%!TEX root = std.tex \rSec0[over]{Overloading}% \indextext{overloading|(} \gramSec[gram.over]{Overloading} \rSec1[over.pre]{Preamble} \pnum \indextext{overloaded function|see{overloading}}% \indextext{function!overloaded|see{overloading}}% \begin{note} Each of two or more entities with the same name in the same scope, which must be functions or function templates, is commonly called an ``overload''. \end{note} \pnum When a function is designated in a call, which function declaration is being referenced and the validity of the call are determined by comparing the types of the arguments at the point of use with the types of the parameters in the declarations in the overload set. This function selection process is called \defn{overload resolution} and is defined in~\ref{over.match}. \begin{note} Overload sets are formed by \grammarterm{id-expression}s naming functions and function templates and by \grammarterm{splice-expression}s designating entities of the same kinds. \end{note} \begin{example} \indextext{overloading!example of}% \begin{codeblock} double abs(double); int abs(int); abs(1); // calls \tcode{abs(int);} abs(1.0); // calls \tcode{abs(double);} \end{codeblock} \end{example} \rSec1[over.match]{Overload resolution}% \rSec2[over.match.general]{General}% \indextext{overloading!resolution|(}% \indextext{resolution|see{overloading, resolution}}% \indextext{ambiguity!overloaded function} \pnum Overload resolution is a mechanism for selecting the best function to call given a list of expressions that are to be the arguments of the call and a set of \defnx{candidate functions}{candidate} that can be called based on the context of the call. The selection criteria for the best function are the number of arguments, how well the arguments match the parameter-type-list of the candidate function, how well (for non-static member functions) the object matches the object parameter, and certain other properties of the candidate function. \begin{note} The function selected by overload resolution is not guaranteed to be appropriate for the context. Other restrictions, such as the accessibility of the function, can make its use in the calling context ill-formed. \end{note} \pnum \indextext{overloading!resolution!contexts}% Overload resolution selects the function to call in seven distinct contexts within the language: \begin{itemize} \item invocation of a function named in the function call syntax\iref{over.call.func}; \item invocation of a function call operator, a pointer-to-function conversion function, a reference-to-pointer-to-function conversion function, or a reference-to-function conversion function on a class object named in the function call syntax\iref{over.call.object}; \item invocation of the operator referenced in an expression\iref{over.match.oper}; \item invocation of a constructor for default- or direct-initialization\iref{dcl.init} of a class object\iref{over.match.ctor}; \item invocation of a user-defined conversion for copy-initialization\iref{dcl.init} of a class object\iref{over.match.copy}; \item invocation of a conversion function for initialization of an object of a non-class type from an expression of class type\iref{over.match.conv}; and \item invocation of a conversion function for conversion in which a reference\iref{dcl.init.ref} will be directly bound\iref{over.match.ref}. \end{itemize} Each of these contexts defines the set of candidate functions and the list of arguments in its own unique way. But, once the candidate functions and argument lists have been identified, the selection of the best function is the same in all cases: \begin{itemize} \item First, a subset of the candidate functions (those that have the proper number of arguments and meet certain other conditions) is selected to form a set of \indextext{function!viable}% viable functions\iref{over.match.viable}. \item Then the best viable function is selected based on the implicit conversion sequences\iref{over.best.ics} needed to match each argument to the corresponding parameter of each viable function. \end{itemize} \pnum If a best viable function exists and is unique, overload resolution succeeds and produces it as the result. Otherwise overload resolution fails and the invocation is ill-formed. When overload resolution succeeds, and the best viable function is not accessible\iref{class.access} in the context in which it is used, the program is ill-formed. \pnum Overload resolution results in a \defnadj{usable}{candidate} if overload resolution succeeds and the selected candidate is either not a function\iref{over.built}, or is a function that is not deleted and is accessible from the context in which overload resolution was performed. \rSec2[over.match.funcs]{Candidate functions and argument lists}% \rSec3[over.match.funcs.general]{General}% \indextext{overloading!candidate functions|(}% \indextext{overloading!argument lists|(} \pnum The subclauses of~\ref{over.match.funcs} describe the set of candidate functions and the argument list submitted to overload resolution in each context in which overload resolution is used. The source transformations and constructions defined in these subclauses are only for the purpose of describing the overload resolution process. An implementation is not required to use such transformations and constructions. \pnum \indextext{member function!overload resolution and}% \indextext{function!overload resolution and}% The set of candidate functions can contain both member and non-member functions to be resolved against the same argument list. If a member function is \begin{itemize} \item an implicit object member function that is not a constructor, or \item a static member function and the argument list includes an implied object argument, \end{itemize} it is considered to have an extra first