); };
the constructed type X<int, bool> has the following members:

array<int>^ a; void G(int i, int t, X<int,bool>^ gt); property bool P { bool get(); void set(bool value); } int H(double d);
179
end example] The inherited members of a constructed type are obtained in a similar way. First, all the members of the immediate base class are determined. If the base class is itself a constructed type, this might involve a recursive application of the current rule. Then, each of the inherited members is transformed by substituting, for each generic-parameter in the member declaration, the corresponding generic-argument of the constructed type. [Example:
generic<typename U> ref class B { public: U F(long index); }; generic<typename T> ref class D : B<array<T>^> { public: T G(String^ s); };
In the above example, the constructed type D<int> has a non-inherited member int G(String^ s) obtained by substituting the type argument int for the type parameter T. D<int> also has an inherited member from the class definition B. This inherited member is determined by first determining the members of the constructed type B<array<T>^> by substituting array<T>^ for U, yielding array<T>^ F(long index). Then, the type argument int is substituted for the type parameter T, yielding the inherited member array<int>^ F(long index). end example] 31.2.5 Accessibility A constructed type C<T1, ...,TN> is accessible when all its parts C, T1, ..., TN are accessible. For instance, if the generic type name C is public and all of the generic-arguments T1, ...,TN are accessible as public, then the constructed type is accessible as public, but if either the type name C or any of the genericarguments has accessibility private then the accessibility of the constructed type is private. If one generic-argument has accessibility protected, and another has accessibility private protected, then the constructed type is accessible only in this class and its subclasses in this assembly. The accessibility domain for a constructed type is the most restrictive access of the open type and its type arguments. Accessibility rules for instantiations of generics are the same as for templates.
31.3 Generic functions

Member functions and non-member functions can be declared generic (31.1). When a generic function is declared inside a ref class, value class, or interface definition, the enclosing type can itself be either generic or non-generic. If a generic function is declared inside a generic type declaration, the body of the function can refer to both the type parameters of the function, and the type parameters of the containing declaration. Not all generic type parameters to a generic function need appear as a parameter type or return type of that function. [Example:
generic<typename T> void f1(T); ref class C1 { generic<typename T, typename U> T f2(T t) { U u; } generic<typename T> T f2(T); };
180
Generics generic<typename T1> ref class C2 { generic<typename T2> void f3(T1, array<T2>^); };
end example] Types not used as a parameter type to a generic function cannot be deduced. Types that cannot be deduced for function templates cannot be deduced for generic functions. When used with a generic function, static, extern, and inline have the same meaning as when used with a non-generic function in the same context. When the type of a parameter or variable is a type parameter, the declaration of that parameter or variable shall use that type parameters name without any pointer, native reference, or handle declarators. [Note: A parameter or variable type is permitted to be a tracking reference to a type parameter. end note] Member access on a parameter or variable whose type is a type parameter shall use the -> operator. [Example:
interface class I1 { void F(); }; generic<typename T> where T : I1 void H(T t1) { // no *, &, or ^ declarator allowed T t2 = t1; // t1->F(); // -> must be used, not . t2->F(); // }
end example] Type parameters can be used in the type of a parameter array. A generic function can be bound to a suitably typed delegate. 31.3.1 Function signature matching rules For the purposes of signature comparisons in function overloading, any constraint-clause-lists are ignored, as are the names of the functions generic-parameters; however, the number of generic type parameters is relevant. [Example:
ref class A {}; ref class B {}; interface class IX { generic<typename T> where T : A void F1(T t); generic<typename T> where T : B void F1(T t); generic<typename T> T F2(T t, int i); generic<typename U> void F2(U u, int i);
// error, constraints are ignored
// error, parameter names and return // type are ignored
void F3(int x); // no type parameters generic<typename T> void F3(int x); // okay, different type parameter count generic<typename T, typename U> void F3(int x); // okay, different type parameter count generic<typename U, typename T> void F3(int x); // error, type parameter names are ignored };
181
end example] Functions can be overloaded; however, this can lead to an ambiguity for certain calls. [Example:
generic<typename T1, typename T2> void F(T1, T2) { } generic<typename T1, typename T2> void F(T2, T1) { } int main() { F<int, double>(10, 20.5); F<double, int>(20.5, 10); F<int, int>(10, 20); } // okay // okay // error, ambiguous
end example] Although a program is permitted to have generic function declarations that could lead to such ambiguities, that program is ill-formed if it uses function calls to create such an ambiguity. Generic functions can be declared abstract, virtual, and override. The signature matching rules described above are used when matching functions for overriding or interface implementation. When a generic function overrides a generic function declared in a base class, or implements a function in a base interface, the constraints given for each function type parameter shall be the same in both declarations. [Example:
ref struct B abstract { generic<typename T, typename U> virtual T F(T t, U u) abstract; generic<typename T> where T : IComparable virtual T G(T t) abstract; }; ref struct D : B { generic<typename X, typename Y> virtual X F(X x, Y y) override; generic<typename T> virtual T G(T t) override; };
// Okay // error, constraint mismatch
The override of F is valid because type parameter names are permitted to differ. The override of G is invalid because the given type parameter constraints (in this case none) do not match those of the function being overridden. end example] 31.3.2 Type deduction A call to a generic function can explicitly specify a type argument list via a generic-id, or it can omit that type argument list using a generic-name only and rely on type deduction to determine the type arguments. [Example:
ref struct X { generic<typename T> static void F(T t) { Console::WriteLine("one"); } generic<typename T> static void F(T t1, T t2) { Console::WriteLine("two"); } generic<typename T> static void F(T t1, int t2) { Console::WriteLine("three"); }
};
182
Generics int main() { X::F<int>(1); X::F(1);
// explicit, prints "one" // deduced, prints "one"
X::F<double>(5.0, 6.0); // explicit, prints "two" X::F(5.0, 6.0); // deduced, prints "two" X::F<double>(5.0, 3); X::F(5.0, 3); X::F<int>(1, 2); X::F(1, 2); X::F<double>(1, 2); } // explicit, prints "three" // deduced, prints "three" // error, ambiguous // error, ambiguous // explicit, prints "three"

interface class IX {}; ref class R : IX {}; generic<typename T> void f(T) {} void g(R^ hR) { f<IX^>(hR); // T is specified to be IX f(hR); // T is deduced to be R }
end example] Type deduction allows a more convenient syntax to be used for calling a generic function, and allows the programmer to avoid specifying redundant type information. In a generic function, if the type of the corresponding argument of the call is either <narrow-string-literaltype> or <wide-string-literal-type>, the deduced type, P, is System::String^. [Note: Type deduction on a string literal for a function template results in an array of characters instead of System::String^. end note] Otherwise, type deduction within generics is handled like type deduction within templates (C++ Standard 14.8.2). If the generic function was declared with a parameter array, then type deduction is first performed against the function using its exact signature. If type deduction succeeds, and the resultant function is applicable, then the function is eligible for overload resolution in its normal form. Otherwise, type deduction is performed against the function in its expanded form. An instance of a delegate can be created that refers to a generic function declaration. The type arguments used when invoking a generic function through a delegate are determined when the delegate is instantiated. The type arguments for a generic delegate can be deduced when invoking the delegate in the same manner as type deduction for invoking a generic function. If type deduction is used, the parameter types of the delegate are used as argument types in the deduction process. The return type of the delegate is not used for deduction. [Example: The following example shows both ways of supplying a type argument to a delegate instantiation expression:
delegate int D(String^ s, int i); delegate int E(); ref class X { public: generic<typename T> static T F(String^ s, T t); generic<typename T> static T G(); { gcnew gcnew gcnew gcnew
};
int main() D^ d1 = D^ d2 = E^ e1 = E^ e2 = }
D(X::F<int>);// D(X::F); // E(X::G<int>);// E(X::G); //
okay, type argument given okay, int deduced as type okay, type argument given error, cannot deduce from
explicitly argument explicitly return type
183
end example] A non-generic delegate type can be instantiated using a generic function. It is also possible to create an instance of a constructed delegate type using a generic function. In all cases, type arguments are given or deduced when the delegate instance is created, and a type-argument-list shall not be supplied when that delegate is invoked.
31.4 Constraints
The set of type arguments that is permitted for any given type parameter in a generic type or function declaration can be restricted via the use of one or more constraints. Such constraints are specified via a constraint-clause-list: constraint-clause-list: constraint-clause-listopt constraint-clause: where identifier constaint-item-list: constraint-item constraint-item-list constraint-item: type-id
refclass refstruct valueclass valuestruct gcnew ( ) :
constraint-clause constraint-item-list
constraint-item
Each constraint-clause consists of the token where, followed by an identifier that shall be the name of a type parameter in the generic type declaration to which this constraint-clause applies, followed by a colon and the list of constraints for that type parameter. There shall be no more than one constraint-clause for each type parameter in any generic declaration, and the constraint-clauses can be listed in any order. The token where is not a keyword. Generic constraints for generic functions are checked after overload resolution. Constraints do not influence overload resolution. [Note: Because value class and value struct are turned into a single token early in the phases of translation, the following code unambiguously has the value class constraint on T:
generic<typename T> where T : value class V F(T t) {}
It is not possible to create a constraint on a type named value followed by a function that uses an elaborated-type-specifier for a native class as a return type. end note] If the type specified by type-id is a ref class type, it is a class constraint. A class constraint shall not be sealed. A constraint-item-list shall contain no more than one class constraint. If the type specified by type-id is an interface class type, it is an interface constraint. The same interface type shall not be specified more than once in a given constraint-clause. If the type specified by type-id is a generic type parameter, it is a naked type parameter constraint. The same naked type parameter shall not be specified more than once in a given constraint-clause. A program is ill-formed if a type parameter results in a constraint upon itself, either directly or indirectly. None of the constraints specified by a naked type parameter shall conflict with other constraints given in a constraintclause. For example, a constraint list shall not have a class constraint and a naked type parameter constraint that itself has a class constraint.
184
Generics
A class or interface constraint can involve any of the type parameters of the associated type or function declaration as part of a constructed type, and can involve the type being declared. Any class or interface type specified as a type parameter constraint shall be at least as accessible as the generic type or function being declared. If the type specified by type-id is anything else, the program is ill-formed. [Example: The following are examples of constraints:
generic<typename T> interface class IComparable { int CompareTo(T value); }; generic<typename T> interface class IKeyProvider { T GetKey(); }; generic<typename T> where T : IPrintable ref class Printer { }; generic<typename T> where T : IComparable<T> ref class SortedList { }; generic<typename K, typename V> where K : IComparable<K> where V : IPrintable, IKeyProvider<K> ref class Dictionary { };
end example] If a type parameter has no constraints associated with it then it is implicitly constrained by System::Object. [Note: having a type parameter constrained in this manner severely limits what you can do with the type within the body of the generic. end note] Generic constraint-items shall not have an elaborated-type-specifier. Constraints on generic type parameters do not have influence on the ordering or on overload resolution. The rules for partial ordering of function templates apply to generic functions. A program that attempts to explicitly specialize a generic function using function template, is ill-formed. 31.4.1 Satisfying constraints Whenever a constructed type or generic function is referenced, the supplied type arguments are checked against the type parameter constraints declared on the generic type or function. For each where clause, the type argument A that corresponds to the named type parameter is checked against each constraint as follows: If the constraint is a class type, an interface type, or a type parameter, let C represent that constraint with the supplied type arguments substituted for any type parameters that appear in the constraint. To satisfy the constraint, it shall be the case that an object of type A is convertible to an object of type C by one of the following: o o o An identity conversion A handle conversion A boxing conversion
[Example:
185
C++/CLI Language Specification interface class I {}; ref class C : I {}; value class V : I {}; generic<typename T> where T : I ref class R {}; R<IF^> r1; // satisfies constraint with identity conversion R<C^> r2; // satisfies constraint with handle conversion R<V> r3; // satisfies constraint with boxing conversion generic<typename U> where U : T ref class Q { R<U> r4; // satisfies constraint, the synthesized type for // U has valid conversions to T's constraint };
end example] If the constraint is the ref class constraint, the type A shall satisfy one of the following: o o
A is a handle type. A is a type parameter that satisfies the ref class constraint (either directly or transitively
because it is constrained by another type parameter that satisfies the ref class constraint). If the constraint is the value class constraint, the type A shall satisfy one of the following: o
A is a value type other than a pointer and is not the generic System::Nullable type. [Note: Note that System::ValueType and System::Enum are reference types so they do
not satisfy this constraint. end note] o

A is a type parameter having the value type constraint (either directly or transitively because
it is constrained by another type parameter that has the value type constraint). If the constraint is the constructor constraint gcnew(), the type argument A shall not be abstract and shall have a public default constructor. This is satisfied if one of the following is true: o o o o o
A is a value type, since all value types have a public default constructor. A is a type parameter having the value type constraint. A is a class that is not abstract, A contains an explicitly declared public default constructor. A is not abstract and has a default constructor. A is a type parameter having the constructor constraint (either directly or transitively
because it is constrained by another type parameter that satisfies the constructor constraint). A program is ill-formed if it contains a generic type one or more of whose type parameters constraints are not satisfied by the given type arguments. Since type parameters are not inherited, constraints are never inherited either. [Example: In the code below,
D shall specify a constraint on its type parameter T, so that T satisfies the constraint imposed by the base class B<T>. In contrast, class E need not specify a constraint, because List<T> implements IEnumerable for any T. generic<typename T> where T: IEnumerable ref class B { }; generic<typename T> where T: IEnumerable ref class D : B<T> { }; generic<typename T> ref class E : B<List<T>^> { };
186
Generics
end example] 31.4.2 Member lookup on type parameters Templates wait to perform lookup with a type parameter until the type parameter is replaced by a type argument. Generics perform lookup at the point of defining the generic rather than the point of specialization. The results of lookup involving a type given by a type parameter T depends on the constraints, if any, specified for T. Lookup replaces the type of the generic type parameter T with a type as specified by one of the following cases: 1. If T has a naked type parameter constraint N, then a type is synthesized for N according to constraints and the rules one two through six below. If the synthesized type for N would satisfy all other constraints of T, then the type synthesized for N replaces T. Otherwise, all the constraints of N are added to the constraints of T and type is synthesized according to rules two through six below. 2. If T has no constraints or only the constructor constraint, System::Object replaces T. If lookup selects the constructor, the type is created by calling System::Activator::CreateInstance. 3. If T has the value class constraint, then a value class V is synthesized with the following characteristics. V replaces T for the purpose of lookup. If T has any interface constraints, V provides an implementation for each interface. If lookup and overload resolution selects one of these functions, the constraint is met by the interface function implemented by the synthesized function.
4. If T has the ref class constraint, then a ref class R is synthesized with the following characteristics. R replaces T for the purpose of lookup. If T has any interface constraints, R provides an implementation for each interface. If lookup and overload resolution selects one of these functions, the constraint is met by the interface function implemented by the synthesized function. If T has the constructor constraint, R provides a public constructor with no parameters. If lookup selects this synthesized constructor, the type is created by calling System::Activator::CreateInstance.
5. If T has a base class constraint B, and if B would satisfy all other constraints of T, then B replaces T. Otherwise, a ref class R immediately deriving from B is synthesized with the following characteristics. R replaces T for the purpose of lookup. If T has any interface constraints, R provides an implementation for each interface function that would not already be satisfied by deriving from B. If lookup and overload resolution selects one of the synthesized functions, the constraint is met by the interface function implemented by the synthesized function. [Note: if a base class constraint and an interface constraint has the same function signature, such that the base class function could implement the interface function, the call to that function through the generic type parameter is made through the base class constraint. end note] If T has the constructor constraint, R provides a public constructor with no parameters. If lookup selects this synthesized constructor, the type is created by calling System::Activator::CreateInstance.
6. If T has neither a ref class constraint, a value class constraint, nor a base class constraint, a class type RV that is both a ref class and a value class is synthesized with the following characteristics. (Such a hybrid class can be synthesized by doing lookup twice using both a ref class and value class and ensuring that the result matches.) If T has any interface constraints, RV provides an implementation for each interface. If lookup and overload resolution selects one of these functions, the constraint is met by the interface function implemented by the synthesized function.
187
If T has the constructor constraint, the ref class represented by RV provides a public constructor with no parameters. If lookup selects this synthesized constructor, the type is created by calling System::Activator::CreateInstance.
[Example: Consider the following code:

interface class IMethod { void F(); }; ref struct R : IMethod { virtual void G() = IMethod::F { Console::WriteLine("R::G"); } void F() { Console::WriteLine("R::F"); } }; generic<typename X> where X : IMethod void G1(X x) { x->F(); } generic<typename X> where X : R, IMethod void G2(X x) { x->F(); } template<typename X> void T(X x) { x->F(); } int main() { R^ r = gcnew R; G1(r); G2(r); T(r); }
The program prints the following output.

R::G R::F R::F G1s type parameter only has one interface constraint, so a synthesized type is created with the function F that implements the constraint. Thus the call to F in the body of G1 is through the interface. G2s type parameter has both a base class constraint and an interface constraint. The base class already implements the interface, and thus X is replaced with the R within the body of G2 for the purpose of lookup. end example]
31.4.3 Type parameters and boxing When a value class type overrides a virtual method inherited from System::Object (such as Equals, GetHashCode, or ToString), invocation of the virtual function through an instance of the value class type doesnt cause boxing to occur. This is true even when the value class is used as a type parameter and the invocation occurs through an instance of the type parameter type. Boxing never implicitly occurs when accessing a member on a constrained type parameter. For example, suppose an interface ICounter contains a function Increment which can be used to modify a value. If ICounter is used as a constraint, the implementation of the Increment function is called with a reference to the variable that Increment was called on, never a boxed copy.
188
Generics
31.4.4 Conversions involving type parameters The conversions that are allowed on a type parameter T depend on the constraints specified for T. For a generic type or function that have both class and interface constraints, type conversions defined in a class constraint are always preferred over those in an interface constraint.
189
32. Standard C and C++ libraries

Except for those requirements described elsewhere in this Standard, the interaction between the CLI library and the Standard C and C++ libraries is unspecified.
190
CLI libraries
33. CLI libraries

33.1 Custom modifiers
Implementations of Standard C++ distinguish between different signatures by using name mangling; however, not only is this a language-specific solution, the mangling scheme used varies from one implementation to the next. As such, this approach is not viable in C++/CLI, where interoperability between different C++ implementations is required, and interoperability between different languages is desired. Custom modifiers address this issue. Custom modifiers (CLI Standard, Partition II, Types and signatures), defined in ILAsm using modreq (required modifier) and modopt (optional modifier), are similar to custom attributes except that custom attributes are attached to a declaration, while custom modifiers are part of that declarations signature. Each custom modifer associates a type reference with an item in the signature. Two signatures that differ only by the addition of a custom modifier (required or optional) shall not be considered to match. Signature matching is discussed further in 33.1.1. Custom modifiers have no other effect on the operation of the VES. 33.1.1 Signature matching Consider the following class definition:
public ref class X { public: static void F(int* p1) { } static void F(const int* p2) { } private: static int* p3; static const int* p4; };
The signatures of these four members are recorded in metadata as follows:

.method public static void F(int32* p1) { } .method public static void F(int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* p2) { } .field private static int32* p3 .field private static int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* p4
[Note: Within the CLI context, the fully qualified name of a type uses dot (.) separators, while within a C++ context, a double colon (::) is used instead. end note] Clearly, the two signatures for F differ, allowing these declarations as overloads. Calls to these functions, and the corresponding code they generate, are as follows:
int* q1 = nullptr; X::F(q1); call void X::F(int32*) const int* q2 = nullptr; X::F(q2); call void X::F(int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)*)
The correct function is called by using an exactly matching signature in the call instruction. (If no matching signature is found at runtime, an exception of type System::MissingMethodException is thrown.) Accesses to the data members are matched in a similar fashion:
191
C++/CLI Language Specification static void F(int* p1) { p3 = p1; p4 = p1; } .method public static void F(int32* p1) { ldarg.0 stsfld int32* X::p3 ldarg.0 stsfld int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* X::p4 } static void F(const int* p2) { p4 = p2; } .method public static void F(int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* p2) { ldarg.0 stsfld int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* X::p4 }
The fields are accessed using an exactly matching signature in the stsfld instruction. (If no matching signature is found at runtime, an exception of type System::MissingFieldException is thrown.) 33.1.2 modreq vs. modopt The distinction between required and optional modifiers is important to tools (such as compilers) that deal with metadata. A required modifier indicates that there is a special semantic to the modified item, which shall not be ignored, while an optional modifier can simply be ignored. For example, volatile-qualified data members shall be marked with the IsVolatile modreq. The presence of this modifier cannot be ignored, as all accesses of such members shall involve the use of the volatile. prefixed instruction (see 33.1.5.9 for an example). On the other hand, the const qualifier can be modelled with a modopt since a const-qualified data member or a parameter that is a pointer to a const-qualified object, requires no special treatment. The CLI itself treats required and optional modifiers in the same manner. 33.1.3 Modifier syntax The following grammar is a subset of that defined by the CLI Standard for fields and methods. For expository purposes, this extract has been significantly simplified. (For the complete, non-simplified, version, refer to Partition II of the CLI Standard.) Field:
.field Type Id
Method:
.method Type MethodName ( Parameters ) { MethodBody }
Parameters: [ Param [ , Param ]* ] Param: Type [ Id ]
192
CLI libraries
Type:
int32 Type * Type [ ] Type modreq ( [ AssemblyName ] NamespaceName . Id ) Type modopt ( [ AssemblyName ] NamespaceName . Id )
The Id in Field refers to the name of the data member. The Id in Param refers to the name of the optional function parameter; this name is not part of that functions signature. The Id in Type for a modopt and modreq refers to the name of the custom modifier type. This type shall be a non-nested ref class having public visibility. [Note: Typically, a modifier class is sealed and has no public members. end note] [Example: Here are some data and function member definitions, and the metadata produced for each of their declarations:
public ref class X { int f1; const int f2; const int* f3; const int** f4; const int* const* f5; array<int>^ f6; array<int*>^ f7; const array<int>^ f8; array<const int>^ f9; const int* F() { } void F(int x, const int* y, array<int>^ z) { }
};
.field private int32 f1 .field private int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst) f2 .field private int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* f3 .field private int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)** f4 .field private int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* f5 .field private int32[] f6 .field private int32*[] f7 .field private int32[] modopt([mscorlib]System.Runtime.CompilerServices.IsConst) f8 .field private int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)[] f9 .method private instance int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* F() { } .method private instance void F(int32 x, int32 modopt([mscorlib]System.Runtime.CompilerServices.IsConst)* y, int32[] z) { }
end example] 33.1.4 Types having multiple custom modifiers A Type can contain multiple modreqs and/or modopts. [Example:
public ref class X { const volatile int m; };
193
C++/CLI Language Specification .field private int32 modreq([mscorlib]System.Runtime.CompilerServices.IsVolatile) modopt([mscorlib]System.Runtime.CompilerServices.IsConst) m
end example] 33.1.5 Standard custom modifiers With the exception of IsVolatile (which is defined by the CLI Standard), all of the modifiers documented in this subclause are C++/CLI-specific. These modifier types are sealed, they are derived from System::Object, their public key is [00 00 00 00 00 00 00 00 04 00 00 00 00 00 00 00], they have the attribute CLSCompliantAttribute(true), they belong to the library RuntimeInfrastructure, they reside in the namespace System::Runtime::CompilerServices, and they are part of the assembly mscorlib. 33.1.5.1 IsBoxed [Note: This modreq type is not required by this Standard; however, at least one implementation provides it to support the handle type punctuator ^ when used with value types. Description: This type is used in the signature of any data member to indicate that member is a handle to a value type. It is also used in a function signature to indicate a return type and parameters that are handles to value types. When emitted, this type shall be immediately preceded by class [mscorlib]System.ValueType and modopt(v), in that order, where v is the value type name.
public value class V {}; public ref class C {}; public ref class X { int* m1; int^ m2; V^ m3; C^ m4; public: void F(int* x) { } void F(int^ x) { } const signed char^ F(V^ v, C^ c) { } }; .field private int32* m1 .field private class [mscorlib]System.ValueType modopt([mscorlib]System.Int32) modreq([mscorlib]System.Runtime.CompilerServices.IsBoxed) m2 .field private class [mscorlib]System.ValueType modopt(V) modreq([mscorlib]System.Runtime.CompilerServices.IsBoxed) m3 .field private class C m4 .method public instance void F(int32* x) { } .method public instance void F(class [mscorlib]System.ValueType modopt([mscorlib]System.Int32) modreq([mscorlib]System.Runtime.CompilerServices.IsBoxed) x) { } .method public instance class [mscorlib]System.ValueType modopt([mscorlib]System.Runtime.CompilerServices.IsConst) modopt([mscorlib]System.SByte) modreq([mscorlib]System.Runtime.CompilerServices.IsBoxed) F(class [mscorlib]System.ValueType modopt(V) modreq([mscorlib]System.Runtime.CompilerServices.IsBoxed) v, class C c) { }
In the case of m2, the signature indicates that the field is a handle to type System::ValueType. The particular kind of value type is then indicated by the value-type special modopt that follows, [mscorlib]System.Int32; that is, type int. Similarly, in the case of m3, this value-type special modopt 194
CLI libraries
is the user-defined type V. The second and third overloads of F also use value-type special modopts, namely [mscorlib]System.Int32 and [mscorlib]System.SByte, to indicate int and signed char, respectively. As suggested by this example, a value-type special modopt can be any value type. As such, C does not result in modopt generation, as that type is a ref type, not a value type. end note] 33.1.5.2 IsByValue This modreq type supports the passing of objects of a ref class type by value. Description: This type is used in the signature of a function. However, it is not used to indicate that a ref class value is returned by a function; for that, see IsUdtReturn (33.1.5.8). [Example:
public ref struct R { static void F(R r) { } }; .class public R { .method public static void F(class R modopt( [mscorlib]System.Runtime.CompilerServices.IsByValue) r) { } }
end example] 33.1.5.3 IsConst This modopt type supports the const qualifier. Description: This type can be used in the signature of any data member or function. Numerous examples of the use of this modifier are shown in 33.1.1, 33.1.3, and 33.1.4. 33.1.5.4 IsExplicitlyDereferenced This modopt type supports the use of interior pointers and pinning pointers. Description: This type can be used in the signature of any function or local variable. [Example:
public ref struct X { void F(interior_ptr<int> x) { } void F(interior_ptr<unsigned char> x) { } }; .method void F(int32& modopt( [mscorlib]System.Runtime.CompilerServices.IsExplicitlyDereferenced) x) { } .method F(unsigned int8& modopt( [mscorlib]System.Runtime.CompilerServices.IsExplicitlyDereferenced) x) { }
end example] 33.1.5.5 IsImplicitlyDereferenced This modopt type supports the reference type punctuators & and %. Description: This type is used in the signature of any data member to indicate that member is a reference. It is also used in a function signature to indicate parameters that are passed by reference or that that function returns by reference. [Example:
195
C++/CLI Language Specification ref class X { int* m1; int& m2; public: void F(int* x) { } void F(int& x) { } void F(X% x) { } int& G() { } }; .field private int32* m1 .field private int32* modopt( [mscorlib]System.Runtime.CompilerServices.IsImplicitlyDereferenced) m2 .method void F(int32* x) { } .method void F(int32* modopt( [mscorlib]System.Runtime.CompilerServices.IsImplicitlyDereferenced) x) { } .method void F(class X modreq([mscorlib] System.Runtime.CompilerServices.IsImplicitlyDereferenced) x) { } .method int32* modopt([mscorlib] System.Runtime.CompilerServices.IsImplicitlyDereferenced) G() { }
end example] 33.1.5.6 IsLong [Note: This modopt type is not part of this Standard; however, it is used by at least one implementation for two unrelated purposes: supporting the types long int and unsigned long int as synonyms for int and unsigned int, respectively, and supporting the type long double as a synonym for double. Description:
IsLong can be used in the signature of any data member or function. public ref class X { int i; long int li; double d; long double ld; public: unsigned int F(unsigned int* pu) { } unsigned long int F(unsigned long int* pul) { } double F(double* pd) { } long double F(long double* pld) { }
};
.field private int32 i .field private int32 modopt([mscorlib]System.Runtime.CompilerServices.IsLong) li .field private float64 d .field private float64 modopt([mscorlib]System.Runtime.CompilerServices.IsLong) ld .method unsigned int32 F(unsigned int32* pu) { } .method unsigned int32 modopt([mscorlib]System.Runtime.CompilerServices.IsLong) F(unsigned int32 modopt([mscorlib]System.Runtime.CompilerServices.IsLong)* pul) { } .method float64 F(float64* pd) { }
196
CLI libraries .method float64 modopt([mscorlib]System.Runtime.CompilerServices.IsLong) F(float64 modopt([mscorlib]System.Runtime.CompilerServices.IsLong)* pld) { }
end note] 33.1.5.7 IsSignUnspecifiedByte This modopt type supports plain chars being a type separate from signed char and unsigned char. Description: This type can be used in the signature of any data member or function. [Example:
public ref class x { char c; signed char sc; unsigned char uc; public: char* F(char* p1) { } char* F(signed char* p2) { } char* F(unsigned char* p2) { } };
The code generated from an implementation in which a plain char is signed, as as follows:
.field private int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte) c .field private int8 sc .field private unsigned int8 uc .method int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* F(int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* p1) { } .method int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* F(int8* p2) { } .method int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* F(unsigned int8* p2) { }
while that generated from an implementation in which a plain char is unsigned, is shown below:
.field private unsigned int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte) c .field private int8 sc .field private unsigned int8 uc .method unsigned int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* F(unsigned int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* p1) { } .method unsigned int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* F(unsigned int8* p2) { } .method unsigned int8 modopt( [mscorlib]System.Runtime.CompilerServices.IsSignUnspecifiedByte)* F(unsigned int8* p2) { }
end example]
197
33.1.5.8 IsUdtReturn This modreq type supports the returning of objects of a ref class type by value. Description: This type is used in the signature of a function. However, it is not used to indicate a ref class value that is passed to a function; for that, see IsByValue (33.1.5.2). [Example:
public ref R() { R(R% r) R F() { }; struct R { } { } }
.method void modreq([mscorlib] System.Runtime.CompilerServices.IsUdtReturn) F(class R& A_1) { }
end example] 33.1.5.9 IsVolatile This modreq type supports the volatile qualifier. (Although IsVolatile is part of the CLI Standard, for convenience, it is documented here as well.) Description: This type can be used in the signature of any data member or function.
volatile-qualified data member, local variable, and parameter declarations shall be marked with this
modreq. Furthermore, each access to such a member, variable, or parameter shall also be marked with this modreq. Any compiler that imports metadata having signature items that contain the volatile modreq is required to use volatile. prefixed instructions when accessing memory locations that are volatile-qualified. [Example:
public ref class x { volatile int* p1; public: void F(volatile int* p2, int* p3) { *p1 = 1; *p2 = 2; *p3 = 3; p1 = 0; } }; .field private int32 modreq([mscorlib]System.Runtime.CompilerServices.IsVolatile)* p1 .method void F(int32 modreq([mscorlib]System.Runtime.CompilerServices.IsVolatile)* p2, int32* p3) { ldarg.0 ldfld int32 modreq([mscorlib] System.Runtime.CompilerServices.IsVolatile)* x::p1 ldc.i4.1 volatile. // prefix instruction needed when dereferencing p1 stind.i4 ldarg.1 ldc.i4.2 volatile. stind.i4 ldarg.2 ldc.i4.3 stind.i4
// prefix instruction needed when dereferencing p2
// no prefix instruction needed when dereferencing p3
198
CLI libraries ldarg.0 ldc.i4.0 stfld int32 modreq([mscorlib] System.Runtime.CompilerServices.IsVolatile)* x::p1 // no prefix instruction needed; not dereferencing p1 ret }
Note that given the declaration volatile int* p1, p1 is not itself volatile-qualified; however, *p1 is. end example]
33.2 Standard attributes

A conforming C++/CLI implementation shall provide the attribute types described below: 33.2.1 NativeCppClass Each native class is encoded in metadata as a value class marked with the attribute NativeCppClass, which is defined as follows:
[System::AttributeUsage(System::AttributeTargets::Struct,Inherited=true)] public ref class NativeCppClassAttribute sealed : System::Attribute { public: NativeCppClassAttribute () { /* */ } };
This type has the following characteristics: Its public key is [00 00 00 00 00 00 00 00 04 00 00 00 00 00 00 00], it has the attribute CLSCompliantAttribute(true), it belongs to the library RuntimeInfrastructure, it resides in the namespace System::Runtime::CompilerServices, and it is part of the assembly mscorlib.
199
34. Metadata
This clause is intended to introduce metadata generation; however, the coverage is not exhaustive. For a definitive description of that topic, refer to the CLI standard, especially Partition II.
34.1 Basic concepts

34.1.1 Importing types from assemblies Ordinarily, when types are referred to in metadata, they are fully qualified using the following form:
[ assembly-name ] namespace-name . type-name
Exceptions are C++/CLI fundamental type names (which are synonyms for CLI built-in type names) and synonyms for CLI built-in type names used directly. [Example:
#using <mscorlib.dll> // redundant #using <System.dll> // needed for Socket #using <System.Xml.dll> // needed for XmlTextReader int main() { System::Text::StringBuilder^ strBld; System::Net::Sockets::Socket^ soc; System::Xml::XmlTextReader^ xtr; int i; System::Int64 j; System::String^ str; System::Object^ obj; // a synonym for System::Int32; // which is equivalent to int32 // equivalent to int64 // " " string // " " object
.method main() { .locals ([0] class [mscorlib]System.Text.StringBuilder V_0, [1] class [System.Xml]System.Xml.XmlTextReader V_1, [2] class [System]System.Net.Sockets.Socket V_2, [3] int32 V_3, [4] int64 V_4, [5] string V_5, [6] object V_6) }
end example]
34.2 Types
34.2.1 Reference types A tracking reference to a ref class or interface class type shall be emitted into metadata as that type with the modopt IsImplicitlyDereferenced (33.1.5.5). A tracking reference to a value class type shall be emitted into metadata as a managed pointer to type without that modopt. [Example:
public ref class R {}; public value class V {}; public interface class I {}; void void void void F1(R% tr1) {} F2(I% tr2) {} F3(V% tr3) {} F4(int% tr3) {}
200
Metadata .method assembly static void F1(class R modreq([mscorlib] System.Runtime.CompilerServices.IsImplicitlyDereferenced) tr1) { } .method assembly static void F2(class I modreq([mscorlib] System.Runtime.CompilerServices.IsImplicitlyDereferenced) tr2) { } .method assembly static void .method assembly static void F3(valuetype V& tr3) { } F4(int32& tr3) { }
end example] 34.2.2 Interior pointers An interior pointer to a type shall be emitted into metadata as a managed pointer to that type with the modopt IsExplicitlyDereferenced (33.1.5.4). [Example:
public ref class R {}; public value class V {}; public interface class I {}; void void void void F1(interior_ptr<R^> ip1) {} F2(interior_ptr<I^> ip2) {} F3(interior_ptr<V> ip3) {} F4(interior_ptr<int> ip3) {}
.method assembly static void F1a(class R& modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) ip1) { } .method assembly static void F2a(class I& modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) ip2) { } .method assembly static void F3a(valuetype V& modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) ip3) { } .method assembly static void F4a(int32& modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) ip3) { }
end example] 34.2.3 Pinning pointers A pinning pointer shall be emitted into metadata with the modifier pinned and the modopt IsExplicitlyDereferenced (33.1.5.4). [Example:
value struct V { int Data; void F() { pin_ptr<V> ppv = this; V* pv = ppv; } }; int main() { V v; pin_ptr<V> ppv = &v; int* pi = &ppv->Data; } .class V { .field public int32 Data .method F() { .locals ([0] valuetype V& pinned modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) V_0, [1] valuetype V* V_1) } }
201
C++/CLI Language Specification .method main() { .locals ([0] valuetype V& pinned modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) V_0, [1] int32* V_1, [2] valuetype V V_2) }
end exaple] 34.2.4 Native arrays The encoding of native arrays in metadata is unspecified. [Note: This does not cause interop problems because such arrays cannot have public visibility. end note]
34.3 Variables
34.3.1 File-scope and namespace-scope variables The encoding of file-scope and namespace-scope variable declarations and definitions in metadata is unspecified. [Note: This does not cause interop problems because such declarations and definitions cannot have public visibility. end note]
34.4 Conversions
34.4.1 String literal conversions When a <narrow-string-literal-type> or <wide-string-literal-type> is converted to System::String^, the result is treated as a CLI string literal. [Example:
void F(String^ s); F("red\t" "car\n"); F("ABC\xFF"); F(L"blue"); F(L"\xFF" L"\xFE"); ldstr call ldstr call ldstr call ldstr call "red\tcar\n" void F(string) bytearray (41 00 42 00 43 00 FF 00 ) void F(string) "blue" void F(string) bytearray (FF 00 FE 00 ) void F(string)
end example] 34.4.2 Boxing conversions A boxing conversion is achieved via the box instruction, as specified in the CLI Standard, Partition III, 4. This causes a runtime bitwise copy of the value class instance to an object on the CLI heap. [Example:
int main() { Console::WriteLine("{0}, {1}", 10, TimeSpan::MinValue); }
202
Metadata .method main() { ldstr "i = {0}" ldc.i4.s 10 box [mscorlib]System.Int32 ldsfld valuetype [mscorlib]System.TimeSpan [mscorlib]System.TimeSpan::MinValue box [mscorlib]System.TimeSpan call void [mscorlib]System.Console::WriteLine(string, object, object) ldc.i4.0 ret }
end example] 34.4.3 Conversion functions In ref classes, implicit conversion functions shall have the name op_Implicit, and explicit conversion functions shall have the name op_Explicit. In native classes, implicit conversion functions shall have the name <op_Implicit>, and explicit conversion functions shall have the name <op_Explicit>. All conversion functions shall be marked specialname. op_Implicit and op_Explicit can be overloaded on their return type. [Example:
public value struct Decimal { static operator Decimal(int value); static explicit operator double(Decimal value); explicit operator float(); }; .class public sequential Decimal { .method public specialname static valuetype Decimal op_Implicit( int32 value) { } .method public specialname static float64 op_Explicit( valuetype Decimal value) { } .method public specialname instance float32 op_Explicit() { } }
end example] Converting constructors are emitted as constructors, never as conversion functions. (Constructors in ref classes and value classes are always explicit.)
34.5 Expressions
34.5.1 Class member access When using an instance of a value type to call a virtual function in a base class (which can only be System::ValueType or System::Object), and that value type does not itself override that function, the instance of the value type shall be boxed. In no other cases shall accessing a member of a value type cause boxing. [Example:
value struct V { virtual int GetHashCode() override { } }; int main() { V v; = v.GetHashCode(); // calls V::GetHashCode = v.ToString(); // calls ValueType::ToString }
203
C++/CLI Language Specification .method main() { .locals ([0] valuetype V V_0) ldloca.s V_0 initobj V ldloca.s V_0 call instance int32 V::GetHashCode() ldloc.0 box callvirt }
V instance string [mscorlib]System.ValueType::ToString()
As V overrides GetHashCode, no box instruction is needed before the call instruction. However, as V does not override ToString, the version from ValueType is used, resulting in a box instruction followed by a callvirt instruction. end example] 34.5.2 Dynamic cast If a run-time check is applied to the cast, and T is a handle or reference to a CLI class type, the run-time check shall be performed using the isinst instruction. 34.5.3 Safe cast When a handle to cv2 B is cast to a handle to cv1 D, a run-time check is performed by the castclass instruction to determine that D inherits from B. The result of the conversion is the result of that instruction. When a cv2 B is cast to a tracking reference to cv1 D, a run-time check is performed by the castclass instruction to determine that D inherits from B. The result is the dereferenced result of castclass. When an rvalue of type handle to cv1 R is converted to an lvalue of type V, the unbox instruction is used.
34.6 Functions
34.6.1 Name lookup On input, the presence or absence of the hidebysig notation in metadata is ignored; all native types are treated as having hidebyname members while all CLI class types are treated as having hidebysig members. [Note: On output, CLI class types shall have each of their members marked hidebysig (34.7.4). end note] 34.6.2 Parameter arrays A function can have a parameter array as its final parameter only. Such a parameter shall result in a
.custom directive for the standard attribute System::ParamArrayAttribute, on the final parameter in the .method directive generated for that function. [Example: void f(... array<Object^>^ p) { } int main() { array<Object^>^ a1 = gcnew array<Object^>(2); array<Object^>^ a2 = gcnew array<Object^>(4); array<Object^>^ a3 = gcnew array<Object^>(8); f(a1); f(a2, a1); f(a1, a3, a2); }
204
Metadata .method assembly static void f(object[] p) { .param [1] .custom instance void [mscorlib]System.ParamArrayAttribute::.ctor() = ( 01 00 00 00 ) }
end example] The final parameter of a function taking a parameter array is a handle to a CLI array of the given type. Calls to such a function shall be translated into an allocation of a CLI array of the given type, with consecutive elements of that array being initialized with the arguments passed to the function, in their lexical order. [Example: Here's an example of using a parameter array with a member function:
public ref struct C { static void F(int val, ... array<String^>^ list) { } static void TestF() { F(10, "red", "blue", "green"); } }; .class public C { .method public static void F(int32 val, string[] list) { .param [2] .custom instance void [mscorlib]System.ParamArrayAttribute::.ctor() = ( 01 00 00 00 ) } .method public static void TestF() { .maxstack 3 .locals (string[] V_0) ldc.i4.3 newarr [mscorlib]System.String stloc.0 ldloc.0 ldc.i4.0 ldstr "red" stelem.ref ldloc.0 ldc.i4.1 ldstr "blue" stelem.ref ldloc.0 ldc.i4.2 ldstr "green" stelem.ref ldc.i4.s 10 ldloc.0 call void C::F(int32, string[]) ret
} }
end example] 34.6.3 Importing native functions If a function has the attribute DllImportAttribute (in namespace System::Runtime::InteropServices), the compiler is required to not preserve that type in metadata as a custom attribute. Instead, the compiler shall emit it directly in the file format. (Consumers of such metadata are required to retrieve this data from the file format and return it as if it were a custom attribute.) The .method directive generated shall be marked with the pinvokeimpl predefined attribute, whose first quoted string is a platform-specific description indicating where the implementation of the function is
205
located, and whose optional second string is the name of the function as it exists on that platform. The body of the method shall be empty. [Example:
// MyCLib.h using namespace System::Runtime::InteropServices; [DllImport("MyCLib.dll", CallingConvention = CallingConvention::StdCall, EntryPoint="Hypot" )] extern "C" double Hypotenuse(double s1, double s2); .method public static pinvokeimpl("MyCLib.dll" as "Hypot" stdcall) float64 Hypotenuse(float64 s1, float64 s2) cil managed {} } // MyCLibApp.cpp #include "MyCLib.h" int main() { Console::WriteLine("Hypotenuse = {0}", Hypotenuse(3, 4)); } .method main() { ldstr "Hypotenuse = {0}" ldc.r8 3. ldc.r8 4. call float64 Hypotenuse(float64, float64) box [mscorlib]System.Double call void [mscorlib]System.Console::WriteLine(string, object) ldc.i4.0 ret }
end example] If a function parameter or return value has the attribute MarshalAsAttribute (in namespace System::Runtime::InteropServices), the compiler is required to not preserve that type in metadata as a custom attribute. Instead, the compiler shall emit it directly in the file format. (Consumers of such metadata are required to retrieve this data from the file format and return it as if it were a custom attribute.) The parameters or return type in the .method directive generated shall be marked with the marshal attribute according to the UnManagedType argument passed. [Example:
using namespace System::Runtime::InteropServices; [DllImport("msvcrt.dll", CallingConvention = CallingConvention::Cdecl)] extern "C" int strcmp([MarshalAs(UnmanagedType::LPStr)] System::String^ s1, [MarshalAs(UnmanagedType::LPStr)] System::String^ s2); .method public static pinvokeimpl("msvcrt.dll" cdecl) int32 strcmp(string marshal(lpstr) s1, string marshal(lpstr) s2) cil managed {}
end example] 34.6.4 Non-member functions The encoding of non-member functions in metadata is unspecified. [Note: This does not cause interop problems because such functions cannot have public visibility. end note]
34.7 Classes and members