parameter, called the \defnadj{implicit}{object parameter}, which represents the object for which the member function has been called. \pnum Similarly, when appropriate, the context can construct an argument list that contains an \defn{implied object argument} as the first argument in the list to denote the object to be operated on. \pnum For implicit object member functions, the type of the implicit object parameter is \begin{itemize} \item ``lvalue reference to \cv{}~\tcode{X}'' for functions declared without a \grammarterm{ref-qualifier} or with the \tcode{\&} \grammarterm{ref-qualifier} \item ``rvalue reference to \cv{}~\tcode{X}'' for functions declared with the \tcode{\&\&} \grammarterm{ref-qualifier} \end{itemize} where \tcode{X} is the class of which the function is a direct member and \cv{} is the cv-qualification on the member function declaration. \begin{example} For a \keyword{const} member function of class \tcode{X}, the extra parameter is assumed to have type ``lvalue reference to \tcode{const X}''. \end{example} For conversion functions that are implicit object member functions, the function is considered to be a member of the class of the implied object argument for the purpose of defining the type of the implicit object parameter. For non-conversion functions that are implicit object member functions nominated by a \grammarterm{using-declaration} in a derived class, the function is considered to be a member of the derived class for the purpose of defining the type of the implicit object parameter. For static member functions, the implicit object parameter is considered to match any object (since if the function is selected, the object is discarded). \begin{note} No actual type is established for the implicit object parameter of a static member function, and no attempt will be made to determine a conversion sequence for that parameter\iref{over.match.best}. \end{note} \pnum \indextext{implied object argument!implicit conversion sequences}% During overload resolution, the implied object argument is indistinguishable from other arguments. The implicit object parameter, however, retains its identity since no user-defined conversions can be applied to achieve a type match with it. \indextext{implied object argument!non-static member function and}% For implicit object member functions declared without a \grammarterm{ref-qualifier}, even if the implicit object parameter is not const-qualified, an rvalue can be bound to the parameter as long as in all other respects the argument can be converted to the type of the implicit object parameter. \begin{note} The fact that such an argument is an rvalue does not affect the ranking of implicit conversion sequences\iref{over.ics.rank}. \end{note} \pnum Because other than in list-initialization only one user-defined conversion is allowed in an implicit conversion sequence, special rules apply when selecting the best user-defined conversion\iref{over.match.best,over.best.ics}. \begin{example} \begin{codeblock} class T { public: T(); }; class C : T { public: C(int); }; T a = 1; // error: no viable conversion (\tcode{T(C(1))} not considered) \end{codeblock} \end{example} \pnum In each case where conversion functions of a class \tcode{S} are considered for initializing an object or reference of type \tcode{T}, the candidate functions include the result of a search for the \grammarterm{conversion-function-id} \tcode{\keyword{operator} T} in \tcode{S}. \begin{note} This search can find a specialization of a conversion function template\iref{basic.lookup}. \end{note} Each such case also defines sets of \defnadj{permissible}{types} for explicit and non-explicit conversion functions; each (non-template) conversion function that \begin{itemize} \item is a non-hidden member of \tcode{S}, \item yields a permissible type, and, \item for the former set, is non-explicit \end{itemize} is also a candidate function. If initializing an object, for any permissible type \cv{} \tcode{U}, any \cvqual{cv2} \tcode{U}, \cvqual{cv2} \tcode{U\&}, or \cvqual{cv2} \tcode{U\&\&} is also a permissible type. If the set of permissible types for explicit conversion functions is empty, any candidates that are explicit are discarded. \pnum In each case where a candidate is a function template, candidate function template specializations are generated using template argument deduction\iref{temp.over,temp.deduct}. If a constructor template or conversion function template has an \grammarterm{explicit-specifier} whose \grammarterm{constant-expression} is value-dependent\iref{temp.dep}, template argument deduction is performed first and then, if the context admits only candidates that are not explicit and the generated specialization is explicit\iref{dcl.fct.spec}, it will be removed from the candidate set. Those candidates are then handled as candidate functions in the usual way. \begin{footnote} The process of argument deduction fully determines the parameter types of the function template specializations, i.e., the parameters of function template specializations contain no template parameter types. Therefore, except where specified otherwise, function template specializations and non-template functions\iref{dcl.fct} are treated equivalently for the remainder of overload resolution. \end{footnote} A given name can refer to, or a conversion can consider, one or more function templates as well as a set of non-template functions. In such a case, the candidate functions generated from each function template are combined with the set of non-template candidate functions. \pnum A defaulted move special member function\iref{class.copy.ctor,class.copy.assign} that is defined as deleted is excluded from the set of candidate functions in all contexts. A constructor inherited from class type \tcode{C}\iref{class.inhctor.init} that has a first parameter of type ``reference to \cvqual{cv1} \tcode{P}'' (including such a constructor instantiated from a template) is excluded from the set of candidate functions when constructing an object of type \cvqual{cv2} \tcode{D} if the argument list has exactly one argument and \tcode{C} is reference-related to \tcode{P} and \tcode{P} is reference-related to \tcode{D}. \begin{example} \begin{codeblock} struct A { A(); // \#1 A(A &&); // \#2 template A(T &&); // \#3 }; struct B : A { using A::A; B(const B &); // \#4 B(B &&) = default; // \#5, implicitly deleted struct X { X(X &&) = delete; } x; }; extern B b1; B b2 = static_cast(b1); // calls \#4: \#1 is not viable, \#2, \#3, and \#5 are not candidates struct C { operator B&&(); }; B b3 = C(); // calls \#4 \end{codeblock} \end{example} \rSec3[over.match.call]{Function call syntax}% \rSec4[over.match.call.general]{General}% \indextext{overloading!resolution!function call syntax|(} \pnum In a function call\iref{expr.call} \begin{ncsimplebnf} postfix-expression \terminal{(} \opt{expression-list} \terminal{)} \end{ncsimplebnf} if the \grammarterm{postfix-expression} names at least one function or function template, overload resolution is applied as specified in \ref{over.call.func}. If the \grammarterm{postfix-expression} denotes an object of class type, overload resolution is applied as specified in \ref{over.call.object}. \pnum If the \grammarterm{postfix-expression} is the address of an overload set, overload resolution is applied using that set as described above. \begin{note} No implied object argument is added in this case. \end{note} If the function selected by overload resolution is an implicit object member function, the program is ill-formed. \begin{note} The resolution of the address of an overload set in other contexts is described in \ref{over.over}. \end{note} \rSec4[over.call.func]{Call to designated function} \pnum Of interest in~\ref{over.call.func} are only those function calls in which the \grammarterm{postfix-expression} ultimately contains an \grammarterm{id-expression} or \grammarterm{splice-expression} that designates one or more functions. Such a \grammarterm{postfix-expression}, perhaps nested arbitrarily deep in parentheses, has one of the following forms: \begin{ncbnf} postfix-expression:\br postfix-expression \terminal{.} id-expression\br postfix-expression \terminal{.} splice-expression\br postfix-expression \terminal{->} id-expression\br postfix-expression \terminal{->} splice-expression\br id-expression\br splice-expression \end{ncbnf} These represent two syntactic subcategories of function calls: qualified function calls and unqualified function calls. \pnum In qualified function calls, the function is designated by an \grammarterm{id-expression} or \grammarterm{splice-expression} $E$ preceded by an \tcode{->} or \tcode{.} operator. Since the construct \tcode{A->B} is generally equivalent to \tcode{(*A).B}, the rest of \ref{over} assumes, without loss of generality, that all member function calls have been normalized to the form that uses an object and the \tcode{.} operator. Furthermore, \ref{over} assumes that the \grammarterm{postfix-expression} that is the left operand of the \tcode{.} operator has type ``\cv{}~\tcode{T}'' where \tcode{T} denotes a class. \begin{footnote} Note that cv-qualifiers on the type of objects are significant in overload resolution for both glvalue and class prvalue objects. \end{footnote} The set of candidate functions either is the set found by name lookup\iref{class.member.lookup} if $E$ is an \grammarterm{id-expression} or is the set determined as specified in~\ref{expr.prim.splice} if $E$ is a \grammarterm{splice-expression}. The argument list is the \grammarterm{expression-list} in the call augmented by the addition of the left operand of the \tcode{.} operator in the normalized member function call as the implied object argument\iref{over.match.funcs}. \pnum In unqualified function calls, the function is designated by an \grammarterm{id-expression} or a \grammarterm{splice-expression} $E$. The set of candidate functions either is the set found by name lookup\iref{basic.lookup} if $E$ is an \grammarterm{id-expression} or is the set determined as specified in~\ref{expr.prim.splice} if $E$ is a \grammarterm{splice-expression}. The set of candidate functions consists either entirely of non-member functions or entirely of member functions of some class \tcode{T}. In the former case or if $E$ is either a \grammarterm{splice-expression} or the address of an overload set, the argument list is the same as the \grammarterm{expression-list} in the call. Otherwise, the argument list is the \grammarterm{expression-list} in the call augmented by the addition of an implied object argument as in a qualified function call. If the current class is, or is derived from, \tcode{T}, and the keyword \keyword{this}\iref{expr.prim.this} refers to it, \begin{itemize} \item if the unqualified function call appears in a precondition assertion of a constructor or a postcondition assertion of a destructor and overload resolution selects a non-static member function, the call is ill-formed; \item otherwise, the implied object argument is \tcode{(*\keyword{this})}. \end{itemize} Otherwise, \begin{itemize} \item if overload resolution selects a non-static member function, the call is ill-formed; \item otherwise, a contrived object of type \tcode{T} becomes the implied object argument. \begin{footnote} An implied object argument is contrived to correspond to the implicit object parameter attributed to member functions during overload resolution. It is not used in the call to the selected function. Since the member functions all have the same implicit object parameter, the contrived