34.7.1 Class definitions A ref class, value class, or interface class shall be emitted using a class directive having the corresponding name and visibility. It can be marked with the following: Any one of the "Marshal string" attributes ansi, autocode, or unicode (34.7.3). Any one of the "Type layout" attributes auto, explicit, or sequential (34.7.3).
206
Metadata
Any combination of the "Special handling" attributes beforefieldinit, rtspecialname, serializable, or specialname. (For more information about serialization, see the note below.)
A nested ref class or value class shall be marked nested, followed by the appropriate accessibility, and shall be defined inside the type in which it is nested. A ref class shall be emitted with an extends clause, which specifies either the explicitly given direct base class or the default base class, [mscorlib]System::Object. If the class implements any interfaces, a corresponding implements clause shall be present. A value class shall extend [mscorlib]System::ValueType, it shall have a type layout of sequential, and it shall be marked sealed. An interface class shall be marked interface and abstract. [Example:
public ref class B { }; public ref struct D : B { ref class N { }; }; private value struct S { }; interface class I { }; .class public auto ansi B extends [mscorlib]System.Object { } .class public auto ansi D extends B { .class auto ansi nested public N extends [mscorlib]System.Object { } } .class private sequential ansi sealed S extends: [mscorlib]System.ValueType { } .class interface private abstract auto ansi I { }
end example] The encoded name of a class includes its parent namespaces, if any, with each pair of identifiers being separated by a period. [Example:
namespace NS1 { public struct N { ref struct R1 { }; }; namespace NS2 { public ref struct R2 { value struct V { }; }; } } .class public sequential ansi sealed NS1.N extends [mscorlib]System.ValueType { .class auto ansi nested public R1 extends [mscorlib]System.Object { } } .class public auto ansi NS1.NS2.R2 extends [mscorlib]System.Object { .class sequential ansi sealed nested public V extends [mscorlib]System.ValueType { } }
end example] For information specific to generic types, see 34.18. 207
[Note: The CLI standard does not define the process of serialization and deserialization. However, it does make provision for such a facility by defining a metadata attribute serializable, which can be applied to a class definition. This attribute indicates that, by default, all the instance data members in that type should be persisted when their parent object is serialized. The CLI standard also defines a metadata attribute notserialized, which can be applied to an instance data member definition, to indicate that that member not be persisted when its parent object is serialized. In an extended implementation, these metadata attributes might be generated, by example, by the compiler's recognizing attributes called System::Runtime::Serialization::SerializableAttribute and System::Runtime::Serialization::NonSerializedAttribute, respectively. All of the types in the CLI standard library are required to have the serializable attribute. end note] 34.7.1.1 Abstract classes A ref class explicitly declared abstract shall be emitted as a class marked abstract. [Example:
public ref struct B abstract { }; .class public abstract B { }
end example] 34.7.1.2 Sealed classes A ref class explicitly declared sealed shall be emitted as a class marked sealed. All value classes shall be marked sealed. [Example:
public ref struct B sealed { }; private value struct C { }; .class public sealed B { } .class private sealed C { }
end example] 34.7.2 Member access Each access-specifier has a corresponding metadata accessibility attribute, as follows: C++/CLI Access Specifier Metadata Accessibility Attribute
private protected public internal protected public public protected protected private private protected private family public assembly famorassem famorassem famandassem famandassem
Each member shall have its own accessibility attribute, as required. [Example:
public ref class C { private: int m1; protected: int m2; public: int m3;
208
Metadata internal: int m4; protected public: int m5; public protected: int m6; private protected: int m7; protected private: int m8; }; .class public C { .field private int32 m1 .field family int32 m2 .field public int32 m3 .field assembly int32 m4 .field famorassem int32 m5 .field famorassem int32 m6 .field famandassem int32 m7 .field famandassem int32 m8 }
end example] 34.7.3 Data members Each data member shall correspond to a field having the corresponding type and accessibility attribute. (For information about accessibility of members see 34.7.2.) A static data member shall have the static attribute, while an instance data member shall not. [Example:
public ref class C { int count; float* pCoeff; array<long long int>^ values; C^ next; System::Exception^ lastException; static int objectCount; static String^ name; }; .class public C { .field private int32 count .field private float32* pCoeff .field private int64[] values .field private class C next .field private class [mscorlib]System.Exception lastException .field private static int32 objectCount .field private static string name }
end example] If a static data member contains an initializer, the initialization of the corresponding field shall be done in the parent class's static constructor. If a ref or value class type has the attribute StructLayoutAttribute (in namespace System::Runtime::InteropServices), the compiler is required to not preserve that type in metadata as a custom attribute. Instead, the compiler shall emit it directly in the file format. (Consumers of such metadata are required to retrieve this data from the file format and return it as if it were a custom attribute.) This attribute can be used to specify the layout of a data structure via the auto, explicit, and sequential attributes on the class definition, the alignment (via a .pack directive), the size (via a .size directive), and the marshalling of strings via the ansi, auto, and unicode attributes on the class definition.
209
An instance data member can have the attribute FieldOffsetAttribute (in namespace System::Runtime::InteropServices), which controls the exact placement of that member. As with the attribute StructLayoutAttribute, the compiler shall emit the affects of FieldOffsetAttribute directly in the file format, rather than emitting the attribute itself. [Example:
using namespace System::Runtime::InteropServices; [StructLayout(LayoutKind::Explicit)] public value class S1 { [FieldOffset(0)] int v; [FieldOffset(4)] unsigned char c; [FieldOffset(8)] int w; }; .class public explicit ansi S1 { .pack .size 0 .field [4] private unsigned int8 c .field [0] private int32 v .field [8] private int32 w } [StructLayout(LayoutKind::Sequential, Pack=4)] public value class S2 { int v; unsigned char c; int w; }; .class public sequential ansi S2 { .pack 4 .size 0 .field private unsigned int8 c .field private int32 v .field private int32 w } [StructLayout(LayoutKind::Explicit, Size=12, CharSet=CharSet::Unicode)] public ref class S3 { [FieldOffset(0)] int* pi; [FieldOffset(0)] unsigned int ptrValue; }; .class public explicit unicode S3 { .pack .size 12 .field [0] private int32* pi .field [0] private unsigned int32 ptrValue }
end example] For information about literal and initonly fields see 34.7.11 and 34.7.12, respectively. A field definition can optionally contain the notserialized attribute. (For more information about serialization, see the note in 34.7.1.) Ordinarily, a field shall not be marked rtspecialname or specialname. However, the instance field called value__ that is emitted in an enum's class shall be marked rtspecialname and specialname. Data members can have applied to them the attribute MarshalAsAttribute (in namespace System::Runtime::InteropServices). For metadata information on this attribute, see 34.6.3. 34.7.4 Functions A function shall be emitted as a .method directive. Ordinarily, a method definition shall not be marked rtspecialname or specialname. (Instance and static constructors are exceptions; see 34.7.9 and
210
Metadata
34.7.10, respectively.) The definition of a static function shall be marked static; that for an instance function shall be marked instance. Member functions of ref classes, value classes, and interface classes shall be marked hidebysig. Virtual member functions of ref classes, value classes, and interface classes shall be marked strict, while non-virtual member functions from those types shall not. [Note: The CLI requires that strict virtual methods can only be overridden if they are also accessible. end note] Ordinarily, the name of the method emitted shall be the same as that in its source declaration; however, instance constructors (34.7.9), static constructors (34.7.10), property accessors (34.7.5), event accessors (34.7.6), and static operators (34.7.7) are exceptions. The return type, and the types and order of the parameters in the parameter list emitted shall correspond directly to that in the function's source declaration. The accessibility of a function shall be reflected in the definition of its .method directive. (See 34.7.2.) A method definition shall be marked with the appropriate implementation attributes, such as cil managed (see discussion below). [Example:
public ref class C { static void compressData(int* p1, String^ p2, Object^ p3) { } public: void Initialize() { } void Initilaize(int i, int j) { } virtual void Display() { } }; .class public C { .method private hidebysig static void compressData(int32* p1, string p2, object p3) cil managed { } .method public hidebysig instance void Initialize() cil managed { } .method public hidebysig instance void Initilaize(int32 i, int32 j) cil managed { } .method public hidebysig strict newslot virtual instance void Display() cil managed { }
end example] 34.7.4.1 Override functions Use of an override-specifier shall always result in an .override directive in the metadata, while use of the function-modifier override without an override-specifier shall not. [Example: Given the following code
public ref struct B { virtual void F() {}; virtual void F(int i) {}; }; public ref struct D1 : B { virtual void F() override {} }; public ref struct D2 : B { virtual void F() override {} virtual void G(int i) = B::F {} }; public ref struct D3 : B { virtual void F() = B::F {} }; // explicitly overrides B::F()
// explicitly overrides B::F() // named override of B::F(int)
// explicitly overrides B::F()
the relevant metadata generated for classes D2 and D3 is as follows:
211
C++/CLI Language Specification .class public D2 extends B { .method public virtual instance void F() { } .method public newslot virtual final instance void G(int32 i) { .override B::F // overrides B::F(int32) } } .class public D3 extends B { .method public newslot virtual final instance void F() { .override B::F // overrides B::F() } }
end example] 34.7.4.2 Sealed function modifier A ref class function explicitly declared sealed shall be emitted as a method marked final. [Example:
public ref struct R { virtual void F() sealed { } }; .class R { .method final instance void F() { } }
end example] 34.7.4.3 Abstract function modifier A ref class function explicitly declared abstract shall be emitted as a method marked abstract. [Example:
public ref virtual virtual virtual }; struct R { void F1() = 0; void F2() abstract; void F3() abstract = 0;
.class abstract R { .method abstract void F1() { } .method abstract void F2() { } .method abstract void F3() { } }
end example] All instance functions in an interface class shall be emitted as methods marked abstract. 34.7.4.4 The newslot attribute The new function modifier corresponds exactly to the CLIs predefined attribute newslot. [Note: According to the CLI Standard, Partition II: A virtual method is introduced in the inheritance hierarchy by defining a virtual method. The versioning semantics differ depending on whether or not the definition is marked as newslot: If the definition is marked newslot then the definition always creates a new virtual method, even if a base class provides a matching virtual method. Any reference to the virtual method created before the new virtual function was defined will continue to refer to the original definition. If the definition is not marked newslot then the definition creates a new virtual method only if there is no virtual method of the same name and signature inherited from a base class. If the
212
Metadata
inheritance hierarchy changes so that the definition matches an inherited virtual function, the definition will be treated as a new implementation of that inherited function. end note] Functions shall be marked newslot in the following cases only: The function is a member of an interface. The function is a virtual function in a ref class or value class and that function's name is not seen by lookup in any of the base classes. [Note: Lookup ignores interfaces, so if the name is specified only in an interface, the function is still marked as newslot. end note] The function is a virtual function declared using new.
34.7.4.5 Special attributes The attributes InAttribute and OutAttribute (both in namespace System::Runtime::InteropServices) can be applied to function parameters. The compiler is required to not preserve these types in metadata as custom attributes. Instead, the compiler shall emit them directly in the file format. (Consumers of such metadata are required to retrieve this data from the file format and return it as if it were a custom attribute.) [Example:
public ref struct C { void F(int* p1, [In] int* p2, [Out] int* p3, [In, Out] int* p4) { } }; .class public C { .method public instance void F(int32* p1, [in] int32* p2, [out] int32* p3, [in][out] int32* p4) { } }
end example] A method definition can be marked with a variety of implementation attributes. Some of these can be specified via the attribute MethodImplAttribute (in namespace System::Runtime::CompilerServices), which takes as an argument, one or a combination of enumerators from the type MethodImplOptions (also in the same namespace). The compiler is required to not preserve this type in metadata as a custom attribute. Instead, the compiler shall emit it directly in the file format. (Consumers of such metadata are required to retrieve this data from the file format and return it as if it were a custom attribute.) [Example:
public ref struct C { [MethodImpl(MethodImplOptions::NoInlining)] void F1() { } [MethodImpl(MethodImplOptions::Synchronized | MethodImplOptions::NoInlining)] void F2() { } }; .class public C { .method public instance void .method public instance void noinlining { } } F1() noinlining { } F2() synchronized
end example] 34.7.5 Properties A property shall be emitted as a .property directive plus one .method directive for each accessor. No other methods shall be emitted. If the property has a get accessor function, the .property directive shall contain a .get directive. If the property has a set accessor function, the .property directive shall contain a .set directive. The method definitions shall be marked specialname. A property itself shall not be marked rtspecialname or specialname.
213
The definition of an instance property shall be marked instance. Any .set and .get directives that property contains shall also be marked instance, as shall the corresponding method definitions. For a static property, only the method definition shall be marked static. For a scalar or named indexed property P, the name of the method emitted for a get accessor function shall be get_P, while that for a set accessor function shall be set_P. For a default-indexed property declared in a type not having the attribute DefaultMemberAttribute, the metadata emitted shall be as if that property were a named indexed property called Item. For a default-indexed property declared in a type having the attribute DefaultMemberAttribute, the metadata emitted shall be as if that property were a named indexed property having the name specified by that attribute. The accessibility of a property shall be reflected in the definitions of its .methods. (See 34.7.2.) [Note: The get and set accessor functions of a property can have different accessibilities. end note] [Example:
public value class Point { static int pointCount = 0; int x; int y; public: property int X { int get() { return x; } void set(int val) { x = val; } } static property int PointCount { int get() { return pointCount; } }
};
.class public Point { .property instance int32 X() { .set instance void Point::set_X(int32) .get instance int32 Point::get_X() } .method public specialname instance int32 get_X() { } .method public specialname instance void set_X(int32 val) { } .property int32 PointCount() { .get int32 Point::get_PointCount() } .method public specialname static int32 get_PointCount() { } }

public ref class IntVector { int length; array<int>^ values; public: property int default[int] { int get(int index) void set(int index, int value) } } .class public IntVector { .field private int32 length .field private int32[] values
{ return values[index]; } { values[index] = value; }
214
Metadata .property instance int32 Item(int32) { .get instance int32 IntVector::get_Item(int32) .set instance void IntVector::set_Item(int32, int32) } .method public int32 get_Item(int32 index) { } .method public void set_Item(int32 index, int32 value) { } }
end example] If a property is declared virtual, the accessor methods it has shall be marked newslot virtual. If a property is not declared virtual, but either of the two of its accessors, or its only accessor is, then the accessor emitted shall be marked newslot virtual. If a property is declared sealed, the accessor methods it has shall be marked newslot virtual final. If a property is not declared sealed, but either of the two of its accessors, or its only accessor is, then the accessor emitted shall be marked newslot virtual final. If a property is declared abstract, the accessor methods it has shall be marked newslot abstract virtual. If a property is not declared abstract, but either of the two of its accessors, or its only accessor is, then the accessor emitted shall be marked newslot abstract virtual. In the case of a trivial scalar property, the private backing storage field allocated shall have a name in the implementer's namespace, and be an instance or static field, as appropriate. [Example:
public ref struct C { property int P; }; .class public C { .field private int32 '<backing_store>P' .property instance int32 P() { .set instance void C2::set_P(int32) .get instance int32 C2::get_P() } .method int32 get_P() { .maxstack 1 .locals (int32 V_0) ldarg.0 ldfld int32 C2::'<backing_store>P' stloc.0 ldloc.0 ret } .method void set_P(int32 __set_formal) { .maxstack 2 ldarg.0 ldarg.1 stfld int32 C2::'<backing_store>P' ret } }
end example] The accessor methods of a property can be marked with a variety of implementation attributes. For more information see 34.7.4. 34.7.6 Events An event is implemented via an .event directive. That directive shall refer to one add and one remove accessor function by using an .addon and a .removeon directive, respectively. For an event having a raise accessor function, that function shall be referred to in the .event directive using a .fire directive. The name of the add, remove, and raise accessor functions shall be add_xx, remove_xx, and raise_xx, respectively, where xx is the declared name of the event. All accessor functions shall be marked
215
C++/CLI Language Specification specialname. If the add or remove accessor functions have the attribute MethodImpl(MethodImplOptions::Synchronized), the resulting methods shall be marked synchronized (see 34.7.4). [Example: public delegate void EvtHandler(Object^ sender, EventArgs^ e); public ref class Button { EvtHandler^ action; public: event EvtHandler^ Click { [MethodImpl(MethodImplOptions::Synchronized)] void add(EvtHandler^ d) {} [MethodImpl(MethodImplOptions::Synchronized)] void remove(EvtHandler^ d) { } void raise(Object^ sender, EventArgs^ e) { } } }; .class public Button { .field private class EvtHandler action .event specialname EvtHandler Click { .addon instance void Button::add_Click(class EvtHandler) .removeon instance void Button::remove_Click(class EvtHandler) .fire instance void Button::raise_Click(object, class [mscorlib]System.EventArgs) } .method public specialname instance void add_Click(class EvtHandler d) synchronized { } .method public specialname instance void remove_Click(class EvtHandler d) synchronized { } .method public specialname instance void raise_Click(object sender, class [mscorlib]System.EventArgs e) { }
end example] A trivial event is handled in much the same way as a non-trivial one, except that for a trivial event, storage shall be allocated for a field to hold the delegate, and add, remove, and raise accessor functions shall be generated to add and remove functions from the delegate field, and raise the event, respectively. The generated add and remove accessor functions shall have the same access specifier as their parent event. The generated raise accessor function shall be marked family. The generated add accessor function shall combine the delegate argument passed to it with the delegate field. The generated remove accessor function shall remove the delegate argument passed to it from the delegate field. The generated raise accessor function shall call the delegate field's Invoke method, passing it the argument list the raise accessor function was given; that accessor function shall return the value returned by that call to Invoke. In order to be thread-safe, the generated add and remove accessor functions shall be marked synchronized. The generated raise access function shall not be so marked. [Example:
public delegate int D(int); public ref struct X { event D^ Ev; }; .class public X { .field private class D '<Ev>' .event specialname D Ev { .addon instance void X::add_Ev(class D) .removeon instance void X::remove_Ev(class D) .fire instance int32 X::raise_Ev(int32) }
216
Metadata .method public specialname instance void add_Ev(class D '<value>') synchronized { ldfld class D X::'<Ev>' call class [mscorlib]System.Delegate [mscorlib]System.Delegate::Combine(class [mscorlib]System.Delegate, class [mscorlib]System.Delegate) stfld class D X::'<Ev>' } .method public specialname instance void remove_Ev(class D '<value>') synchronized { ldfld class D X::'<Ev>' call class [mscorlib]System.Delegate [mscorlib]System.Delegate::Remove(class [mscorlib]System.Delegate, class [mscorlib]System.Delegate) stfld class D X::'<Ev>' } .method family specialname instance int32 raise_Ev(int32 value0) { ldfld class D X::'<Ev>' callvirt instance int32 D::Invoke(int32) ret }
end example] 34.7.7 Static operators When an implementation emits metadata for a CLS-compliant operator, it shall translate the C++ operator function identifier to its respective CLS-compliant name, as shown in Table 19-1: CLS-Compliant Unary Operators and Table 19-2: CLS-Compliant Binary Operators. When an implementation imports functions from metadata, it shall rewrite that function's CLS-compliant name as its corresponding C++ operator function identifier, as indicated by these tables.. If an operator function does not match the criteria for a CLS-compliant operator (19.7.5.1), the operator is C++-dependent. Table 19-4: C++-Dependent Unary Operators and Table 19-5: C++-Dependent Binary Operators identify these functions. When an implementation imports C++-dependent functions (Table 19-4: C++-Dependent Unary Operators and Table 19-5: C++-Dependent Binary Operators) from metadata, these functions shall be treated using their corresponding C++ identifiers. If such a function does not make sense as an operator function (for example, it takes three arguments), the function name shall not be changed to the internal operator function name, and the function shall be callable by the name it has in the metadata. All static operator functions shall be marked static and specialname. [Example:
217
C++/CLI Language Specification public ref class IntVector { public: static IntVector^ operator+(IntVector^ iv, int i); static IntVector^ operator+(int i, IntVector^ iv); static IntVector^ operator+(IntVector^ iv1, IntVector^ iv2); static IntVector^ operator-(IntVector^ iv); static IntVector^ operator++(IntVector^ iv); }; .class public IntVector { .method public specialname static class IntVector op_Addition( class IntVector iv, int32 val) { } .method public specialname static class IntVector op_Addition( int32 val, class IntVector iv) { } .method public specialname static class IntVector op_Addition( class IntVector iv1, class IntVector iv2) { } .method public specialname static class IntVector op_UnaryNegation( class IntVector iv) { } .method public specialname static class IntVector op_Increment( class IntVector iv) { } }
end example] 34.7.8 Non-static operators The metadata for non-static operators implemented as member functions is just like that for static operators, except that in the former case, the function is implemented as an instance method instead of a static one. All non-static operator functions shall be marked specialname. As with Standard C++, instance versions of operator++ and operator-- have to be implemented separately for prefix and postfix notation. [Example:
public ref class IntVector { public: IntVector^ operator+(int val); static IntVector^ operator+(int val, IntVector^ iv); IntVector^ operator+(IntVector^ iv2); IntVector^ operator-(); IntVector^ operator++(); IntVector^ operator++(int); }; .class public IntVector { .method public specialname class IntVector op_Addition(int32 val) { } .method public specialname static class IntVector op_Addition( int32 val, class IntVector iv) { } .method public specialname class IntVector op_Addition( class IntVector iv2) { } .method public specialname class IntVector op_UnaryNegation() { } .method public specialname class IntVector op_Increment() { } } .method public specialname class IntVector op_Increment(int32) { }
The function operator+(int, Intvector^) cannot be implemented as an instance method as its first parameter is not of the parent class type or a handle to that type. end example]
218
Metadata
In the case of operators implemented as global functions, they shall be marked assembly, and their names shall be the exact spelling of their source language token; '+' for operator+, '-' for operator-, '++' for operator++, and so on. As with Standard C++, instance versions of operator++ and operator-have to be implemented separately for prefix and postfix notation. [Example:
public ref class IntVector { }; IntVector^ IntVector^ IntVector^ IntVector^ IntVector^ IntVector^ operator+(IntVector^ iv, int val); operator+(int val, IntVector^ iv); operator+(IntVector^ iv1, IntVector^ iv2); operator-(IntVector^ iv); operator++(IntVector^ iv); operator++(IntVector^ iv, int);
.class public IntVector { } .class public abstract '' { .method assembly specialname static class IntVector '+'( class IntVector iv, int32 val) { } .method assembly specialname static class IntVector '+'( int32 val, class IntVector iv) { } .method assembly specialname static class IntVector '+'( class IntVector iv1, class IntVector iv2) { } .method assembly specialname static class IntVector '-'( class IntVector iv) { } .method assembly specialname static class IntVector '++'( class IntVector iv) { } .method assembly specialname static class IntVector '++'( class IntVector iv, int32) { } }
end example] 34.7.9 Instance constructors An instance constructor of a ref class shall be emitted as an instance method, called .ctor, of its class. The accessibility of the constructor shall be reflected in its definition (see 34.7.2). The method shall be marked specialname, rtspecialname, instance, cil, and managed, and shall have a void return type and corresponding parameter list. [Example:
public ref class C { int v; C() { } public: C(int i) : v(i) { } }; .class public C { .method private specialname rtspecialname instance void .ctor() { .maxstack ldarg.0 call instance void [mscorlib]System.Object::.ctor() ret }
219
C++/CLI Language Specification .method public specialname rtspecialname instance void .ctor(int32 i) { .maxstack ldarg.0 call instance void [mscorlib]System.Object::.ctor() ldarg.0 ldarg.1 stfld int32 C::v ret }
end example] An instance constructor can be marked with a variety of implementation attributes. For more information see 34.7.4. 34.7.10 Static constructors A static constructor of a ref or value class shall be emitted as a private static method, called .cctor, of its class. The method shall be marked specialname, rtspecialname, static, cil, and managed, and shall have a void return type and no arguments. The class itself shall be marked beforefieldinit. [Example:
public ref class B { static B() { } public: }; .class public beforefieldinit B { .method private specialname rtspecialname static void .cctor() cil managed { } }
end example] A static constructor can be marked with a variety of implementation attributes. For more information see 34.7.4. 34.7.11 Literal fields A literal field shall be implemented as a public static literal field with the specified initial value. [Example:
public ref struct X { literal int Count = 100; literal String^ Greeting = "Hello"; }; .class public X { .field public static literal int32 Count = int32(0x00000064) .field public static literal string Greeting = "Hello" }
end example] For information about metadata generation for data members in general, see 34.7.3. 34.7.12 Initonly fields An initonly field shall be implemented as an instance or static initonly field, as appropriate. The accessibility of the field shall be reflected in its definition. The initialization code placed in the static constructor for each explicitly initialized static initonly field shall cause those fields to be initialized in their declaration lexical order. [Example:
220
Metadata public ref class X { initonly static int V1 = 5, V2 = V1; initonly static int V3 = V2 + 1; initonly static int V4; public: initonly int V5; static X() { V4 = V1 + V3; } X(int i) { V5 = i; } }; .class public X { .field private static initonly int32 .field private static initonly int32 .field private static initonly int32 .field private static initonly int32 .field public initonly int32 V5 V1 V2 V3 V4
.method private specialname rtspecialname static void .cctor() { .maxstack 2 ldc.i4.5 stsfld int32 X::V1 ldsfld stsfld ldsfld ldc.i4.1 add stsfld ldsfld ldsfld add stsfld ret } } int32 X::V1 int32 X::V2 int32 X::V2 int32 X::V3 int32 X::V1 int32 X::V3 int32 X::V4
In the static constructor, V1, V2, and V3 shall be initialized in that order, all before the assignment to V4. end example] For information about metadata generation for data members in general, see 34.7.3. 34.7.13 Destructors and finalizers 34.7.13.1 CLI dispose pattern C++/CLI implements the destructor and finalizer semantics in ref classes by using the CLI dispose pattern. This pattern makes use of three functions upon which all languages targeting the CLI agree. These functions are
void Dispose(); void Dispose(bool); void Finalize();
and their definitions are generated by the compiler, as required. Two other C++/CLI-specific private helper functions are also generated, and used by Dispose(bool); they are:
void __identifier(~T)() void __identifier(!T)()
where T is the parent class name. Many languages have constructs that support this dispose pattern directly. Since C++/CLI fully supports this dispose pattern, any CLI class type authored in C++/CLI can be used by other languages, and any CLI class type authored in other languages and having this dispose pattern, supports C++ destructor cleanup semantics when used in C++/CLI code. The CLI dispose pattern requires the following:
221
A function Dispose() that implements System::IDisposable::Dispose(). A function Finalize() that overrides System::Object::Finalize(). A function Dispose(bool), which is a member of a class that has a Dispose() function that implements System::IDisposable::Dispose(), or is a member of a class that has a Finalize() function that overrides System::Object::Finalize(), or the Dispose(bool) function itself overrides a Dispose(bool) function in a base class that does have such a Dispose() or Finalize() function.
A C++/CLI program that contains a definition for a function having any of these signatures is ill-formed. [Note: It would be helpful to the programmer if the diagnostic issued in such cases encouraged the programmer to define a destructor and/or finalizer instead. end note] Function definitions having these signatures can exist, however. If a function definition having any of these signatures fulfills the corresponding requirement above, it shall be used to implement the CLI dispose pattern, and a C++/CLI program that calls such a function is illformed. [Note: It would be helpful to the programmer if the diagnostic issued in such cases encouraged the programmer to call the destructor instead. end note] If a function definition having any of these signatures does not fulfill the corresponding requirement above, it shall not be used to implement the CLI dispose pattern, and a C++/CLI program is permitted to call that function directly. The System::IDisposable interface is used by the CLI dispose pattern as an entry point for destruction. However, because C++/CLI provides direct support for cleanup via destructors and finalizers, the System::IDisposable interface need never be used directly. A C++/CLI program shall not use this interface. [Example:
public ref class B { protected: !B() {} public: ~B() {} }; public ref class D : B { protected: !D() {} public: ~D() {} }; .class B implements [mscorlib]System.IDisposable { .method void '!B'() { } .method void Dispose(bool marshal( unsigned int8) A_1) { ldarg.1 brfalse.s IL_000b ldarg.0 call instance void B::'~B'() br.s IL_001b IL_000b: nop .try { ldarg.0 call leave.s }
instance void B::'!B'() IL_001b
finally { ldarg.0 call instance void [mscorlib]System.Object::Finalize() endfinally } IL_001b: ret }
222
Metadata .method void Dispose() { ldarg.0 ldc.i4.1 callvirt instance void B::Dispose(bool) ldarg.0 call void [mscorlib]System.GC::SuppressFinalize(object) ret } .method void Finalize() { ldarg.0 ldc.i4.0 callvirt instance void B::Dispose(bool) ret } .method void } .class D extends B { .method void '!D'() { } .method void Dispose(bool marshal( unsigned int8) A_1) { ldarg.1 brfalse.s IL_0015 .try { ldarg.0 call leave.s } '~B'() { }
instance void D::'~D'() IL_0013
finally { ldarg.0 ldc.i4.1 call instance void B::Dispose(bool) endfinally } IL_0013: br.s IL_0015: nop .try { ldarg.0 call leave.s } IL_0026
instance void D::'!D'() IL_0026
finally { ldarg.0 ldc.i4.0 call instance void B::Dispose(bool) endfinally } IL_0026: ret } .method void } '~D'() { }
end example] 34.7.13.2 Destructors A ref class with a user-defined or compiler-generated destructor shall be marked as implementing System::IDisposable. Destruction of an instance of a ref class shall always begin by dynamically casting that object to System::IDisposable. If that cast succeeds, the Dispose() function shall be called through the result
223
of the cast. If that cast fails, the destructor does nothing. [Note: As a result, a destructor can be called on an instance of any ref class, value class, or interface class. end note] The compiler shall not generate code to call a destructor except through the
System::IDisposable::Dispose function.
Although a value class cannot have a destructor, if a value class indirectly implements
System::IDisposable (as the result of another interfaces implementing System::IDisposable), the compiler shall emit a corresponding Dispose() function that implements the interface; however, that Dispose() function shall do nothing.
For an interface class declaring a destructor, no method shall be emitted for that destructor; however, the interface shall be marked as implementing System::IDisposable. 34.7.13.3 Finalizers A finalizer for a class shall be generated if and only if the user writes a finalizer for that class. Calls to a finalizer in any ref class T result in direct calls to the __identifier(!T) function (34.7.13.9). 34.7.13.4 Functions generated to support the dispose pattern The CLI dispose pattern uses three primary functions: Dispose(), Finalize(), and Dispose(bool). Two secondary functions, __identifier(~T)() and __identifier(!T)(), are called by Dispose(bool). The definitions of all five functions are generated by the compiler, as specified below. 34.7.13.5 The Dispose() function This member function is the starting point for cleanup done via destruction. This function shall only be emitted for any ref class T in the following scenarios: The Dispose(bool) function is being introduced by class T (Cases #2 and #3 below), or If Case #1 was used and no base class that used Case #1 has already introduced a public virtual Dispose() that implements System::IDisposable. If the dispose pattern already exists, and A Dispose() that is part of the dispose pattern also exists, and The class explicitly implements System::IDisposable.
This function shall not be emitted
This function shall be emitted as if it were written in C++/CLI, inside the definition of T, as follows:
public: virtual void Dispose() sealed { this->Dispose(true); System::GC::SuppressFinalize(this); }
The parent class of any Dispose() function emitted by the compiler, shall be marked as implementing System::IDisposable. If a base class of T has a Dispose() method that does not implement System::IDisposable, that base class function shall be hidden by the one emitted for T. The Dispose() function shall be marked newslot in metadata unless the function can override a base classs implementation of Dispose() that implements System::IDisposable. 34.7.13.6 The Finalize() function This function is the starting point for cleanup done via finalization.
224
Metadata
This function shall only be emitted for any ref class T if the following criteria are met: The compiler will generate an __identifier(!T) function for class T, and Class T is introducing the dispose pattern (Cases #2 and #3 below), or if class T is extending the dispose pattern (Case #1 below), no base class with the dispose pattern has already introduced a Finalize() function.
This function shall be emitted as if it were written in C++/CLI, inside the definition of T, as follows:
protected: virtual void Finalize() override { this->Dispose(false); }
The Finalize() function shall never be marked newslot in metadata. 34.7.13.7 The Dispose(bool) function For any ref class T, this function is generated if and only if either or both of the functions __identifier(~T)() and __identifier(!T)() are generated for this class or the compiler needs to generate a non-trivial destructor to clean up members of that class. This function has three possible forms, as shown in Case #1, Case #2, and Case #3, below. (In each Case, the base class of T is assumed to be Base. It is also assumed that class T has both a destructor and a finalizer. If one or the other of these functions is omitted, the corresponding call to __identifier(~T) or __identifier(~T) shall be omitted.) The decision tree following these Cases shows how each Case is chosen. Case #1: Extending the dispose pattern, existing Dispose(bool) that is part of the dispose pattern
protected: virtual void Dispose(bool calledFromDispose) override { if (calledFromDispose) { try { this->__identifier("~T")(); } finally { try { this->Base::Dispose(true); } finally { // member cleanup goes here } } } else { try { this->__identifier("!T")(); } finally { this->Base::Dispose(false); } } }
Case #2: Introducing dispose pattern, no public Dispose() that implements System::IDisposable
225
C++/CLI Language Specification protected: virtual void Dispose(bool calledFromDispose) { if (calledFromDispose) { this->__identifier("~T")(); } else { try { try { this->__identifier("!T")(); } finally { // member cleanup goes here } } finally { this->Base::Finalize(); } } }
Case #3: Introducing dispose pattern, existing callable Dispose()

protected: virtual void Dispose(bool calledFromDispose) { if (calledFromDispose) { try { this->__identifier("~T")(); } finally { try { this->Base::Dispose(); } finally { // member cleanup goes here } } } else { try { this->__identifier("!T")(); } finally { this->Base::Finalize(); } } }
226
Metadata
227
34.7.13.8 The __identifier(~T)() function This function shall be emitted for any ref class T, but only if that class has a user-defined destructor. The body of this function shall correspond exactly to that of the user-defined destructor. The compiler shall not generate calls to functions in the base class in this function. This function shall be emitted as if it were written in C++/CLI, inside the definition of T, as follows:
private: void __identifier("~T")() { // user-defined destructor body goes here }
34.7.13.9 The __identifier(!T)() function This function shall be emitted for any ref class T, but only if that class has a user-defined finalizer. The body of this function shall correspond exactly to that of the user-defined finalizer. The compiler shall not generate any other code in this function. This function shall be emitted as if it were written in C++/CLI, inside the definition of T, as follows:
private: void __identifier("!T")() { // user-defined finalizer body goes here }
34.8 Native classes

A native class shall be emitted as a value class (even though a native class is not a value class) with the corresponding name and visibility (34.6.3). It shall be marked with the following: The "Marshal string" attributes ansi (34.7.3), and The "Type layout" attribute sequential (34.7.3),
however, the corresponding attribute, StructLayoutAttribute (and FieldOffsetAttribute), from namespace System::Runtime::InteropServices cannot be applied to a native class at the source code level. A nested native class or value class shall be marked nested, followed by the appropriate accessibility, and shall be defined inside the type in which it is nested. Like a value class, a native class shall extend [mscorlib]System::ValueType. The value class used to encode the native class shall contain an explicit .size directive whose value is determined by the implementation, as the size needed to represent an instance of that class. The value class used to encode the tnative class shall have attached to it the NativeCppClass (33.2.1) attribute, from namespace System::Runtime::CompilerServices. The encoding for a native class is not required to have any other characteristics. In particular, it is not required to have a constructor or the members of the class encoded. [Example:
public class N1 { char c[2]; int i; double d; public: void F() { } }; .class public sequential ansi sealed N1 extends [mscorlib]System.ValueType { .size 16 .custom instance void [mscorlib]System.Runtime.CompilerServices. NativeCppClassAttribute::.ctor() = ( )}
228
Metadata
The size 16 bytes is based on an implementation in which a char occupies 1 byte, an int occupies 4 bytes, a double occupies 8 bytes, a char can be aligned on any boundary, an int is aligned on a 4-byte boundary, and a double is aligned on an 8-byte boundary. (That is, two 1-byte chars, two bytes of padding, one 4-byte int, and one 8-byte double.)
namespace MyApp { public class N2 { char c[3]; double d; int i; public: void F(int i) { } class N3 { short int s; public: void F(int i) { }; }; }
.class public sequential ansi sealed MyApp.N2 extends [mscorlib]System.ValueType { .size 24 .custom instance void [mscorlib]System.Runtime.CompilerServices. NativeCppClassAttribute::.ctor() = ( ) .class sequential ansi sealed nested public N3 extends [mscorlib]System.ValueType { .size 2 .custom instance void [mscorlib]System.Runtime.CompilerServices. NativeCppClassAttribute::.ctor() = ( ) } }
The size 24 bytes comes from three 1-byte chars, five bytes of padding, one 8-byte double, one 4-byte int, one 2-byte short, and two bytes of padding. The size 2 bytes comes from one 2-byte short.
template<typename T> public class N4 { T m1; T m2[2]; public: void F(T t, T* pt) {} }; N4<char> n4a; N4<int> n4b; .class public sequential ansi sealed 'N4<char>' extends [mscorlib]System.ValueType { .size 3 .custom instance void [mscorlib]System.Runtime.CompilerServices. NativeCppClassAttribute::.ctor() = ( ) } .class public sequential ansi sealed 'N4<int>' extends [mscorlib]System.ValueType { .size 12 .custom instance void [mscorlib]System.Runtime.CompilerServices. NativeCppClassAttribute::.ctor() = ( ) }
The encodings of n4a and n4b are not shown. end example] Metadata for template classes is described in 34.17.
229
34.9 Ref classes

[Note: For implementations providing the IsBoxed modifier: Any member function of a ref class, value class, or interface class having a parameter declaration or return type involving a handle to a value type shall have that parameter and/or return type marked with the modifier IsBoxed (33.1.5.1). end note] Any member function of a ref class, value class, or interface class having a ref class type parameter passed by value shall have the corresponding parameter marked with the modifier IsByValue (33.1.5.2). Any member function of a ref class, value class, or interface class having a const-qualified parameter or returning a const-qualified type shall have the corresponding parameter and/or return type marked with the modifier IsConst (33.1.5.3), as appropriate. However, parameter qualification at the top level shall not be so marked. [Example: A parameter such as const int* ci shall be marked, but one such as const int i shall not. end example] Any data member of a ref class, value class, or interface class having a const-qualified type shall be marked with the modifier IsConst (33.1.5.3). Any member function of a ref class, value class, or interface class having a parameter that is an interior pointer or pinning pointer shall have the corresponding parameter marked with the modifier IsExplicitlyDereferenced (33.1.5.4). Any member function of a ref class, value class, or interface class having a parameter that is a reference or tracking reference, or returning a reference or tracking reference shall have the corresponding parameter and/or return type marked with the modifier IsImplicitlyDereferenced (33.1.5.5). Any data member of a ref class, value class, or interface class that is a reference or tracking reference shall be marked with the modifier IsImplicitlyDereferenced (33.1.5.5). [Note: For implementations providing the IsLong modifier: Any member function of a ref class, value class, or interface class having a parameter declaration or return type involving a long int or long double shall have that parameter and/or return type marked with the modifier IsLong (33.1.5.6). Any data member of a ref class, value class, or interface class involving a long int or long double shall have that parameter and/or return type marked with the modifier IsLong (33.1.5.6). end note] Any member function of a ref class, value class, or interface class having a parameter declaration or return type involving a plain char shall have that parameter and/or return type marked with the modifier IsSignUnspecifiedByte (33.1.5.7). Any data member of a ref class, value class, or interface class involving a plain char shall be marked with the modifier IsSignUnspecifiedByte (33.1.5.7). Any member function of a ref class, value class, or interface class returning an instance of a ref class type by value shall be marked with the modifier IsUdtReturn (33.1.5.8). Any member function of a ref class, value class, or interface class having a volatile-qualified parameter or returning a volatile-qualified type shall have the corresponding parameter and/or return type marked with the modifier IsVolatile (33.1.5.9), as appropriate. However, parameter qualification at the top level shall not be so marked. [Example: A parameter such as volatile int* vi shall be marked, but one such as volatile int v shall not. end example] Any data member of a ref class, value class, or interface class having a volatile-qualified type shall be marked with the modifier IsVolatile (33.1.5.9). For more information, see 34.7.1.
34.10 Value classes