object will not be the cause to select or reject a function. \end{footnote} \end{itemize} \begin{example} \begin{codeblock} struct C { bool a(); void b() { a(); // OK, \tcode{(*this).a()} } void c(this const C&); // \#1 void c() &; // \#2 static void c(int = 0); // \#3 void d() { c(); // error: ambiguous between \#2 and \#3 (C::c)(); // error: as above (&(C::c))(); // error: cannot resolve address of overloaded \tcode{this->C::c}\iref{over.over} (&C::c)(C{}); // selects \#1 (&C::c)(*this); // error: selects \#2, and is ill-formed\iref{over.match.call.general} (&C::c)(); // selects \#3 } void f(this const C&); void g() const { f(); // OK, \tcode{(*this).f()} f(*this); // error: no viable candidate for \tcode{(*this).f(*this)} this->f(); // OK } static void h() { f(); // error: contrived object argument, but overload resolution // picked a non-static member function f(C{}); // error: no viable candidate C{}.f(); // OK } void k(this int); operator int() const; void m(this const C& c) { c.k(); // OK } C() pre(a()) // error: implied \keyword{this} in constructor precondition pre(this->a()) // OK post(a()); // OK ~C() pre(a()) // OK post(a()) // error: implied \keyword{this} in destructor postcondition post(this->a()); // OK }; \end{codeblock} \end{example} \rSec4[over.call.object]{Call to object of class type} \pnum If the \grammarterm{postfix-expression} \tcode{E} in the function call syntax evaluates to a class object of type ``\cv{}~\tcode{T}'', then the set of candidate functions includes at least the function call operators of \tcode{T}. The function call operators of \tcode{T} are the results of a search for the name \tcode{\keyword{operator}()} in the scope of \tcode{T}. \pnum In addition, for each non-explicit conversion function declared in \tcode{T} of the form \begin{ncsimplebnf} \keyword{operator} conversion-type-id \terminal{(\,)} \opt{cv-qualifier-seq} \opt{ref-qualifier} \opt{noexcept-specifier} \opt{attribute-specifier-seq} \terminal{;} \end{ncsimplebnf} where the optional \grammarterm{cv-qualifier-seq} is the same cv-qualification as, or a greater cv-qualification than, \cv{}, and where \grammarterm{conversion-type-id} denotes the type ``pointer to function of ($\tcode{P}_1, \dotsc, \tcode{P}_n$) returning \tcode{R}'', or the type ``reference to pointer to function of ($\tcode{P}_1, \dotsc, \tcode{P}_n$) returning \tcode{R}'', or the type ``reference to function of ($\tcode{P}_1, \dotsc, \tcode{P}_n$) returning \tcode{R}'', a \defn{surrogate call function} with the unique name \placeholder{call-function} and having a declaration of the form \begin{ncbnf} \terminal{R} \placeholder{call-function} \terminal{(} conversion-type-id \ % \terminal{F, P$_1$ a$_1$, $\dotsc$, P$_n$ a$_n$);} \end{ncbnf} is also considered as a candidate function. \begin{note} If a surrogate call function is selected by overload resolution, the behavior is as described in~\ref{over.call}. \end{note} Similarly, surrogate call functions are added to the set of candidate functions for each non-explicit conversion function declared in a base class of \tcode{T} provided the function is not hidden within \tcode{T} by another intervening declaration. \begin{footnote} Note that this construction can yield candidate call functions that cannot be differentiated one from the other by overload resolution because they have identical declarations or differ only in their return type. The call will be ambiguous if overload resolution cannot select a match to the call that is uniquely better than such undifferentiable functions. \end{footnote} \pnum The argument list submitted to overload resolution consists of the argument expressions present in the function call syntax preceded by the implied object argument \tcode{(E)}. \begin{note} When comparing the call against the function call operators, the implied object argument is compared against the object parameter of the function call operator. When comparing the call against a surrogate call function, the implied object argument is compared against the first parameter of the surrogate call function. \end{note} \begin{example} \begin{codeblock} int f1(int); int f2(float); typedef int (*fp1)(int); typedef int (*fp2)(float); struct A { operator fp1() { return f1; } operator fp2() { return f2; } } a; int i = a(1); // calls \tcode{f1} via pointer returned from conversion function \end{codeblock} \end{example} \indextext{overloading!resolution!function call syntax|)} \rSec3[over.match.oper]{Operators in expressions}% \indextext{overloading!resolution!operators} \pnum If no operand of an operator in an expression has a type that is a class or an enumeration, the operator is assumed to be a built-in operator and interpreted according to \ref{expr.compound}. \begin{note} Because \tcode{.}, \tcode{.*}, and \tcode{::} cannot be overloaded, these operators are always built-in operators interpreted according to \ref{expr.compound}. \tcode{?:} cannot be overloaded, but the rules in this subclause are used to determine the conversions to be applied to the second and third operands when they have class or enumeration type\iref{expr.cond}. \end{note} \begin{example} \begin{codeblock} struct String { String (const String&); String (const char*); operator const char* (); }; String operator + (const String&, const String&); void f() { const char* p= "one" + "two"; // error: cannot add two pointers; overloaded \tcode{\keyword{operator}+} not considered // because neither operand has class or enumeration type int I = 1 + 1; // always evaluates to \tcode{2} even if class or enumeration types exist // that would perform the operation. } \end{codeblock} \end{example} \pnum If either operand has a type that is a class or an enumeration, a user-defined operator function can be declared that implements this operator or a user-defined conversion can be necessary to convert the operand to a type that is appropriate for a built-in operator. In this case, overload resolution is used to determine which operator function or built-in operator is to be invoked to implement the operator. Therefore, the operator notation is first transformed to the equivalent function-call notation as summarized in \tref{over.match.oper} (where \tcode{@} denotes one of the operators covered in the specified subclause). However, the operands are sequenced in the order prescribed for the built-in operator\iref{expr.compound}. \begin{floattable}{Relationship between operator and function call notation}{over.match.oper} {l|l|l|l} \topline \hdstyle{Subclause} & \hdstyle{Expression} & \hdstyle{As member function} & \hdstyle{As non-member function} \\ \capsep \ref{over.unary} & \tcode{@a} & \tcode{(a).\keyword{operator}@ (\,)} & \tcode{\keyword{operator}@(a)} \\ \ref{over.binary} & \tcode{a@b} & \tcode{(a).\keyword{operator}@ (b)} & \tcode{\keyword{operator}@(a, b)} \\ \ref{over.assign} & \tcode{a=b} & \tcode{(a).\keyword{operator}= (b)} & \\ \ref{over.sub} & \tcode{a[b]} & \tcode{(a).\keyword{operator}[](b)} & \\ \ref{over.ref} & \tcode{a->} & \tcode{(a).\keyword{operator}->(\,)} & \\ \ref{over.inc} & \tcode{a@} & \tcode{(a).\keyword{operator}@ (0)} & \tcode{\keyword{operator}@(a, 0)} \\ \end{floattable} \pnum For a unary operator \tcode{@} with an operand of type \cvqual{cv1} \tcode{T1}, and for a binary operator \tcode{@} with a left operand of type \cvqual{cv1} \tcode{T1} and a right operand of type \cvqual{cv2} \tcode{T2}, four sets of candidate functions, designated \defnx{member candidates}{member candidate}, \defnx{non-member candidates}{non-member candidate}, \defnx{built-in candidates}{built-in candidate}, and \defnx{rewritten candidates}{rewritten candidate}, are constructed as follows: \begin{itemize} \item If \tcode{T1} is a complete class type or a class currently being defined, the set of member candidates is the result of a search for \tcode{\keyword{operator}@} in the scope of \tcode{T1}; otherwise, the set of member candidates is empty. \item For the operators \tcode{=}, \tcode{[]}, or \tcode{->}, the set of non-member candidates is empty; otherwise, it includes the result of unqualified lookup for \tcode{\keyword{operator}@} in the rewritten function call\iref{basic.lookup.unqual,basic.lookup.argdep}, ignoring all member functions. However, if no operand has a class type, only those non-member functions in the lookup set that have a first parameter of type \tcode{T1} or ``reference to \cv{}~\tcode{T1}'', when \tcode{T1} is an enumeration type, or (if there is a right operand) a second parameter of type \tcode{T2} or ``reference to \cv{}~\tcode{T2}'', when \tcode{T2} is an enumeration type, are candidate functions. \item For the operator \tcode{,}, the unary operator \tcode{\&}, or the operator \tcode{->}, the built-in candidates set is empty. For all other operators, the built-in candidates include all of the candidate operator functions defined in~\ref{over.built} that, compared to the given operator, \begin{itemize} \item have the same operator name, and \item accept the same number of operands, and \item accept operand types to which the given operand or operands can be converted according to \ref{over.best.ics}, and \item do not have the same parameter-type-list as any non-member candidate or rewritten non-member candidate that is not a function template specialization. \end{itemize} \item The rewritten candidate set is determined as follows: \begin{itemize} \item For the relational\iref{expr.rel} operators, the rewritten candidates include all non-rewritten candidates for the expression \tcode{x <=> y}. \item For the relational\iref{expr.rel} and three-way comparison\iref{expr.spaceship} operators, the rewritten candidates also include a synthesized candidate, with the order of the two parameters reversed, for each non-rewritten candidate for the expression \tcode{y <=> x}. \item For the \tcode{!=} operator\iref{expr.eq}, the rewritten candidates include all non-rewritten candidates for the expression \tcode{x == y} that are rewrite targets with first operand \tcode{x} (see below). \item For the equality operators, the rewritten candidates also include a synthesized candidate, with the order of the two parameters reversed, for each non-rewritten candidate for the expression \tcode{y == x} that is a rewrite target with first operand \tcode{y}. \item For all other operators, the rewritten candidate set is empty. \end{itemize} \begin{note} A candidate synthesized from a member candidate has its object parameter as the second parameter, thus implicit conversions are considered for the first, but not for the second, parameter. \end{note} \end{itemize} \pnum A non-template function or function template \tcode{F} named \tcode{\keyword{operator}==} is a rewrite target with first operand \tcode{o} unless a search for the name \tcode{\keyword{operator}!=} in the scope $S$ from the instantiation context of the operator expression finds a function or function template that would correspond\iref{basic.scope.scope} to \tcode{F} if its name were \tcode{\keyword{operator}==}, where $S$ is the scope of the class type of \tcode{o} if \tcode{F} is a class member, and the namespace scope of which \tcode{F} is a member otherwise. A function template specialization named \tcode{\keyword{operator}==} is a rewrite target if its function template is a rewrite target. \begin{example} \begin{codeblock} struct A {}; template bool operator==(A, T); // \#1 bool a1 = 0 == A(); // OK, calls reversed \#1 template bool operator!=(A, T); bool a2 = 0 == A(); // error, \#1 is not a rewrite target struct B { bool operator==(const B&); // \#2 }; struct C : B { C(); C(B); bool operator!