For more information, see 34.7.1 and 34.9.
230
Metadata
34.11 CLI arrays

CLI arrays are encoded in metadata according to the CLI standard, primarily in Partitions I, II, and III. [Note: A CLI array type shall be defined by specifying the element type of the CLI array, the rank of the CLI array, and the upper and lower bounds of each dimension of the CLI array. CLI array elements shall be laid out within the CLI array object in row-major order. The actual storage allocated for each CLI array element can include platform-specific padding. The VES shall provide two constructors for arrays: The first takes a sequence of integers giving the number of elements in each dimension (a lower bound of zero is assumed). The second takes twice as many arguments. These arguments occur in pairsone pair per dimensionwith the first argument of each pair specifying the lower bound for that dimension, and the second argument specifying the total number of elements in that dimension.
In addition to array constructors, the VES provides the instance methods Get, Set, and Address to access specific elements and compute their addresses. These methods take a number for each dimension, to specify the target element. In addition, Set takes an additional final argument specifying the value to be stored into the target element. end note] [Example:
ref class R { array<int>^ m1; array<array<String^>^, 2>^ m2; public: array<String^, 2>^ F(array<R^, 3>^ ary) { } }; .class R { .field private int32[] m1 .field private string[][0...,0...] m2 .method public instance string[0...,0...] F(class R[0...,0...,0...] ary) { } } array<int>^ array1D = gcnew array<int>(10); array<int, 3>^ array3D = gcnew array<int, 3>(10, 20, 30); pin_ptr<int> pp1; .method { .locals ([0] int32[0...,0...,0...] V_0, [1] int32[] V_1) [2] int32& pinned modopt([mscorlib] System.Runtime.CompilerServices.IsExplicitlyDereferenced) V_2) ldnull stloc.1 ldnull stloc.0 ldc.i4.s newarr stloc.1 ldloc.1 ldc.i4.5 ldc.i4.s stelem.i4 ldc.i4.s ldc.i4.s ldc.i4.s newobj stloc.0 10 [mscorlib]System.Int32
10 10 20 30 instance void int32[0...,0...,0...]::.ctor(int32, int32, int32)
231
C++/CLI Language Specification array1D[5] = 10; array3D[1,2,3] = array3D[4,5,6]; ldloc.0 ldc.i4.1 ldc.i4.2 ldc.i4.3 ldloc.0 ldc.i4.4 ldc.i4.5 ldc.i4.6 call call instance int32 int32[0...,0...,0...]::Get(int32, int32, int32) instance void int32[0...,0...,0...]::Set(int32, int32, int32, int32)
pp1 = &array1D[8]; pp1 = &array3D[7,6,5]; stloc.0 ldloc.1 ldc.i4.8 ldelema stloc.2 ldloc.0 ldc.i4.7 ldc.i4.6 ldc.i4.5 call
[mscorlib]System.Int32
instance int32& int32[0...,0...,0...]::Address(int32, int32, int32)
end example]
34.12 Interfaces
An interface class shall be emitted as a class with the corresponding name and visibility. It shall be marked interface. As an interface class is a class, see 34.7 and its subordinate subclauses, and 34.9 for metadata details pertaining to classes and their members. All interface class member functions shall be emitted as .methods marked as newslot, abstract, and
virtual. [Example: public interface struct I { void F(); property int P { int get(); void set(int value); } }; .class interface public abstract I { .method public newslot abstract virtual instance void F() { } .property instance int32 P() { .get instance int32 I::get_P() .set instance void I::set_P(int32) } .method public newslot abstract virtual int32 get_P() { } .method public newslot abstract virtual void set_P(int32 value) { }
232
Metadata public interface struct I1 { void F(); }; public interface struct I2 : I1 { void G(); void K(); }; public ref struct B { virtual void K() { } }; public ref virtual virtual virtual }; public ref struct E abstract : I1 { virtual void F() abstract; }; .class interface public abstract I1 { .method public newslot abstract virtual instance void F() { } } .class interface public abstract I2 implements I1 { .method public newslot abstract virtual instance void G() { } .method public newslot abstract virtual instance void K() { } } .class public B { .method public newslot virtual instance void K() { } } .class public D extends B implements I2 { .method public virtual instance void F() { } .method public newslot virtual final instance void H() { .override I2::G } .method public newslot virtual instance void G() { } } .class public abstract E implements I1 { .method public abstract virtual instance void F() { } } struct D void F() void H() void G() : B, I2 { { } = I2::G { } new { } // // // // implements I1::F implements I2::G a new G I2::K implemented by B::K
end example]
34.13 Enums
Both native and CLI enums shall be implemented as sealed classes that derive from System::Enum. The visibility of the enum type shall be reflected in its class's definition. Each enum type's class shall contain a public instance field called value__ whose type shall be that of the enum's underlying type, which shall be a CLS-compliant integer type. That field shall be marked rtspecialname and specialname. (For information specific to fields, see 34.7.3.) Each enumerator in a CLI enum shall have a corresponding public static literal field of the same name, whose type is that of the parent enum type, and whose value is as defined in the enum-specifier. [Note Enumerators in native enums have no such corresponding fields. As a result, to share their values across separate compilations, a header must be used. end note] [Example:
public enum Suit : short { Hearts = 1, Spades, Clubs, Diamonds}; enum class Direction { North, South = 10, East, West = 20 };
233
C++/CLI Language Specification .class public sealed Suit extends [mscorlib]System.Enum { .field public specialname rtspecialname int16 value__ } .class private sealed Direction extends [mscorlib]System.Enum { .field public static literal valuetype Direction East = int32(0x0B) .field public static literal valuetype Direction North = int32(0x00) .field public static literal valuetype Direction South = int32(0x0A) .field public static literal valuetype Direction West = int32(0x14) .field public specialname rtspecialname int32 value__ }
end example]
34.14 Delegates
A delegate shall be implemented as a sealed class that (ultimately) derives from System::Delegate. [Note: A delegate class need not derive directly from this class, however. A conforming implementation of the CLI is permitted to extend the required type hierarchy by including intermediate types. For example, a conforming implementation of the CLI could provide a type System::MulticastDelegate, which, in turn, is derived from System::Delegate. As such, a conforming C++/CLI implementation could derive its delegate classes from System::MulticastDelegate, or from a class derived from that class. end note] The visibility of the delegate type shall be reflected in its class's definition. For each delegate type class, a conforming implementation shall provide a constructor, a method called Invoke, and the methods BeginInvoke and EndInvoke (used for asynchronous processing), as defined by the CLI standard. [Example:
public delegate Object^ D(int* pi, array<int>^ a); .class public sealed D extends [mscorlib]System.Delegate { .method public specialname rtspecialname instance void .ctor(object A_1, native int A_2) runtime managed forwardref {} .method public newslot virtual instance class [mscorlib]System.IAsyncResult BeginInvoke(int32* pi, int32[] a, class [mscorlib]System.AsyncCallback callback, object obj) runtime managed forwardref {} .method public newslot virtual instance object EndInvoke(class [mscorlib]System.IAsyncResult result) runtime managed forwardref {} .method public newslot virtual instance object Invoke(int32* pi, int32[] a) runtime managed forwardref {} }
end example] In 27.2, it states "Each delegate type shall have two constructors, as follows: " The library class System::Delegate has no constructors defined. Instead, as we can see from the metadata example above, one, and only one, constructor is generated for a delegate, and its implementation attributes are runtime managed instead of cil managed. This is because the constructor is generated at runtime by the VES. Although the C++/CLI syntax supports delegate constructor calls having either one or two arguments, both forms shall be converted to a call to the one constructor that actually exists in metadata. The C++/CLI constructor taking one argument shall be emitted as a call to the two-argument version with nullptr as the first argument. [Example:
delegate void D(int i);
234
Metadata ref struct R { static void M1(int a) { } void M2(int b) { } virtual void M3(int c) { } }; int main() { R^ r = gcnew R; D^ d; d = gcnew D(&R::M1); d = gcnew D(r, &R::M2); d += gcnew D(r, &R::M3); } .method main() { .locals ([0] class D V_0, [1] class R V_1) ldnull stloc.1 ldnull stloc.0 newobj stloc.1 ldnull ldftn newobj stloc.0 ldloc.1 ldftn newobj stloc.0 ldloc.0 ldloc.1 dup ldvirtftn newobj call
instance void R::.ctor()
void R::M1(int32) instance void D::.ctor(object, native int)
instance void R::M2(int32) instance void D::.ctor(object, native int)
instance void R::M3(int32) instance void D::.ctor(object, native int) class [mscorlib]System.Delegate [mscorlib]System.Delegate::Combine( class [mscorlib]System.Delegate, class [mscorlib]System.Delegate) D
castclass stloc.0 }
end example]
34.15 Exceptions
try, catch, and finally shall be emitted using one or more .try directives. [Example: int main() { try { // ... } catch (NullReferenceException^ ex1) { // ... } catch (IndexOutOfRangeException^ ex2) { // ... }
235
C++/CLI Language Specification finally { // ... } } .method main() { .locals ([0] class [mscorlib]System.IndexOutOfRangeException ex2, [1] class [mscorlib]System.NullReferenceException ex1) .try { .try { leave.s }
L8
catch [mscorlib]System.NullReferenceException { stloc.1 leave.s Le } catch [mscorlib]System.IndexOutOfRangeException { stloc.0 leave.s La } L8: La: Lc: Ie: I10: } br.s leave.s br.s leave.s leave.s Lc L13 L10 L13 L13
finally { endfinally } L13: }
end example] The metadata encoding for exception-declarations that declare non-ref class types, or have the form ..., is unspecified.
34.16 Attributes
If it is not required to be consumed by the compiler, an attribute on a program element shall be emitted into metadata via a .custom directive on that element, or, in some cases, to the immediately preceding element declaration. If a program element has multiple attributes, and multiple attributes are permitted, that element shall have one .custom directive for each; their ordering is irrelevant. A custom attribute is declared using the directive .custom, followed by the method declaration for a type constructor (i.e., that method's name shall be .ctor), optionally followed by an equals sign (=) and a set of byte values in parentheses. The values of the constructor's arguments, if any, shall be specified in the set of bytes in the format specified by the CLI Standard. If there are no arguments, the equals sign and parenthesized set of bytes shall be omitted. As a constructor is an instance method, its .custom directive shall contain the instance attribute. [Example:
236
Metadata [AttributeUsage(AttributeTargets::All, AllowMultiple = true, Inherited = true)] public ref class XAttribute : Attribute { String^ name; public: XAttribute(String^ name) : name(name) {} property String^ Name { String^ get() { return name;} } }; .class public XAttribute extends [mscorlib]System.Attribute { .custom instance void [mscorlib]System.AttributeUsageAttribute::.ctor(valuetype [mscorlib]System.AttributeTargets) = ( 01 00 FF 7F 00 00 02 00 54 02 0D 41 6C 6C 6F 77 4D 75 6C 74 69 70 6C 65 01 54 02 09 49 6E 68 65 72 69 74 65 64 01) } [X("refclass")] public ref class R { [X("field")] int count; public: [X("constructor")] R() {} }; .class R { .custom instance void XAttribute::.ctor(string) = ( 01 00 08 72 65 66 63 6C 61 73 73 00 00 ) // refclass .field private int32 count .custom instance void XAttribute::.ctor(string) = ( 01 00 05 66 69 65 6C 64 00 00 ) // field .method public specialname rtspecialname instance void .ctor() cil { .custom instance void XAttribute::.ctor(string) = ( 01 00 0B 63 6F 6E 73 74 72 75 63 74 6F 72 00 00 ) // constructor } } [X("valueclass")] public value struct V { [X("method1"),X("method2")] [returnvalue:X("returnvalue")] void Display([X("parameter")] int i) {} }; .class V { .custom instance void XAttribute::.ctor(string) = ( 01 00 0A 76 61 6C 75 65 63 6C 61 73 73 00 00 ) // valueclass .method void Display(int32 i) { .custom instance void XAttribute::.ctor(string) = ( 01 00 07 6D 65 74 68 6F 64 32 00 00 ) // method2 .custom instance void XAttribute::.ctor(string) = ( 01 00 07 6D 65 74 68 6F 64 31 00 00 ) // method1 .param [0] .custom instance void XAttribute::.ctor(string) = ( 01 00 0B 72 65 74 75 72 6E 76 61 6C 75 65 00 00 ) // returnvalue .param [1] .custom instance void XAttribute::.ctor(string) = ( 01 00 09 70 61 72 61 6D 65 74 65 72 00 00 ) // parameter } } .param [0] represents the function's return value, while the actual parameter attributes start with .param [1].
237
C++/CLI Language Specification [X("interfaceclass")] public interface class I { [X("property")]property int Count { [X("getter")]int get(); } }; .class interface I { .custom instance void XAttribute::.ctor(string) = ( 01 00 0E 69 6E 74 65 72 66 61 63 65 63 6C 61 73 73 00 00 ) // interfaceclass .property instance int32 Count() { .custom instance void XAttribute::.ctor(string) = ( 01 00 08 70 72 6F 70 65 72 74 79 00 00 ) // property .get instance int32 I::get_Count() } .method public get_Count() { .custom instance void XAttribute::.ctor(string) = ( 01 00 06 67 65 74 74 65 72 00 00 ) // getter } } [X("nativeclass")] public class N { [X("field")] int count; public: [X("constructor")] N() { } [X("method")][returnvalue:X("returnvalue")] void Display([X("parameter")] int) {} }; .class N { .custom instance void XAttribute::.ctor(string) = ( 01 00 0B 6E 61 74 69 76 65 63 6C 61 73 73 00 00 ) // nativeclass }
As member information for a native class need not be emitted in metadata, only the .custom directive for the class itself need be present. end example] Since attributes can be used to customize metadata, they are often referred to as custom attributes. There are two kinds of custom attributes: genuine custom attributes and pseudo-custom attributes. Custom attributes and pseudo-custom attributes are treated differently, at the time they are defined, as follows: A custom attribute is stored directly into the metadata; the blob which holds its defining data is stored as-is. That blob can be retrieved later. A pseudo-custom attribute is recognized because its name is one of a short list. Rather than store its blob directly in metadata, that blob is parsed, and the information it contains is used to set bits and/or fields within metadata tables. The blob is then discarded; it cannot be retrieved later.
Pseudo-custom attributes therefore serve to capture user directives, using the same familiar syntax the compiler provides for genuine custom attributes, but these user directives are then stored into the more space-efficient form of metadata tables. Tables are also faster to check at runtime than are genuine custom attributes. Many custom attributes are invented by higher layers of software. They are stored and returned by the CLI, without its knowing or caring what they mean. However, all pseudo-custom attributes, plus a collection of genuine custom attributes, are of special interest to compilers and to the CLI. The CLI Standard, Partition II, subclause 21 lists the pseudo-custom attributes and distinguished custom attributes, where distinguished means that the CLI and/or compilers need to pay direct attention to them, and their behavior is affected in some way. The special processing needed for various pseudo-custom attributes is described elsewhere in this clause. Examples include DllImportAttribute, FieldOffsetAttribute, InAttribute, MarshalAsAttribute, MethodImplAttribute, OutAttribute, and StructLayoutAttribute. 238
Metadata
A conforming implementation needs to be aware of the attribute AttributeUsageAttribute (from namespace System). The parameter array ellipses notation (...) involves the generation of a .custom directive for the attribute ParamArrayAttribute, (in namespace System). See 34.6.2.
34.17 Templates
The metadata encoding for template classes and functions is unspecified except that the name of any template class emitted shall not be spelled in a CLS-compliant manner.
34.18 Generics
The name of a generic type shall be that type's name as specified in the C++/CLI source, plus a suffix of the form `n, where n is a decimal integer constant (without leading zeros) representing the arity of that type. The name in metadata of a non-generic type shall not have such a suffix. [Example:
ref class X { }; // metadata type name is X .class public X { } generic<typename T> public ref class X { }; // metadata type name is X`1 .class public X`1< T> { } generic<typename T, typename U> public ref class X { public: ref class Y { }; generic<typename A> ref class Z { }; }; // metadata type name is X`2 .class public X`2< T, U> { // metadata type name is Y .class nested public Y<( T, U> { } // metadata type name is Z`1 .class nested public Z`1<( } T, U, A> { }
end example]
239
Annex A. Grammar
A.1 Keywords
typedef-name: identifier namespace-name: original-namespace-name namespace-alias original-namespace-name: identifier namespace-alias: identifier class-name: identifier template-id enum-name: identifier template-name: identifier property-or-event-name: identifier
default
A.2 Lexical conventions

hex-quad: hexadecimal-digit hexadecimal-digit hexadecimal-digit hexadecimal-digit
universal-character-name: \u hex-quad \U hex-quad hex-quad preprocessing-token: header-name identifier pp-number character-literal string-literal preprocessing-op-or-punc each non-white-space character that cannot be one of the above token identifier keyword literal operator punctuator
240
Grammar
header-name: <h-char-sequence> "q-char-sequence" h-char-sequence: h-char h-char-sequence h-char h-char: any member of the source character set except new-line and > q-char-sequence q-char q-char-sequence q-char q-char: any member of the source character set except new-line and " pp-number: digit . digit pp-number pp-number pp-number pp-number pp-number
digit nondigit e sign E sign

.
identifier: nondigit identifier nondigit identifier digit nondigit: one of universal-character-name

_ a n A N 1 b o B O 2 c p C P 3 d q D Q 4 e r E R 5 f s F S 6 g t G T 7 h u H U 8 i v I V 9 ## %:%: .* ^ -= >>= -not ( ; & *= <<= , not_eq ) : | /= == ->* j w J W k x K X l y L Y m z M Z
digit: one of
0 { <: new + ! ^= <= and or
preprocessing-op-or-punc: one of
} :> delete = &= >= and_eq or_eq [ <% ? * < |= && bitand xor ] %> :: / > << || bitor xor_eq # %: . % += >> ++ compl ... ~ %= != ->
literal: integer-literal character-literal floating-literal string-literal boolean-literal null-literal
241
integer-literal: decimal-literal integer-suffixopt octal-literal integer-suffixopt hexadecimal-literal integer-suffixopt decimal-literal: nonzero-digit decimal-literal digit octal-literal:
0
octal-literal octal-digit hexadecimal-literal: 0x hexadecimal-digit 0X hexadecimal-digit hexadecimal-literal hexadecimal-digit nonzero-digit: one of

1 0 0 a A 2 1 1 b B 3 2 2 c C 4 3 3 d D 5 4 4 e E 6 5 5 f F 7 6 6 8 7 7 8 9 9
octal-digit: one of hexadecimal-digit: one of
integer-suffix: unsigned-suffix long-suffixopt unsigned-suffix long-long-suffixopt long-suffix unsigned-suffixopt long-long-suffix unsigned-suffixopt unsigned-suffix: one of
u l ll U L LL
long-suffix: one of long-long suffix: one of character-literal: 'c-char-sequence' L'c-char-sequence' c-char-sequence: c-char c-char-sequence c-char c-char: any member of the source character set except the single-quote ', backslash \, or new-line character escape-sequence universal-character-name escape-sequence: simple-escape-sequence octal-escape-sequence hexadecimal-escape-sequence
242
Grammar
simple-escape-sequence: one of
\ \a \" \b \? \f \\ \n \r \t \v
octal-escape-sequence: \ octal-digit \ octal-digit octal-digit \ octal-digit octal-digit octal-digit hexadecimal-escape-sequence: \x hexadecimal-digit hexadecimal-escape-sequence hexadecimal-digit floating-literal: fractional-constant exponent-partopt floating-suffixopt digit-sequence exponent-part floating-suffixopt fractional-constant: digit-sequenceopt . digit-sequence digit-sequence . exponent-part: e signopt digit-sequence E signopt digit-sequence sign: one of
+ -
digit-sequence: digit digit-sequence digit floating-suffix: one of

f l F L
string-literal: "s-char-sequenceopt" L"s-char-sequenceopt" s-char-sequence: s-char s-char-sequence s-char s-char: any member of the source character set except the double-quote ", backslash \, or new-line character escape-sequence universal-character-name boolean-literal:
false true
null-literal:
nullptr
A.3 Basic concepts