=(const B&); // \#3 }; bool c1 = B() == C(); // OK, calls \#2; reversed \#2 is not a candidate // because search for \tcode{\keyword{operator}!=} in \tcode{C} finds \#3 bool c2 = C() == B(); // error: ambiguous between \#2 found when searching \tcode{C} and // reversed \#2 found when searching \tcode{B} struct D {}; template bool operator==(D, T); // \#4 inline namespace N { template bool operator!=(D, T); // \#5 } bool d1 = 0 == D(); // OK, calls reversed \#4; \#5 does not forbid \#4 as a rewrite target \end{codeblock} \end{example} \pnum For the first parameter of the built-in assignment operators, only standard conversion sequences\iref{over.ics.scs} are considered. \pnum For all other operators, no such restrictions apply. \pnum The set of candidate functions for overload resolution for some operator \tcode{@} is the union of the member candidates, the non-member candidates, the built-in candidates, and the rewritten candidates for that operator \tcode{@}. \pnum The argument list contains all of the operands of the operator. The best function from the set of candidate functions is selected according to~\ref{over.match.viable} and~\ref{over.match.best}. \begin{footnote} If the set of candidate functions is empty, overload resolution is unsuccessful. \end{footnote} \begin{example} \begin{codeblock} struct A { operator int(); }; A operator+(const A&, const A&); void m() { A a, b; a + b; // \tcode{\keyword{operator}+(a, b)} chosen over \tcode{int(a) + int(b)} } \end{codeblock} \end{example} \pnum If a rewritten \tcode{\keyword{operator}<=>} candidate is selected by overload resolution for an operator \tcode{@}, \tcode{x @ y} is interpreted as \tcode{0 @ (y <=> x)} if the selected candidate is a synthesized candidate with reversed order of parameters, or \tcode{(x <=> y) @ 0} otherwise, using the selected rewritten \tcode{\keyword{operator}<=>} candidate. Rewritten candidates for the operator \tcode{@} are not considered in the context of the resulting expression. \pnum If a rewritten \tcode{\keyword{operator}==} candidate is selected by overload resolution for an operator \tcode{@}, its return type shall be \cv{} \tcode{bool}, and \tcode{x @ y} is interpreted as: \begin{itemize} \item if \tcode{@} is \tcode{!=} and the selected candidate is a synthesized candidate with reversed order of parameters, \tcode{!(y == x)}, \item otherwise, if \tcode{@} is \tcode{!=}, \tcode{!(x == y)}, \item otherwise (when \tcode{@} is \tcode{==}), \tcode{y == x}, \end{itemize} in each case using the selected rewritten \tcode{\keyword{operator}==} candidate. \pnum If a built-in candidate is selected by overload resolution, the operands of class type are converted to the types of the corresponding parameters of the selected operation function, except that the second standard conversion sequence of a user-defined conversion sequence\iref{over.ics.user} is not applied. Then the operator is treated as the corresponding built-in operator and interpreted according to \ref{expr.compound}. \begin{example} \begin{codeblock} struct X { operator double(); }; struct Y { operator int*(); }; int *a = Y() + 100.0; // error: pointer arithmetic requires integral operand int *b = Y() + X(); // error: pointer arithmetic requires integral operand \end{codeblock} \end{example} \pnum The second operand of operator \tcode{->} is ignored in selecting an \tcode{\keyword{operator}->} function, and is not an argument when the \tcode{\keyword{operator}->} function is called. When \tcode{\keyword{operator}->} returns, the operator \tcode{->} is applied to the value returned, with the original second operand. \begin{footnote} If the value returned by the \tcode{\keyword{operator}->} function has class type, this can result in selecting and calling another \tcode{\keyword{operator}->} function. The process repeats until an \tcode{\keyword{operator}->} function returns a value of non-class type. \end{footnote} \pnum If the operator is the operator \tcode{,}, the unary operator \tcode{\&}, or the operator \tcode{->}, and there are no viable functions, then the operator is assumed to be the built-in operator and interpreted according to \ref{expr.compound}. \pnum \begin{note} The lookup rules for operators in expressions are different than the lookup rules for operator function names in a function call, as shown in the following example: \begin{codeblock} struct A { }; void operator + (A, A); struct B { void operator + (B); void f (); }; A a; void B::f() { operator+ (a,a); // error: global operator hidden by member a + a; // OK, calls global \tcode{\keyword{operator}+} } \end{codeblock} \end{note} \rSec3[over.match.ctor]{Initialization by constructor}% \indextext{overloading!resolution!initialization} \pnum When objects of class type are direct-initialized\iref{dcl.init}, copy-initialized from an expression of the same or a derived class type\iref{dcl.init}, or default-initialized\iref{dcl.init}, overload resolution selects the constructor. For direct-initialization or default-initialization (including default-initialization in the context of copy-list-initialization), the candidate functions are all the constructors of the class of the object being initialized. Otherwise, the candidate functions are all the non-explicit constructors\iref{class.conv.ctor} of that class. The argument list is the \grammarterm{expression-list} or \grammarterm{assignment-expression} of the \grammarterm{initializer}; for default-initialization, the argument list is empty. For default-initialization in the context of copy-list-initialization, if an explicit constructor is chosen, the initialization is ill-formed. \rSec3[over.match.copy]{Copy-initialization of class by user-defined conversion}% \indextext{overloading!resolution!initialization} \pnum