translation-unit: declaration-seqopt
243
A.4 Expressions
primary-expression: literal
this ( expression )
id-expression id-expression: unqualified-id qualified-id unqualified-id: identifier operator-function-id conversion-function-id ~ class-name ! class-name template-id generic-id
default
qualified-id: ::opt nested-name-specifier templateopt unqualified-id :: identifier :: operator-function-id :: template-id nested-name-specifier: class-or-namespace-name :: nested-name-specifieropt class-or-namespace-name :: template nested-name-specifier class-or-namespace-name: class-name namespace-name property-or-event-name postfix-expression: primary-expression postfix-expression [ expression-list ] postfix-expression ( expression-listopt ) simple-type-specifier ( expression-listopt ) typename ::opt nested-name-specifier identifier ( expression-listopt ) typename ::opt nested-name-specifier templateopt template-id ( expression-listopt ) postfix-expression . templateopt id-expression postfix-expression -> templateopt id-expression postfix-expression . pseudo-destructor-name postfix-expression -> pseudo-destructor-name postfix-expression ++ postfix-expression -dynamic_cast < type-id > ( expression ) static_cast < type-id > ( expression ) reinterpret_cast < type-id > ( expression ) const_cast < type-id > ( expression ) typeid ( expression ) typeid ( type-id ) typenameopt ::opt nested-name-specifier identifier :: typeid typenameopt ::opt nested-name-specifier templateopt template-id :: typeid
244
Grammar
expression-list: assignment-expression expression-list , assignment-expression pseudo-destructor-name: ::opt nested-name-specifieropt type-name :: ~ type-name ::opt nested-name-specifier template template-id :: ~ type-name ::opt nested-name-specifieropt ~ type-name unary-expression: postfix-expression ++ cast-expression -- cast-expression unary-operator cast-expression sizeof unary-expression sizeof ( type-id ) new-expression delete-expression unary-operator: one of
* & + ! ~
new-expression:
::opt new new-placementopt new-type-id new-initializeropt ::opt new new-placementopt ( type-id ) new-initializeropt gcnew type-specifier-seq new-initializeropt array-initopt
new-placement: ( expression-list ) new-type-id: type-specifier-seq new-declaratoropt new-declarator: ptr-operator new-declaratoropt direct-new-declarator direct-new-declarator: [ expression ] direct-new-declarator [ constant-expression ] new-initializer: ( expression-listopt ) array-init: { initializer-list ,opt }
{ }
delete-expression:
::opt delete cast-expression ::opt delete [ ] cast-expression
cast-expression: unary-expression ( type-id ) cast-expression pm-expression: cast-expression pm-expression .* cast-expression pm-expression ->* cast-expression
245
multiplicative-expression: pm-expression multiplicative-expression * pm-expression multiplicative-expression / pm-expression multiplicative-expression % pm-expression additive-expression: multiplicative-expression additive-expression + multiplicative-expression additive-expression - multiplicative-expression shift-expression: additive-expression shift-expression << additive-expression shift-expression >> additive-expression relational-expression: shift-expression relational-expression relational-expression relational-expression relational-expression
< shift-expression > shift-expression <= shift-expression >= shift-expression
equality-expression: relational-expression equality-expression == relational-expression equality-expression != relational-expression and-expression: equality-expression and-expression & equality-expression exclusive-or-expression: and-expression exclusive-or-expression ^ and-expression inclusive-or-expression: exclusive-or-expression inclusive-or-expression | exclusive-or-expression logical-and-expression: inclusive-or-expression logical-and-expression && inclusive-or-expression logical-or-expression: logical-and-expression logical-or-expression || logical-and-expression conditional-expression: logical-or-expression logical-or-expression ? expression : assignment-expression assignment-expression: conditional-expression logical-or-expression assignment-operator assignment-expression throw-expression assignment-operator: one of
= *= /= %= += -= >>= <<= &= ^= |=
246
Grammar
expression: assignment-expression expression , assignment-expression constant-expression: conditional-expression
A.5 Statements
statement: labeled-statement expression-statement compound-statement selection-statement iteration-statement jump-statement declaration-statement try-block labeled-statement: identifier : statement case constant-expression : statement default : statement expression-statement: expressionopt ; compound-statement: { statement-seqopt } statement-seq: statement statement-seq statement selection-statement: if ( condition ) statement if ( condition ) statement else statement switch ( condition ) statement condition: expression type-specifier-seq declarator = assignment-expression iteration-statement: while ( condition ) statement do statement while ( expression ) ; for ( for-init-statement conditionopt ; expressionopt ) statement foreach ( type-specifier-seq declarator in assignment-expression ) statement for-init-statement: expression-statement simple-declaration jump-statement:
break ; continue ; return expressionopt ; goto identifier ;
declaration-statement: block-declaration
247
A.6 Declarations
declaration-seq: declaration declaration-seq declaration declaration: block-declaration function-definition template-declaration generic-declaration explicit-instantiation explicit-specialization linkage-specification namespace-definition block-declaration: simple-declaration asm-definition namespace-alias-definition using-declaration using-directive simple-declaration: attributesopt decl-specifier-seqopt init-declarator-listopt ; decl-specifier: storage-class-specifier type-specifier function-specifier
friend typedef
decl-specifier-seq: decl-specifier-seqopt decl-specifier storage-class-specifier:

auto register static extern mutable
function-specifier:
inline virtual explicit
typedef-name: identifier type-specifier: simple-type-specifier class-specifier enum-specifier elaborated-type-specifier cv-qualifier delegate-specifier
248
Grammar
simple-type-specifier: ::opt nested-name-specifieropt type-name ::opt nested-name-specifier template template-id

char wchar_t bool short int long signed unsigned float double void
type-name: class-name enum-name typedef-name elaborated-type-specifier: attributesopt class-key ::opt nested-name-specifieropt identifier attributesopt class-key ::opt nested-name-specifieropt templateopt template-id attributesopt enum-key ::opt nested-name-specifieropt identifier attributesopt typename ::opt nested-name-specifieropt identifier attributesopt typename ::opt nested-name-specifier templateopt template-id enum-name: identifier enum-specifier: attributesopt top-level-visibilityopt enum-key identifieropt enum-baseopt { enumerator-listopt } enum-key:
enum enumclass enumstruct
enum-base: : type-specifier-seq enumerator-list: enumerator-definition enumerator-list , enumerator-definition enumerator-definition: enumerator enumerator = constant-expression enumerator: attributesopt identifier namespace-name: original-namespace-name namespace-alias original-namespace-name: identifier
249
namespace-definition: named-namespace-definition unnamed-namespace-definition named-namespace-definition: original-namespace-definition extension-namespace-definition original-namespace-definition: namespace identifier { namespace-body } extension-namespace-definition: namespace original-namespace-name { namespace-body } unnamed-namespace-definition: namespace { namespace-body } namespace-body: declaration-seqopt namespace-alias: identifier namespace-alias-definition: namespace identifier = qualified-namespace-specifier ; qualified-namespace-specifier: ::opt nested-name-specifieropt namespace-name using-declaration:
using typenameopt ::opt nested-name-specifier unqualified-id ; using :: unqualified-id ;
using-directive: using namespace ::opt nested-name-specifieropt namespace-name ; asm-definition: asm ( string-literal ) ; linkage-specification: extern string-literal { declaration-seqopt } extern string-literal declaration
A.7 Declarators
init-declarator-list: init-declarator init-declarator-list , init-declarator init-declarator: declarator initializeropt declarator: direct-declarator ptr-operator declarator direct-declarator: declarator-id direct-declarator ( parameter-declaration-clause ) cv-qualifier-seqopt exception-specificationopt direct-declarator [ constant-expressionopt ] ( declarator )
250
Grammar
ptr-operator: * cv-qualifier-seqopt ^ cv-qualifier-seqopt

& % ::opt nested-name-specifier * cv-qualifier-seqopt
cv-qualifier-seq: cv-qualifier cv-qualifier-seqopt cv-qualifier:

const volatile
declarator-id: id-expression ::opt nested-name-specifieropt type-name type-id: type-specifier-seq abstract-declaratoropt type-specifier-seq: type-specifier type-specifier-seqopt abstract-declarator: ptr-operator abstract-declaratoropt direct-abstract-declarator direct-abstract-declarator: direct-abstract-declaratoropt ( parameter-declaration-clause ) cv-qualifier-seqopt exception-specificationopt direct-abstract-declaratoropt [ constant-expressionopt ] ( abstract-declarator ) parameter-declaration-clause: parameter-declaration-listopt ...opt parameter-declaration-list , ... parameter-array parameter-declaration-list , parameter-array parameter-declaration-list: parameter-declaration parameter-declaration-list , parameter-declaration parameter-declaration: attributesopt decl-specifier-seq attributesopt decl-specifier-seq attributesopt decl-specifier-seq attributesopt decl-specifier-seq declarator declarator = assignment-expression abstract-declaratoropt abstract-declaratoropt = assignment-expression
parameter-array: attributesopt ... parameter-declaration function-definition: attributesopt decl-specifier-seqopt declarator function-modifiersopt override-specifierop ctor-initializeropt function-body attributesopt decl-specifier-seqopt declarator function-modifiersopt override-specifieropt function-try-block function-body: compound-statement
251
initializer: = initializer-clause ( expression-list ) initializer-clause: assignment-expression { initializer-list ,opt }

{ }
initializer-list: initializer-clause initializer-list , initializer-clause
A.8 Classes
class-name: identifier template-id class-specifier: attributesopt top-level-visibilityopt class-head { member-specificationopt } top-level-visibility:
public private
class-head: class-key identifieropt class-modifiersopt base-clauseopt class-key nested-name-specifier identifier class-modifiersopt base-clauseopt class-key nested-name-specifieropt template-id class-modifiersopt base-clauseopt class-key:
class struct union refclass refstruct valueclass valuestruct interfaceclass interfacestruct
class-modifiers: class-modifiersopt class-modifier class-modifier:

abstract sealed
member-specification: member-declaration member-specificationopt access-specifier : member-specificationopt
252
Grammar
member-declaration: attributesopt initonly-or-literalopt decl-specifier-seqopt member-declarator-listopt ; function-definition ;opt ::opt nested-name-specifier templateopt unqualified-id ; using-declaration template-declaration generic-declaration delegate-specifier event-definition property-definition initonly-or-literal:
initonly literal
member-declarator-list: member-declarator member-declarator-list , member-declarator member-declarator: declarator function-modifiersopt override-specifieropt declarator constant-initializeropt identifieropt : constant-expression function-modifiers: function-modifiersopt function-modifier function-modifier:
abstract new override sealed
override-specifier: = overridden-name-list pure-specifier overridden-name-list: id-expression overridden-name-list , id-expression pure-specifier:

= 0
constant-initializer: = constant-expression
A.9 Properties and events

property-definition: attributesopt property-modifiersopt property type-specifier-seq declarator propertyindexesopt { accessor-specification } attributesopt property-modifiersopt property type-specifier-seq declarator ; property-modifiers: property-modifiersopt property-modifier property-modifier:
static virtual
253
property-indexes: [ property-index-parameter-list ] property-index-parameter-list: type-id property-index-parameter-list , type-id accessor-specification: accessor-declaration accessor-specificationopt access-specifier : accessor-specificationopt accessor-declaration: attributesopt decl-specifier-seqopt member-declarator-listopt ; function-definition event-definition: attributesopt event-modifiersopt event event-type identifier { accessor-specification } attributesopt event-modifiersopt event event-type identifier ; event-modifiers: event-modifiersopt event-modifier event-modifier:
static virtual
event-type: ::opt nested-name-specifieropt type-name ^opt ::opt nested-name-specifieropt template template-id ^
A.10 Derived classes

base-clause: : base-specifier-list base-specifier-list: base-specifier base-specifier-list , base-specifier base-specifier: ::opt nested-name-specifieropt class-name virtual access-specifieropt ::opt nested-name-specifieropt class-name access-specifier virtualopt ::opt nested-name-specifieropt class-name access-specifier:
private protected public internal protected public public protected private protected protected private
A.11 Special member functions

conversion-function-id: operator conversion-type-id conversion-type-id: type-specifier-seq conversion-declaratoropt
254
Grammar
conversion-declarator: ptr-operator conversion-declaratoropt ctor-initializer: : mem-initializer-list mem-initializer-list: mem-initializer mem-initializer , mem-initializer-list mem-initializer: mem-initializer-id ( expression-listopt ) mem-initializer-id: ::opt nested-name-specifieropt class-name identifier
A.12 Overloading
operator-function-id: operator operator operator operator < template-argument-listopt > operator: one of
new + ! = <= () delete * = < &= |= >= && [] new[] / > << || % += >> ++ delete[] ^ & -= *= >>= <<= -, | /= == ->* ~ %= != ->
A.13 Delegates
delegate-specifier: attributesopt top-level-visibilityopt delegate type-specifier-seq declarator ;
A.14 Templates
template-declaration:
exportopt template < template-parameter-list > declaration
template-parameter-list: template-parameter template-parameter-list , template-parameter template-parameter: type-parameter parameter-declaration type-parameter: class identifieropt class identifieropt = type-id typename identifieropt typename identifieropt = type-id template < template-parameter-list > class identifieropt template < template-parameter-list > class identifieropt = id-expression template-id: template-name < template-argument-listopt > template-name: identifier
255
template-argument-list: template-argument template-argument-list , template-argument template-argument: assignment-expression type-id id-expression explicit-instantiation: template declaration explicit-specialization:
template < > declaration
A.15 Generics
generic-declaration: generic < generic-parameter-list > constraint-clause-listopt declaration generic-parameter-list: generic-parameter generic-parameter-list , generic-parameter generic-parameter: attributesopt class identifier attributesopt typename identifier generic-id: generic-name < generic-argument-list > generic-name: identifier operator-function-id generic-argument-list: generic-argument generic-argument-list , generic-argument generic-argument: type-id constraint-clause-list: constraint-clause-listopt constraint-clause: where identifier constaint-item-list: constraint-item constraint-item-list constraint-item: type-id
refclass refstruct valueclass valuestruct gcnew ( ) :
constraint-clause constraint-item-list
constraint-item
256
Grammar
A.16 Exception handling

try-block:
try compound-statement handler-seq try compound-statement finally-clause try compound-statement handler-seq finally-clause
function-try-block: try ctor-initializeropt function-body handler-seq try ctor-initializeropt function-body finally-clause try ctor-initializeropt function-body handler-seq finally-clause handler-seq: handler handler-seqopt handler: catch ( exception-declaration ) compound-statement exception-declaration: type-specifier-seq declarator type-specifier-seq abstract-declarator type-specifier-seq
...
finally-clause:
finally compound-statement
throw-expression: throw assignment-expressionopt exception-specification: throw ( type-id-listopt ) type-id-list: type-id type-id-list , type-id
A.17 Attributes
attributes: attribute-sections attribute-sections: attribute-sectionsopt attribute-section attribute-list
]
attribute-section: [ attribute-target-specifieropt attribute-target-specifier: : attribute-target
257
attribute-target:
assembly class constructor delegate enum event field interface method parameter property returnvalue struct
attribute-list: attribute attribute-list attribute: attribute-name attribute-name: type-name
attribute attribute-argumentsopt
attribute-arguments: ( positional-argument-listopt ) positional-argument-list , named-argument-list ( ( named-argument-list ) positional-argument-list: positional-argument positional-argument-list
positional-argument
positional-argument: attribute-argument-expression named-argument-list: named-argument named-argument-list named-argument: = identifier
named-argument
attribute-argument-expression
attribute-argument-expression: assignment-expression
A.18 Preprocessing directives

preprocessing-file: groupopt group: group-part group group-part group-part: pp-tokensopt new-line if-section control-line
258
Grammar
if-section: if-group elif-groupsopt else-groupopt endif-line if-group:

# if constant-expression new-line groupopt # ifdef identifier new-line groupopt # ifndef identifier new-line groupopt
elif-groups: elif-group elif-groups elif-group elif-group:

# elif constant-expression new-line groupopt
else-group:
# else new-line groupopt
endif-line:
# endif new-line
control-line:
# # # # # # # # # include pp-tokens new-line using pp-tokens new-line define identifier replacement-list new-line define identifier lparen identifier-listopt ) replacement-list new-line undef identifier new-line line pp-tokens new-line error pp-tokensopt new-line pragma pp-tokensopt new-line
new-line
lparen: the left-parenthesis character without preceding white-space replacement-list: pp-tokensopt pp-tokens: preprocessing-token pp-tokens preprocessing-token new-line: the new-line character
259
Annex B. Verifiable code

[Note: Reserved for future use. end note]
260
Documentation comments
Annex C. Documentation comments

This annex is informative.
C.1 Introduction
Comments having a special form can be used to direct a tool to produce XML from those comments and the source code elements they precede. Such comments are single-line comments that start with exactly three slashes (///). They shall immediately precede a user-defined type (such as a class, delegate, or interface) or a member (such as a field, event, property, or function) that they annotate. Attribute sections are considered part of declarations, so documentation comments shall precede attributes applied to a type or member. Alternatively, comments (possibly multi-line) that start with a slash and exactly two asterisks may also contain XML document comments. These comments may only be applied to CLI class types and members within those types. While processing such comments, if they are applied to unsupported types, the compiler shall issue a warning. Documentation comments in a header are processed only if that header were included using the "" form of
#include.
Syntax: single-line-doc-comment: /// intput-charactersopt delimited-doc-comment: /** delimited-comment-charactersopt */ In a single-line-doc-comment, if there is a white-space character following the /// characters on each of the single-line-doc-comments adjacent to the current single-line-doc-comment, then that one white-space character is not included in the XML output. In a delimited-doc-comment, if the first non-white-space character on the second line is an asterisk and the same pattern of optional white-space characters and an asterisk character is repeated at the beginning of each of the lines within the delimited-doc-comment, then the characters of the repeated pattern are not included in the XML output. The pattern can include white-space character after, as well as before, the asterisk character. Example:
/** <remarks> Class <c>Point</c> models a point in a two-dimensional plane. </remarks> */ public ref class Point { public: /// <remarks>Method <c>Draw</c> renders the point.</remarks> void Draw() { /*...*/ } };
The text within documentation comments shall be well-formed according to the rules of XML (http://www.w3.org/TR/REC-xml). If the XML is ill-formed, a warning is generated and the documentation file will contain a comment saying that an error was encountered. Although developers are free to create their own set of tags, a recommended set is defined in C.2. Some of the recommended tags have special meanings:
261
The <param> tag is used to describe parameters. If such a tag is used, the documentation generator shall verify that the specified parameter exists and that all parameters are described in documentation comments. If such verification fails, the documentation generator issues a warning. The cref attribute can be attached to any tag to provide a reference to a code element. The documentation generator shall verify that this code element exists. If the verification fails, the documentation generator issues a warning. When looking for a name described in a cref attribute, the documentation generator shall respect namespace visibility according to using statements appearing within the source code. The <summary> tag is intended to be used by a documentation viewer to display additional information about a type or member.
Note carefully that the documentation file does not provide full information about the type and members (for example, it does not contain any type information). To get such information about a type or member, the documentation file shall be used in conjunction with reflection on the actual type or member.
C.2 Recommended tags

The documentation generator shall accept and process any tag that is valid according to the rules of XML. The following tags provide commonly used functionality in user documentation. (Of course, other tags are possible.)
Tag
<c>
Section C.2.1 <code> C.2.2 <example> C.2.3 <exception> C.2.4 <list> C.2.5 <para> C.2.6 <param> C.2.7 <paramref> C.2.8 <permission> C.2.9 <remarks> C.2.10 <returns> C.2.11 <see> C.2.12 <seealso> C.2.13 <summary> C.2.14 <typeparam> C.2.15 <typeparamref> C.2.16 <value> C.2.17 C.2.1 <c>
Purpose Set text in a code-like font Set one or more lines of source code or program output Indicate an example Identifies the exceptions a function can throw Create a list or table Permit structure to be added to text Describe a parameter for a function or constructor Identify that a word is a parameter name Document the security accessibility of a member Describe a type Describe the return value of a function Specify a link Generate a See Also entry Describe a member of a type Describe a generic type parameter Identify that a word is a type parameter name Describe a property
This tag provides a mechanism to indicate that a fragment of text within a description should be set in a special font such as that used for a block of code. For lines of actual code, use <code> (C.2.2). Syntax:
<c>text to be set like code</c>
Example:
262
Documentation comments /// <remarks> /// Class <c>Point</c> models a point in a two-dimensional plane. /// </remarks> ref class Point { // ... };
C.2.2 <code> This tag is used to set one or more lines of source code or program output in some special font. For small code fragments in narrative, use <c> (C.2.1). Syntax:
<code>source code or program output</code>
Example:
/// <summary> /// Changes the Point's location by the given x- and y-offsets. /// <example> /// The following code: /// <code> /// Point p(3,5); /// p.Translate(-1,3); /// </code> /// results in <c>p</c>'s having the value (2,8). /// </example> /// </summary> void Translate(int xord, int yord) { X += xord; Y += yord; }
C.2.3 <example> This tag allows example code within a comment, to specify how a function or other library member may be used. Ordinarily, this would also involve use of the tag <code> (C.2.2) as well. Syntax:
<example>description</example>
Example: See <code> (C.2.2) for an example. C.2.4 <exception> This tag provides a way to document the exceptions a function can throw. Syntax:
<exception cref="member">description</exception>
where
cref="member"
The name of a member. The documentation generator checks that the given member exists and translates member to the canonical element name in the documentation file.
description
A description of the circumstances in which the exception is thrown. Example:
263
C++/CLI Language Specification public ref class DataBaseOperations { /// <exception cref="MasterFileFormatCorruptException"></exception> /// <exception cref="MasterFileLockedOpenException"></exception> static void ReadRecord(int flag) { if (flag == 1) throw new MasterFileFormatCorruptException(); else if (flag == 2) throw new MasterFileLockedOpenException(); // } };
C.2.5 <list> This tag is used to create a list or table of items. It may contain a <listheader> block to define the heading row of either a table or definition list. (When defining a table, only an entry for term in the heading need be supplied.) Each item in the list is specified with an <item> block. When creating a definition list, both term and description shall be specified. However, for a table, bulleted list, or numbered list, only description need be specified. Syntax:
<list type="bullet" | "number" | "table"> <listheader> <term>term</term> <description>description</description> </listheader> <item> <term>term</term> <description>description</description> </item> <item> <term>term</term> <description>description</description> </item> </list>
where
term
The term to define, whose definition is in description.

description
Either an item in a bullet or numbered list, or the definition of a term. Example:

public ref class MyClass { public: /// <remarks> /// Here is an example of a bulleted list: /// <list type="bullet"> /// <item> /// <description>First item.</description> /// </item> /// <item> /// <description>Second item.</description> /// </item> /// </list> /// </remarks> static void F() { // ... } };
264
C.2.6 <para> This tag is for use inside other tags, such as <remarks> (C.2.10) or <returns> (C.2.11), and permits structure to be added to text. Syntax:
<para>content</para>
where
content
The text of the paragraph. Example:

/// <summary> /// <para> /// This is the entry point of the Point class testing program. /// </para> /// <para> /// This program tests each function and operator, and is intended /// to be run after any non-trivial maintenance has been performed /// on the Point class. /// </para> /// </summary> int main() { // ... }
C.2.7 <param> This tag is used to describe a parameter for a function, constructor, or indexer. Syntax:
<param name="name">description</param>
where
name
The name of the parameter.

description
A description of the parameter. Example:

/// <summary> /// This function changes the point's location to the given coordinates. /// </summary> /// <param name="xord"><c>xord</c> is the new x-coordinate.</param> /// <param name="yord"><c>yord</c> is the new y-coordinate.</param> void Move(int xord, int yord) { X = xord; Y = yord; }
C.2.8 <paramref> This tag is used to indicate that a word is a parameter. The documentation file can be processed to format this parameter in some distinct way. Syntax:
<paramref name="name"/>
where
265
C++/CLI Language Specification name
The name of the parameter. Example:

/// <summary> /// This constructor initializes the new Point to /// (<paramref name="xord"/>,<paramref name="yord"/>). /// </summary> /// <param name="xord"> /// <c>xord</c> is the new Point's x-coordinate. /// </param> /// <param name="yord"> /// <c>yord</c> is the new Point's y-coordinate. /// </param> Point(int xord, int yord) { X = xord; Y = yord; }
C.2.9 <permission> This tag allows the security accessibility of a member to be documented. Syntax:
<permission cref="member">description</permission>
where
cref="member"
The name of a member. The documentation generator checks that the given code element exists and translates member to the canonical element name in the documentation file.
description
A description of the access to the member. Example:

/// <permission cref="System::Security::PermissionSet"> /// Everyone can access this function. /// </permission> static void Test() { // ... }
C.2.10 <remarks> This tag is used to specify overview information about a type. (Use <summary> (C.2.14) to describe the members of a type.) Syntax:
<remarks>description</remarks>
where
description
The text of the remarks. Example:

/// <remarks> /// Class <c>Point</c> models a point in a two-dimensional plane. /// </remarks> public ref class Point { // ... };
266
C.2.11 <returns> This tag is used to describe the return value of a function. Syntax:
<returns>description</returns>
where
description
A description of the return value. Example:

/// <summary> /// Report a point's location as a string. /// </summary> /// <returns> /// A string representing a point's location, in the form (x,y), /// without any leading, trailing, or embedded whitespace. /// </returns> String^ ToString() override { return String::Format("({0},{1})", X, Y); }
C.2.12 <see> This tag allows a link to be specified within text. Use <seealso> (C.2.13) to indicate text that is to appear in a See Also subclause. Syntax:
<see cref="member"/>
where
cref="member"
The name of a member. The documentation generator checks that the given code element exists and changes member to the element name in the generated documentation file. Example:
/// <summary> /// This function changes the point's location to the given coordinates. /// Use the <see cref="Translate"/> function to apply a relative change. /// </summary> void Move(int xord, int yord) { X = xord; Y = yord; } /// <summary> /// This function changes the point's location by the given offsets. /// Use the <see cref="Move"/> function to directly set the coordinates. /// </summary> void Translate(int xord, int yord) { X += xord; Y += yord; }
C.2.13 <seealso> This tag allows an entry to be generated for the See Also section. Use <see> (C.2.12) to specify a link from within text. Syntax:
267
C++/CLI Language Specification <seealso cref="member"/>
where
cref="member"
The name of a member. The documentation generator checks that the given code element exists and changes member to the element name in the generated documentation file. Example:
/// <summary> /// This function determines whether two Points have the same location. /// </summary> /// <seealso cref="operator=="/> /// <seealso cref="operator!="/> bool Equals(Object^ o) override { // ... }
C.2.14 <summary> This tag can be used to describe a member for a type. Use <remarks> (C.2.10) to describe the type itself. Syntax:
<summary>description</summary>
where
description
A summary of the member. Example:

/// <summary> /// This constructor initializes the new Point to (0,0). /// </summary> Point() { // }
C.2.15 <typeparam> This tag is used to describe a type parameter for a generic type or function. Syntax:
<typeparam name="name">description</typeparam>
where
name
The name of the type parameter.

description
A description of the type parameter. Example:

/// <summary> /// A single linked list that stores unique elements. /// </summary> /// <typeparam name="T">Each element of the list is a <c>T</c>.</typeparam> generic<typename T> ref class List { /* ... */ };
268
C.2.16 <typeparamref> This tag is used to indicate that a word is a type parameter. The documentation file can be processed to format this parameter in some distinct way. Syntax:
<typeparamref name="name"/>
where
name
The name of the parameter. C.2.17 <value> This tag allows a property to be described. Syntax:
<value>property description</value>
where
property description
A description for the property. Example:

/// <value> /// The point's x-coordinate. /// </value> property int X { int get() { return x; } void set(int value) { x = value; } }
C.3 Processing the documentation file

The following information is intended for C++/CLI implementations targeting the CLI. The documentation generator generates an ID string for each element in the source code that is tagged with a documentation comment. This ID string uniquely identifies a source element. A documentation viewer can use an ID string to identify the corresponding metadata/reflection item to which the documentation applies. The documentation file is not a hierarchical representation of the source code; rather, it is a flat list with a generated ID string for each element. C.3.1 ID string format The documentation generator observes the following rules when it generates the ID strings: No white space is placed in the string. The first part of the string identifies the kind of member being documented, via a single character followed by a colon. The following kinds of members are defined: Description Event Field Method (including constructors, destructors, finalizers, functions, and operators) Namespace Property (including indexers)
Character E F M N P
269
D T !
Typedef Type (such as class, delegate, enum, interface, and struct) Error string; the rest of the string provides information about the error. For example, the documentation generator generates error information for links that cannot be resolved.
The second part of the string is the fully qualified name of the element, starting at the root of the namespace. The name of the element, its enclosing type(s), and namespace are separated by periods. If the name of the item itself has periods, they are replaced by NUMBER SIGN # (U+0023) characters. (It is assumed that no element has this character in its name.) For functions and properties with arguments, the argument list follows, enclosed in parentheses. For those without arguments, the parentheses are omitted. The arguments are separated by commas. The encoding of each argument is the same as a CLI signature, as follows: Arguments are represented by their fully qualified name. For example, int is System.Int32, and so on. Tracking reference arguments have an @ following their type name. Arguments passed by value or via param arrays have no special notation. Arguments that are CLI arrays are represented as [ lowerbound : size , , lowerbound : size ] where the number of commas is the rank less one, and the lower bounds and size of each dimension, if known, are represented in decimal. If a lower bound or size is not specified, it is omitted. If the lower bound and size for a particular dimension are omitted, the : is omitted as well. Jagged arrays are represented by one [] per level. Arguments that have pointer types other than void are represented using a * following the type name. A void pointer is represented using a type name of System.Void.
C.3.2 ID string examples The following examples each show a fragment of C++ code, along with the ID string produced from each source element capable of having a documentation comment: Types are represented using their fully qualified name.
enum class Color { Red, Blue, Green }; namespace Acme { interface class IProcess { /*...*/ }; value class ValueType { /*...*/ }; ref class Widget : IProcess { public: ref class NestedClass { /*...*/ }; interface class IMenuItem { /*...*/ }; delegate void Del(int i); enum class Direction { North, South, East, West }; }; } "T:Color" "T:Acme.IProcess" "T:Acme.ValueType" "T:Acme.Widget" "T:Acme.Widget.NestedClass" "T:Acme.Widget.IMenuItem" "T:Acme.Widget.Del" "T:Acme.Widget.Direction"
Fields are represented by their fully qualified name.

namespace Acme { value class ValueType { private: int total; };
270
ref class Widget: IProcess { public: ref class NestedClass { private: int value; }; private: String^ message; static Color^ defaultColor; literal double PI = 3.14159; initonly double monthlyAverage; array<long>^ array1; array<Widget^,2>^ array2; int *pCount; float **ppValues; }; } "F:Acme.ValueType.total" "F:Acme.Widget.NestedClass.value" "F:Acme.Widget.message" "F:Acme.Widget.defaultColor" "F:Acme.Widget.PI" "F:Acme.Widget.monthlyAverage" "F:Acme.Widget.array1" "F:Acme.Widget.array2" "F:Acme.Widget.pCount" "F:Acme.Widget.ppValues"
Constructors.
namespace Acme { ref class Widget : IProcess { static Widget() { /*...*/ } public: Widget() { /*...*/ } Widget(String^ s) { /*...*/ } }; } "M:Acme.Widget.#cctor" "M:Acme.Widget.#ctor" "M:Acme.Widget.#ctor(System.String)"
Finalizers.
namespace Acme { ref class Widget : IProcess protected: !Widget() { /*...*/ } }; } "M:Acme.Widget.Finalize" {
Methods.
namespace Acme { value class ValueType { public: void M(int i) { /*...*/ } }; ref class Widget : IProcess { public: ref class NestedClass { public: void M(int i) { /*...*/ }
271
C++/CLI Language Specification }; static void M0() { /*...*/ } void M1(wchar_t c, float% f, ValueType% v) { /*...*/ } void M2(array<short>^ x1, array<int,2>^ x2, array<array<int>^>^ x3) { /*...*/ } void M3(array<array<int>^> x3, array<array<Widget^,3>^>^ x4) { /*...*/ } void M4(wchar_t *pc, Color **pf) { /*...*/ } void M5(void *pv, array<array<double*,2>^ > pd) { /*...*/ } void M6(int i, ... array<Object^>^ args) { /*...*/ } }; } "M:Acme.ValueType.M(System.Int32)" "M:Acme.Widget.NestedClass.M(System.Int32)" "M:Acme.Widget.M0" "M:Acme.Widget.M1(System.Char,System.Single@,Acme.ValueType@)" "M:Acme.Widget.M2(System.Int16[],System.Int32[0:,0:],System.Int64[][])" "M:Acme.Widget.M3(System.Int64[][],Acme.Widget[0:,0:,0:][])" "M:Acme.Widget.M4(System.Char*,Color**)" "M:Acme.Widget.M5(System.Void*,System.Double*[0:,0:][])" "M:Acme.Widget.M6(System.Int32,System.Object[])"
Properties and indexers.

namespace Acme { ref class Widget : IProcess { public: property int Width { int get() { /*...*/ } void set(int value) { /*...*/ } } property int default[int] { int get(int i) { /*...*/ } void set(int i, int value) { /*...*/ } } property int default[String^, int] { int get(String^ s, int i) { /*...*/ } void set(String^ s, int i, int value) { /*...*/ } }
}; }
"P:Acme.Widget.Width" "P:Acme.Widget.Item(System.Int32)" "P:Acme.Widget.Item(System.String,System.Int32)"
Events.
namespace Acme { ref class Widget : IProcess { public: event Del^ AnEvent; }; }
"E:Acme.Widget.AnEvent"
Unary operators. (The complete set of unary operator function names used is listed in Table 19-1: CLS-Compliant Unary Operators.)
namespace Acme { ref class Widget : IProcess { public: static Widget^ operator+(Widget^ x) { /*...*/ }
272
Documentation comments }; } "M:Acme.Widget.op_UnaryPlus(Acme.Widget)"
Binary operators. (The complete set of binary operator function names used is listed in Table 19-2: CLS-Compliant Binary Operators.)
namespace Acme { ref class Widget : IProcess { public: static Widget^ operator+(Widget^ x1, Widget^ x2) { /*...*/ } }; } "M:Acme.Widget.op_Addition(Acme.Widget,Acme.Widget)"
Conversion operators have a trailing ~ followed by the return type.

namespace Acme { ref class Widget : IProcess { public: static explicit operator int(Widget^ x) { /*...*/ } static operator long long(Widget^ x) { /*...*/ } }; } "M:Acme.Widget.op_Explicit(Acme.Widget)~System.Int32" "M:Acme.Widget.op_Implicit(Acme.Widget)~System.Int64"
C.4 An example
C.4.1 C++ source code The following example shows the source code of a Point class:
273
C++/CLI Language Specification namespace Graphics { /// <remarks> /// Class <c>Point</c> models a point in a two-dimensional plane. /// </remarks> public ref class Point { public: /// <value> /// The Point's x-coordinate. /// </value> property int X; /// <value> /// The Points' y-coordinate. /// </value> property int Y; /// <summary> /// This constructor initializes the new Point to (0,0). /// </summary> Point() { X = 0; Y = 0; } /// <summary> /// This constructor initializes the new Point to /// (<paramref name="xord"/>,<paramref name="yord"/>). /// </summary> /// <param name="xord"> /// <c>xord</c> is the new Point's x-coordinate. /// </param> /// <param name="yord"> /// <c>yord</c> is the new Point's y-coordinate. /// </param> Point(int xord, int yord) { X = xord; Y = yord; } /// <summary> /// This function changes the point's location to the given /// coordinates. /// </summary> /// <param name="xord"> /// <c>xord</c> is the new x-coordinate. /// </param> /// <param name="yord"> /// <c>yord</c> is the new y-coordinate. /// </param> /// <seealso cref="Translate"/> void Move(int xord, int yord) { X = xord; Y = yord; } /// /// /// /// /// /// /// /// /// /// /// /// /// /// <summary> This function changes the point's location by the given x- and y-offsets. </summary> <example> The following code: <code> Point p(3,5); p.Translate(-1,3); </code> results in <c>p</c>'s having the value (2,8). </example> <param name="xord"> <c>xord</c> is the relative x-offset.
274
Documentation comments /// </param> /// <param name="yord"> /// <c>yord</c> is the relative y-offset. /// </param> /// <seealso cref="Move"/> void Translate(int xord, int yord) { X += xord; Y += yord; } /// <summary> /// This function determines whether two Points have the same /// location. /// </summary> /// <param name="o"> /// <c>o</c> is the object to be compared to the current object. /// </param> /// <returns> /// True if the Points have the same location; otherwise, false. /// </returns> /// <seealso cref="operator =="/> /// <seealso cref="operator !="/> bool Equals(Object^ o) override { Point^ p = dynamic_cast<Point^>(o); if (!p) return false; return (X == p->X) && (Y == p->Y); } /// <summary> /// Computes the hash code for a Point. /// </summary> /// <returns> /// A hash code computed from the x and y coordinates. /// </returns> int GetHashCode() override { return X ^ Y; } /// <summary> /// Report a point's location as a string. /// </summary> /// <returns> /// A string representing a point's location, in the form (x,y), /// without any leading, training, or embedded whitespace. /// </returns> String^ ToString() override { return String::Format("({0},{1})", X, Y); } /// <summary> /// This operator determines whether two Points have the same /// location. /// </summary> /// <param name="p1">The first Point to be compared.</param> /// <param name="p2">The second Point to be compared.</param> /// <returns> /// True if the Points have the same location; otherwise, false. /// </returns> /// <seealso cref="Equals"/> /// <seealso cref="operator !="/> static bool operator==(Point^ p1, Point^ p2) { if ((Object^)p1 == nullptr || (Object^)p2 == nullptr) return false; return (p1->X == p2->X) && (p1->Y == p2->Y); } /// <summary> /// This operator determines whether two Points have the same /// location.
275
C++/CLI Language Specification /// </summary> /// <param name="p1">The first Point to be compared.</param> /// <param name="p2">The second Point to be compared.</param> /// <returns> /// True if the Points do not have the same location; /// otherwise, false. /// </returns> /// <seealso cref="Equals"/> /// <seealso cref="operator =="/> static bool operator!=(Point^ p1, Point^ p2) { return !(p1 == p2); } }; }
C.4.2 Resulting XML Here is the output produced by one documentation generator when given the source code for class Point, shown above: <?xml version="1.0"?> <doc> <assembly> Point </assembly> <members> <member name="T:Graphics.Point"> <remarks> Class <c>Point</c> models a point in a two-dimensional plane. </remarks> </member> <member name="M:Graphics.Point.get_X"> <value> The Point's x-coordinate. </value> </member> <member name="M:Graphics.Point.get_Y"> <value> The Points' y-coordinate. </value> </member> <member name="M:Graphics.Point.#ctor"> <summary> This constructor initializes the new Point to (0,0). </summary> </member> <member name="M:Graphics.Point.#ctor(System.Int32,System.Int32)"> <summary> This constructor initializes the new Point to (<paramref name="xord"/>,<paramref name="yord"/>). </summary> <param name="xord"> <c>xord</c> is the new Point's x-coordinate. </param> <param name="yord"> <c>yord</c> is the new Point's y-coordinate.
276
</param> </member> <member name="M:Graphics.Point.Move(System.Int32,System.Int32)"> <summary> This function changes the point's location to the given coordinates. </summary> <param name="xord"> <c>xord</c> is the new x-coordinate. </param> <param name="yord"> <c>yord</c> is the new y-coordinate. </param> <seealso cref="M:Graphics.Point.Translate(System.Int32,System.Int32)"/> </member> <member name="M:Graphics.Point.Translate(System.Int32,System.Int32)"> <summary> This function changes the point's location by the given x- and y-offsets. </summary> <example> The following code: <code> Point p(3,5); p.Translate(-1,3); </code> results in <c>p</c>'s having the value (2,8). </example> <param name="xord"> <c>xord</c> is the relative x-offset. </param> <param name="yord"> <c>yord</c> is the relative y-offset. </param> <seealso cref="M:Graphics.Point.Move(System.Int32,System.Int32)"/> </member> <member name="M:Graphics.Point.Equals(System.Object)"> <summary> This function determines whether two Points have the same location. </summary> <param name="o"> <c>o</c> is the object to be compared to the current object. </param> <returns> True if the Points have the same location; otherwise, false. </returns> <seealso cref="M:Graphics.Point.op_Equality(Graphics.Point,Graphics.Point)"/> <seealso cref="M:Graphics.Point.op_Inequality(Graphics.Point,Graphics.Point)"/> </member>
277
<member name="M:Graphics.Point.GetHashCode"> <summary> Computes the hash code for a Point. </summary> <returns> A hash code computed from the x and y coordinates. </returns> </member> <member name="M:Graphics.Point.ToString"> <summary> Report a point's location as a string. </summary> <returns> A string representing a point's location, in the form (x,y), without any leading, training, or embedded whitespace. </returns> </member> <member name="M:Graphics.Point.op_Equality(Graphics.Point,Graphics.Point)"> <summary> This operator determines whether two Points have the same location. </summary> <param name="p1">The first Point to be compared.</param> <param name="p2">The second Point to be compared.</param> <returns> True if the Points have the same location; otherwise, false. </returns> <seealso cref="M:Graphics.Point.Equals(System.Object)"/> <seealso cref="M:Graphics.Point.op_Inequality(Graphics.Point,Graphics.Point)"/> </member> <member name="M:Graphics.Point.op_Inequality(Graphics.Point,Graphics.Point)"> <summary> This operator determines whether two Points have the same location. </summary> <param name="p1">The first Point to be compared.</param> <param name="p2">The second Point to be compared.</param> <returns> True if the Points do not have the same location; otherwise, false. </returns> <seealso cref="M:Graphics.Point.Equals(System.Object)"/> <seealso cref="M:Graphics.Point.op_Equality(Graphics.Point,Graphics.Point)"/> </member> </members> </doc>
278
Non-normative references
Annex D. Non-normative references

ECMA-334:2005, C# Programming language.
279
Annex E. CLI naming guidelines

This annex is informative. Information on this topic can be found at the following location: http://msdn.microsoft.com/library/default.asp?url=/library/enus/cpgenref/html/cpconnetframeworkdesignguidelines.asp End of informative text
280
Future directions
Annex F. Future directions

This annex is informative. This annex contains information about features that might be considered for a future revision of this Standard.
F.1 Expressions
F.1.1 Class member access A named indexed property could be accessed like any other member of a class. [Note: As expected, an expression of the form p->NamedIndexer[index] is equivalent to (*p).NamedIndexer[index]. end note] F.1.2 Type identification Consider having a way for typeid on CLI class types produce a std::type_info. F.1.3 Pointer type portability The hardware architecture running the program determines the size of pointers. With the CLI, it is possible to use pointer types in programs that can run on multiple hardware architectures where pointer sizes are different. In order to support such programs, sizeof expressions on pointers would turn into a runtime expression instead of a compile time constant.
F.2 Statements
F.2.1 The checked and unchecked statements Statements of the form checked { } and unchecked { } could be used to control the overflowchecking context for integral-type arithmetic operations and conversions.
F.3 Classes
F.3.1 Delegating constructors Tutorial: When implementing a class, it is not unusual to have a number of constructors share some common code. For example, consider the case of the following point class:
class point { int x_; int y_; void commonCode(); public: point(); point(int x, int y); point(const point& p); };
All three constructors need to initialize the two private members, x_ and y_; they might also perform other actions, some of which they share, and some of which are unique. One approach is as follows:
point::point() : x_(0), y_(0) { commonCode(); // custom code goes here }
281
C++/CLI Language Specification point::point(int x, int y) : x_(x), y_(y) { commonCode(); } point::point(const point& p) : x_(p.x_), y_(p.y_) { commonCode(); // custom code goes here }
Certainly, the constructor with no parameters can be eliminated by adding default argument values to the constructor having two. However, that is not an entirely satisfactory approach for all classes. Specifically, it allows the two-argument constructor to be called with only the first argument, but not with only the second, which, philosophically, is asymmetric. As shown above, a common approach to implementing such a family of constructors is to place their common code in a private member function, such as commonCode, and have each of them call that function. C++/CLI could help solve this problem by providing delegating constructors. Simply stated, prior to executing its body, a delegating constructor can call one of its sibling constructors as though it were a base constructor. That is, it delegates part of the Objects initialization to another constructor, gets control back, and then optionally performs other actions as well. Using this approach, the constructors shown earlier can be re-implemented as follows:
point::point() : point(0, 0) { // custom code goes here } point::point(int x, int y) : x_(x), y_(y) { // common code goes here } point::point(const point& p) : point(p.x_, p.y_) { // custom code goes here }
Note how the ctor-initializer construct has been extended to accommodate a call to a sibling constructor, using the exact same approach as for a call to a base class constructor. The common code statements can now be part of the body of the second constructor, where they will be executed by calls to all three constructors. When the first and third constructors are called, they transfer control to the second. When that returns control to its caller, that callers body is executed. Any constructor can delegate to any of its siblings; however, a class shall have at least one non-delegating constructor (no diagnostic is required), and that constructor can still have a ctor-initializer that calls one or more base class constructors. A delegating constructor cannot also have a ctor-initializer that contains a comma-separated list of member initializers. Specification: The definition of ctor-initializer is augmented to accommodate the addition of delegating constructors to C++/CLI; however, no change is necessary in the Standard C++ (8.4) grammar. Prior to executing its body, a constructor can call one of its sibling constructors to initialize members. That is, it delegates the objects initialization to another constructor, gets control back, and then optionally performs other actions as well. A constructor that delegates in this manner is called a delegating constructor, and the constructor to which it delegates is called a target constructor. A delegating constructor can also be a target constructor of some other delegating constructor. [Example:
class FullName { string firstName_; string middleName_; string lastName_; public: FullName(string firstName, string middleName, string lastName); FullName(string firstName, string lastName); FullName(const FullName& name); };
282
Future directions FullName::FullName(string firstName, string middleName, string lastName) : firstName_(firstName), middleName_(middleName), lastName_(lastName) { } // delegating copy constructor FullName::FullName(const FullName& name) : FullName(name.firstName, name.middleName, name.lastName) { } // delegating constructor FullName::FullName(string firstName, string lastName) : FullName(firstName, "", lastName) { }
end example] If a mem-initializer-id designates the class being defined, it shall be the only mem-initializer. The resulting ctor-initializer signifies that the constructor being defined is a delegating constructor. A delegating constructor causes a constructor from the class itself to be invoked. The target constructor is selected by overload resolution and template argument deduction, as usual. If a delegating constructor definition includes a ctor-initializer that directly or indirectly invokes the constructor itself, the program is ill-formed; however, no diagnostic is required. [Example: When using constructors that are templates, deduction works as usual:
class X { template<class T> X(T, T) : l_(first, last) { /* Common Init */ } list<int> l_; public: X(vector<short>&); }; X::X(vector<short>& v) : X(v.begin(), v.end()) { } // T is deduced as vector<short>::iterator
end example] The objects lifetime begins when all construction is successfully completed. For the purposes of the C++ Standard (3.8), the constructor call has completed means the originally invoked constructor call. [Rationale: Even if a target constructor completes, an outer delegating constructor can still throw an exception, and if so the caller did not get the object that was requested. The foregoing decision also preserves the Standard C++ rule that an exception emitted from a constructor means that the objects lifetime never began. end rationale] F.3.2 Properties Allowing properties in native classes. Allowing the modifiers abstract, new, override, and sealed to be applied directly to a property as well as or instead of to one or more of its accessors. F.3.3 Events Allowing the modifiers abstract, new, override, and sealed to be applied directly to an event as well as or instead of to one or more of its accessors. F.3.4 Unsupported CLS-recommended operators
Function Name in Assembly 283
op_SignedRightShift op_UnsignedRightShift op_MemberSelection op_PointerToMemberSelection
undefined undefined undefined undefined
Regarding op_MemberSelection and op_PointerToMemberSelection, the C++ Standard only permits non-static member declarations of these operators. F.3.5 Operators true and false Add the ability to define operator true and operator false.
F.4 Generic types

Although the CLI permits the retrieval of a System::Type object that is associated with an open constructed generic type (31.2.1), C++/CLI provides no syntax for doing this. However, such syntax might be considered in future.
F.5 Custom modifiers

F.5.1 IsPinned This modopt type supports the use of the type pin_ptr as a parameter. Description: This type is used in the signature of any function. [Example:
public ref class X { public: void F(pin_ptr<int> x) { } };
end example]
F.6 Attributes
Add the ability to chose unambiguously between two attributes called X and XAttribute. End of informative text
284
Portability issues
Annex G. Portability issues

This annex is informative. This annex collects some information about portability that appears in this Standard.
G.1 Undefined behavior

The committee that produced this standard did not intend to introduce any new undefined behavior.
G.2 Implementation-defined behavior

A conforming implementation is required to document its choice of behavior in each of the areas listed in this subclause. The following are implementation-defined: 1. Except for plain char, signed char, and unsigned char, the mapping of fundamental types to CLI types. (12.1.1) 2. If the pre-defined macro __cplusplus_cli is the subject of a #define or a #undef preprocessing directive. (11.1)
G.3 Unspecified behavior

The behavior is unspecified in the following circumstances: 1. Whether the replacement of an __identifier construct takes place before or after translation phase 4. (9.1.1) 2. Whether white space generated by comments, documentation comments, and macro invocations is permitted in the position signified by the symbol. (9.1.2) 3. The semantics of any attribute target specifiers other than those described in this standard. (29.2) 4. The interaction between the CLI library and the Standard C and C++ libraries (except for those requirements described elsewhere in this Standard). (32) End of informative text
285
Annex H. Index
This annex is informative. ... ..........................................................See ellipsis .addon...............................................................215 .class .................................................................207 .custom .....................................................204, 236 .event ................................................................215 .field .................................................................209 .fire ...................................................................215 .get....................................................................213 .locals ...............................................................200 .method.....................................................205, 211 .override ...................................................211, 233 .pack .................................................................209 .param.......................................................205, 237 .property ...........................................................213 .removeon.........................................................215 .set ....................................................................213 .size ..................................................................209 .try ....................................................................235 __identifier("!T")().............................26, 101, 228 __identifier("~T")()............................26, 101, 228 __identifier() ..................................................38 += event handler addition ....................................24 -= event handler removal ....................................24 abstract class............... See class modifier, abstract abstract function ....See function modifier, abstract access assembly.........................................................44 family .............................................................44 family and assembly.......................................44 family or assembly .........................................44 narrower .........................................................44 private.......................................................43, 44 protected.........................................................43 public........................................................43, 44 wider...............................................................44 accessor function add ............................ See add accessor function get.............................. See get accessor function property ..... 21, 109, 111, See also get accessor function; set accessor function remove................ See remove accessor function set ...............................See set accessor function add accessor function .........................................24 add_* reserved names ......................................101 application ............................................................4 286 application domain .............................................. 4 argument list function call ................................................... 72 variable-length ................... See parameter array array ........................................................... 38, 142 creation ........................................................ 143 element access ............................................. 143 initialization ................................................. 144 members....................................................... 144 parameter ..................................................... 104 Standard C++........................................... 4, 142 storage layout............................................... 231 Array.................................................. 90, 142, 144 array covariance......................................... 62, 144 assembly ........................................................ 4, 30 attribute...................4, 32, 159, See also Attribute class naming convention.............................. 159 compilation of an ......................................... 165 delegate........................................................ 163 event............................................................. 163 function........................................................ 163 genuine custom ............................................ 238 instance of an ............................................... 165 name of an ................................................... 162 property........................................................ 163 pseudo custom ............................................. 238 reserved........................................................ 166 specification of an........................................ 161 Attribute................................................... 159, 166 attribute class ................................................... 159 multi-use .............................................. 159, 160 parameter named....................................................... 160 positional ................................................. 160 single-use ..................................................... 159 attribute section................................................ 161 Attribute suffix ................................................ 164 attribute target.................................................. 163 assembly ...................................................... 162 class ............................................................. 162 constructor ................................................... 162 delegate........................................................ 162 enum ............................................................ 162 event............................................................. 162 field.............................................................. 163 interface ....................................................... 163 method ......................................................... 163
Index
parameter......................................................163 property ........................................................163 returnvalue....................................................163 struct.............................................................163 AttributeTargets ...............................................166 AttributeUsage ......... See AttributeUsageAttribute AttributeUsageAttribute...........................159, 166 behavior implementation-defined ...............................285 undefined......................................................285 unspecified ...................................................285 block finally exception thrown from ...............................91 boxing.............................................................4, 14 Byte ....................................................................50 C# Standard......................................................279 callable entity ...................................................153 class abstract ................... See class modifier, abstract attribute ................................. See attribute class enum.......................................... See enum class generic operator and..............................................175 initialization of a ............................................26 interface......................................... See interface native ........................................ See native class ref ...................................................See ref class sealed........................ See class modifier, sealed struct versus....................................................28 default values ...........................................139 inheritance ................................................139 meaning of this.........................................139 class definition ...................................................98 class modifier .....................................................99 abstract ...........................................................99 sealed............................................................100 CLI array ..............................................................4 CLI dispose pattern ....................................26, 221 cli::interior_ptr ..............................See interior_ptr cli::pin_ptr........................................... See pin_ptr cli::safe_cast..................................... See safe_cast CLS ........... See Common Language Specification CLS compliance ...................................................4 collection ......................................................19, 89 System::Array ................................................90 Common Intermediate Language .........................8 Common Language Infrastructure .................... xii Common Language Specification ........................8 Common Type System.................................5, 6, 8 ConditionalAttribute ........................................167 const .......................................... See also constant constant null pointer .....................................................63 constraint ............................................................34 287
class ............................................................... 34 constructor ..................................................... 34 interface ......................................................... 34 constructor delegating..................................................... 282 instance ........................................................ 126 static....................................................... 26, 127 default ...................................................... 128 target ............................................................ 282 conversion boxing ............................................................ 66 explicit ........................................................... 67 implicit constant expression.................................... 66 CTS.............................See Common Type System Current ............................................................... 89 DefaultMemberAttribute ......................... 101, 214 definition non-inline................ See definition, out-of-class out-of-class ...................................................... 4 delegate.............. 4, 19, 23, 153, See also Delegate combining of.................................................. 82 equality of ........ See operator, equality, delegate removal of a ................................................... 82 sealedness of a ............................................. 154 Delegate ..................................................... 19, 153 members of .................................................... 43 destructor ........................................... 25, 131, 223 Dispose()............................................ 26, 101, 224 Dispose(bool)..................................... 26, 101, 225 ellipsis................................................................ 95 enum .................................................................. 11 enum class........................................................ 150 enum struct ...................................................... 150 event ...................................................... 4, 23, 115 abstract......................................................... 117 accessing an ................................................... 71 instance ........................................................ 116 non-trivial .................................................... 115 override........................................................ 117 reserved names ............................................ 101 sealed ........................................................... 117 static............................................................. 116 trivial.............................................. 24, 115, 117 event handler.................................................... 115 examples .............................................................. 9 exception types thrown by certain operations ...... 157, 158 Execution Engine... See Virtual Execution System explicit interface member .................................. 29 field...................................................................... 4 initonly....................................See initonly field literal..........................................See literal field Finalize()............................................ 26, 101, 224 finalizer.............................................. 25, 131, 224
function abstract .............................................................4 pure virtual ...................... See function, abstract reserved names .....................................101, 102 function member ................................................70 function modifier..............................................104 abstract .........................................................107 new ...............................................................108 override ........................................................104 sealed............................................................107 garbage collection ..........................................5, 18 gc-lvalue.......................................... See lvalue, gc generic method ...................... See method, generic generics ....................................................171, 172 get accessor function ..................................21, 111 get_* reserved names .......................................100 get_Item ...........................................................101 GetEnumerator ...................................................89 handle ...................................................................5 null .................................................................40 operations on a .....................................119, 126 heap CLI ...................................................................5 native ................................................................5 hidebyname ........................................................44 hidebysig ....................................................45, 211 IDisposable...............................................131, 222 IEC ...................See International Electrotechnical Commission IEC 60559 standard..............................................3 IEEE .....See Institute of Electrical and Electronics Engineers IEEE 754 standard........... See IEC 60559 standard IEnumerable::GetEnumerator See GetEnumerator IEnumerator::Current ..........................See Current IEnumerator::MoveNext ................See MoveNext inheritance ..........................................................51 initonly field ...............................................21, 129 literal field versus.................................128, 130 instance.................................................................5 Institute of Electrical and Electronics Engineers .8 Int32 ...................................................................12 Int64 ...................................................................12 interface......................................................28, 146 base...............................................................146 delegate ........................................................148 event .............................................................147 function ........................................................147 implementation.............................................148 member.........................................................146 abstract .............................................146, 147 virtual ...............................................146, 147 property ........................................................147 interface class .................................... See interface interface struct................................... See interface 288
interior_ptr ............................................. 16, 38, 54 internal ............................................................... 44 International Electrotechnical Commission......... 8 International Organization for Standardization ... 8 invocation list .................................................. 153 ISO...................See International Organization for Standardization ISO/IEC 10646 .................................................... 3 keyword ............................................................. 38 literal field.................................................. 20, 128 initonly field versus ............................. 128, 130 interdependency of ...................................... 128 restrictions on type of a................................ 128 versioning of a ............................................. 130 lvalue ................................................................... 5 gc 5, 58 MarshalAsAttribute ........................................... 97 member data....................................................... See field member declaration ........................... 99, 128, 129 member name reserved........................................................ 100 metadata............................................................... 5 method generic ........................................................... 35 virtual........................................................... 212 modifier optional ........................................................ 191 required........................................................ 191 modopt ................................See modifier, optional modreq ............................... See modifier, required MoveNext .......................................................... 89 namespace.......................................................... 30 native class....................................................... 133 NativeCppClassAttribute........................... 42, 228 new class member hiding and ............................... 21 new function ...............See function modifier, new newslot............................................................. 212 normative text ...................................................... 9 notes..................................................................... 9 null type ............................................................. 53 null value ........................................................... 62 null value constant ............................................. 40 nullptr null pointer constant and................................ 63 NullReferenceException for each and ................................................... 89 object ................................................................. 13 object reference.....................................See handle Obsolete ..............................See ObsoleteAttribute ObsoleteAttribute............................................. 166 operator equality delegate...................................................... 84
Index
static .............................................................117 C++-dependent.........................................125 CLS-compliant .........................................123 decrement .........................................120, 126 increment..........................................120, 126 synthesis of a ............................................123 output formatted ........................................................11 overload resolution.............................................71 override function ..See function modifier, override override specifier..............................................104 parameter array.............................................16, 94 type parameter and .......................................181 pin_ptr ..........................................................38, 55 pinning .................................................................5 pointer interior .......................................See interior_ptr pinning ............................................ See pin_ptr private type...................See type visibility, private property ..................................................5, 21, 109 abstract .........................................................113 accessing a......................................................71 indexed ...................................................21, 109 accessing an................................................71 default.................................................22, 110 named .......................................................110 instance.........................................................111 read-only ......................................................112 read-write .....................................................112 reserved names .............................................100 scalar ......................................................21, 109 trivial ........................................................114 static .............................................................111 trivial ..............................................................22 write-only .....................................................112 protected public......................see public protected public protected..................................................44 public type.....................See type visibility, public raise_* reserved names ....................................101 rank...................................................................231 rebinding ..............................................................5 ref class ....................................................135, 146 base...............................................................135 restricted types .........................................135 member.........................................................135 ref struct .............................................See ref class remove accessor function ...................................24 remove_* reserved names ................................101 rvalue....................................................................5 safe_cast .......................................................38, 76 SByte............................................................43, 50 members of.....................................................43 sealed class ................... See class modifier, sealed sealed function ........ See function modifier, sealed set accessor function ..........................................21 289
set_* reserved names ....................................... 100 set_Item ........................................................... 101 standard C# ............................................See C# Standard IEC 60559....................See IEC 60559 standard IEEE 754......................See IEC 60559 standard Unicode........................... See Unicode standard strict ................................................................. 211 struct ............................................................ 11, 28 class versus .................................................... 28 default values........................................... 139 inheritance ............................................... 139 meaning of this ........................................ 139 enum .........................................See enum struct inheritance and............................................. 139 ref................................................... See ref class value........................................... See value class System::ArithmeticException.......................... See ArithmeticException System::Array........................................ See Array System::ArrayTypeMismatch.......................... See ArrayTypeMismatch System::Attribute.............................. See Attribute System::AttributeTargets...... See AttributeTargets System::AttributeUsageAttribute .................... See AttributeUsageAttribute System::Delegate .............................. See Delegate System::DivideByZeroException .................... See DivideByZeroException System::ExecutionEngineException................ See ExecutionEngineException System::IDisposable .................... See IDisposable System::IndexOutOfRangeException.............. See IndexOutOfRangeException System::Int32.......................................... See Int32 System::Int64.......................................... See Int64 System::InvalidCastException......................... See InvalidCastException System::MissingFieldException ...................... See MissingFieldException System::MissingMethodException.................. See MissingMethodException System::NullReferenceException.................... See NullReferenceException System::ObsoleteAttribute..See ObsoleteAttribute System::OutOfMemoryException ................... See OutOfMemoryException System::OverflowException............................ See OverflowException System::Reflection::DefaultmemberAttribute ..See DefaultMemberAttribute System::Runtime::InteropServices::MarshalAsSee MarshalAsAttribute System::SByte ....................................... see SByte System::SecurityException See SecurityException
System::StackOverflowException ................... See StackOverflowException System::Type...........................................See Type System::TypeInitializationException ............... See TypeInitializationException System::TypeLoadException ........................... See TypeLoadException System::ValueType ........................see ValueType this constructor call explicit......................................................283 type of in ref class ........................................139 type of in value class ......................................55 ToString .............................................................13 tracking.................................................................5 type array ....................................................See array boxed ................................................................5 class ..................................................... See class any ................................................................5 CLI ...............................................................5 interface........................................................5 native ............................................................5 ref .................................................................5 value .............................................................5 closed ...........................................................178 collection ..................................... See collection constructed .....................................................33 bases of.............................................173, 179 delegate ..........................................................51 element ...........................................................89 fundamental......................................................6 mapping to system class.......................43, 50 members of a ..............................................43 handle ...............................................................6 instance.........................................................172 interface..........................................................51 mixed............................................................141 open ..............................................................178 pointer
native ........................................................... 6 private ...................... See type visibility, private public ........................ See type visibility, public raw ................................................................. 52 reference native ........................................................... 6 tracking ........................................................ 6 simple struct type and.................................... 28, 138 struct ...................................................See struct value class boxed ........................................................... 5 simple........................................................... 5 Type ................................................................... 74 type argument .................................................... 33 type inferencing ................................................. 36 type parameter ................................................... 33 boxing and ................................................... 188 conversion and ............................................. 189 member lookup on ....................................... 187 type visibility ............................................... 57, 98 class ............................................................... 57 default ...................................................... 12, 57 delegate.......................................................... 57 enum .............................................................. 57 interface ......................................................... 57 private ...................................................... 12, 57 public ....................................................... 12, 57 struct .............................................................. 57 unboxing ........................................................ 6, 14 value class member .......................................................... 43 value struct..................................... See value class ValueType ............. 43, 51, 77, 135, 138, 139, 184 variable local ............................................................... 11 versioning .......................................................... 31 VES........................ See Virtual Execution System Virtual Execution System ............................ 5, 6, 8 where ............................................................... 184
290

C++ Cli

Uploaded by

Copyright:

Available Formats

C++ Cli

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

C++ Cli

Uploaded by

Copyright:

Available Formats

ECMA-372

1st Edition / December 2005

C++/CLI Language Specification

C++/CLI Language Specification

Rue du Rhne 114

T/F: +41 22 849 6000/01

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

C++/CLI Language Specification

6. Acronyms and abbreviations

End of informative text

C++/CLI Language Specification

8.1 Getting started

shows this difference. The output produced is

Corresponding CLI Value class type

(with modopt for IsSignUnspecifiedByte)

(with modopt IsLong)

(with modopt IsLong)

Extra-precision floating point type A 16-bit Unicode code unit

An example of using this pseudo-template is:

The output produced is:

// pc4 is the null pointer value // error, no composite type

// error, no implicit conversion to // error, no composite type

// calls h(char*, Object^); // calls h(Object^, int);

C++/CLI Language Specification

The expression %lvalue yields a T^.

{ } Object^ a) { } Object^ a, Object^ b) { } Object^ a, Object^ b, Object^ c)

s, ... array<Object^>^ args) { }

8.4 Automatic memory management

d -= gcnew MyFunction(b, &B::G); // invocation list is A::F A::F d(40);

8.8 Native and ref classes

which produces the following output:

{ return Xor; } { Xor = value; }

{ return Yor; } { Yor = value; }

For a trivial property declaration such as

C++/CLI Language Specification

// D::f overrides B::f // D::g overrides B::g

8.9 Value classes

C++/CLI Language Specification

8.12 Namespaces and assemblies

public ref struct Derived : Base { virtual void F() { Console::WriteLine("Derived::F"); } };

C++/CLI Language Specification

shows several uses of the attribute Help.

C++/CLI Language Specification

C++/CLI Language Specification

End of informative text.

C++/CLI Language Specification

C++/CLI Language Specification

10. Basic concepts

10.2 Application entry point

10.3 Importing types from assemblies

C++/CLI Language Specification

10.4 Reserved names