Under the conditions specified in~\ref{dcl.init}, as part of a copy-initialization of an object of class type, a user-defined conversion can be invoked to convert an initializer expression to the type of the object being initialized. Overload resolution is used to select the user-defined conversion to be invoked. \begin{note} The conversion performed for indirect binding to a reference to a possibly cv-qualified class type is determined in terms of a corresponding non-reference copy-initialization. \end{note} Assuming that ``\cvqual{cv1} \tcode{T}'' is the type of the object being initialized, with \tcode{T} a class type, the candidate functions are selected as follows: \begin{itemize} \item The non-explicit constructors\iref{class.conv.ctor} of \tcode{T} are candidate functions. \item When the type of the initializer expression is a class type ``\cv{}~\tcode{S}'', conversion functions are considered. The permissible types for non-explicit conversion functions are \tcode{T} and any class derived from \tcode{T}. When initializing a temporary object\iref{class.mem} to be bound to the first parameter of a constructor where the parameter is of type ``reference to \cvqual{cv2} \tcode{T}'' and the constructor is called with a single argument in the context of direct-initialization of an object of type ``\cvqual{cv3} \tcode{T}'', the permissible types for explicit conversion functions are the same; otherwise there are none. \end{itemize} \pnum In both cases, the argument list has one argument, which is the initializer expression. \begin{note} This argument will be compared against the first parameter of the constructors and against the object parameter of the conversion functions. \end{note} \rSec3[over.match.conv]{Initialization by conversion function}% \indextext{overloading!resolution!initialization} \pnum Under the conditions specified in~\ref{dcl.init}, as part of an initialization of an object of non-class type, a conversion function can be invoked to convert an initializer expression of class type to the type of the object being initialized. Overload resolution is used to select the conversion function to be invoked. Assuming that ``\cv{} \tcode{T}'' is the type of the object being initialized, the candidate functions are selected as follows: \begin{itemize} \item The permissible types for non-explicit conversion functions are those that can be converted to type \tcode{T} via a standard conversion sequence\iref{over.ics.scs}. For direct-initialization, the permissible types for explicit conversion functions are those that can be converted to type \tcode{T} with a (possibly trivial) qualification conversion\iref{conv.qual}; otherwise there are none. \end{itemize} \pnum The argument list has one argument, which is the initializer expression. \begin{note} This argument will be compared against the object parameter of the conversion functions. \end{note} \rSec3[over.match.ref]{Initialization by conversion function for direct reference binding}% \indextext{overloading!resolution!initialization} \pnum Under the conditions specified in~\ref{dcl.init.ref}, a reference can be bound directly to the result of applying a conversion function to an initializer expression. Overload resolution is used to select the conversion function to be invoked. Assuming that ``reference to \cvqual{cv1} \tcode{T}'' is the type of the reference being initialized, the candidate functions are selected as follows: \begin{itemize} \item Let $R$ be the set of all \begin{itemize} \item lvalue reference types (when converting to an lvalue) and \item non-reference types and rvalue reference types (when converting to an rvalue or an lvalue of function type). \end{itemize} The permissible types for non-explicit conversion functions are the members of $R$ having the form ``\cv{}~\tcode{T2}'' or ``reference to \cvqual{cv2}~\tcode{T2}'' where ``\cvqual{cv1} \tcode{T}'' is reference-compatible\iref{dcl.init.ref} with ``\cvqual{cv2} \tcode{T2}''. For direct-initialization, the permissible types for explicit conversion functions are the members of $R$ having the form ``\cv{}~\tcode{T2}'' or ``reference to \cvqual{cv2}~\tcode{T2}'' where \tcode{T2} can be converted to type \tcode{T} with a (possibly trivial) qualification conversion\iref{conv.qual}; otherwise there are none. \end{itemize} \pnum The argument list has one argument, which is the initializer expression. \begin{note} This argument will be compared against the object parameter of the conversion functions. \end{note} \rSec3[over.match.list]{Initialization by list-initialization}% \indextext{overloading!resolution!initialization} \pnum When objects of non-aggregate class type \tcode{T} are list-initialized such that \ref{dcl.init.list} specifies that overload resolution is performed according to the rules in this subclause or when forming a list-initialization sequence according to \ref{over.ics.list}, overload resolution selects the constructor in two phases: \begin{itemize} \item If the initializer list is not empty or \tcode{T} has no default constructor, overload resolution is first performed where the candidate functions are the initializer-list constructors\iref{dcl.init.list} of the class \tcode{T} and the argument list consists of the initializer list as a single argument. \item Otherwise, or if no viable initializer-list constructor is found, overload resolution is performed again, where the candidate functions are all the constructors of the class \tcode{T} and the argument list consists of the elements of the initializer list. \end{itemize} In copy-list-initialization, if an explicit constructor is chosen, the initialization is ill-formed. \begin{note} This differs from other situations\iref{over.match.ctor,over.match.copy}, where only non-explicit constructors are considered for copy-initialization. This restriction only applies if this initialization is part of the final result of overload resolution. \end{note} \rSec3[over.match.class.deduct]{Class template argument deduction}% \indextext{deduction!class template arguments}% \pnum When resolving a placeholder for a deduced class type\iref{dcl.type.class.deduct} where the \grammarterm{template-name} or \grammarterm{splice-type-specifier} designates a primary class template \tcode{C}, a set of functions and function templates, called the guides of \tcode{C}, is formed comprising: \begin{itemize} \item If \tcode{C} is defined, for each constructor of \tcode{C}, a function template with the following properties: \begin{itemize} \item The template parameters are the template parameters of \tcode{C} followed by the template parameters (including default template arguments) of the constructor, if any. \item The associated constraints\iref{temp.constr.decl} are the conjunction of the associated constraints of \tcode{C} and the associated constraints of the constructor, if any. \begin{note} A \grammarterm{constraint-expression} in the \grammarterm{template-head} of \tcode{C} is checked for satisfaction before any constraints from the \grammarterm{template-head} or trailing \grammarterm{requires-clause} of the constructor. \end{note} \item The \grammarterm{parameter-declaration-clause} is that of the constructor. \item The return type is the class template specialization designated by \tcode{C} and template arguments corresponding to the template parameters of \tcode{C}. \end{itemize} \item If \tcode{C} is not defined or does not declare any constructors, an additional function template derived as above from a hypothetical constructor \tcode{C()}. \item An additional function template derived as above from a hypothetical constructor \tcode{C(C)}, called the \defn{copy deduction candidate}. \item For each \grammarterm{deduction-guide}, a function or function template with the following properties: \begin{itemize} \item The \grammarterm{template-head}, if any, and \grammarterm{parameter-declaration-clause} are those of the \grammarterm{deduction-guide}. \item The return type is the \grammarterm{simple-template-id} of the \grammarterm{deduction-guide}. \end{itemize} \end{itemize} In addition, if \tcode{C} is defined and its definition satisfies the conditions for an aggregate class\iref{dcl.init.aggr} with the assumption that any dependent base class has no virtual functions and no virtual base classes, and the initializer is a non-empty \grammarterm{braced-init-list} or parenthesized \grammarterm{expression-list}, and there are no \grammarterm{deduction-guide}{s} for \tcode{C}, the set contains an additional function template, called the \defnadj{aggregate deduction}{candidate}, defined as follows. Let $x_1, \dotsc, x_n$ be the elements of the \grammarterm{initializer-list} or \grammarterm{designated-initializer-list} of the \grammarterm{braced-init-list}, or of the \grammarterm{expression-list}. For each $x_i$, let $e_i$ be the corresponding aggregate element of \tcode{C} or of one of its (possibly recursive) subaggregates that would be initialized by $x_i$\iref{dcl.init.aggr} if \begin{itemize} \item brace elision is not considered for any aggregate element that has \begin{itemize} \item a dependent non-array type, \item an array type with a value-dependent bound, or \item an array type with a dependent array element type and $x_i$ is a string literal; and \end{itemize} \item each non-trailing aggregate element that is a pack expansion is assumed to correspond to no elements of the initializer list, and \item a trailing aggregate element that is a pack expansion is assumed to correspond to all remaining elements of the initializer list (if any). \end{itemize} If there is no such aggregate element $e_i$ for any $x_i$, the aggregate deduction candidate is not added to the set. The aggregate deduction candidate is derived as above from a hypothetical constructor $\tcode{C}(\tcode{T}_1, \dotsc, \tcode{T}_n)$, where \begin{itemize} \item if $e_i$ is of array type and $x_i$ is a \grammarterm{braced-init-list}, $\tcode{T}_i$ is an rvalue reference to the declared type of $e_i$, and \item if $e_i$ is of array type and $x_i$ is a \grammarterm{string-literal}, $\tcode{T}_i$ is an lvalue reference to the const-qualified declared type of $e_i$, and \item otherwise, $\tcode{T}_i$ is the declared type of $e_i$, \end{itemize} except that additional parameter packs of the form $\tcode{P}_j \tcode{...}$ are inserted into the parameter list in their original aggregate element position corresponding to each non-trailing aggregate element of type $\tcode{P}_j$ that was skipped because it was a parameter pack, and the trailing sequence of parameters corresponding to a trailing aggregate element that is a pack expansion (if any) is replaced by a single parameter of the form $\tcode{T}_n \tcode{...}$. In addition, if \tcode{C} is defined and inherits constructors\iref{namespace.udecl} from a direct base class denoted in the \grammarterm{base-specifier-list} by a \grammarterm{class-or-decltype} \tcode{B}, let \tcode{A} be an alias template whose template parameter list is that of \tcode{C} and whose \grammarterm{defining-type-id} is \tcode{B}. %% FIXME: the following sentence is very hard to follow; rewrite! If \tcode{A} is a deducible template\iref{dcl.type.simple}, the set contains the guides of \tcode{A} with the return type \tcode{R} of each guide replaced with \tcode{typename CC::type} given a class template \begin{codeblock} template class CC; \end{codeblock} whose primary template is not defined and with a single partial specialization whose template parameter list is that of \tcode{A} and whose template argument list is a specialization of \tcode{A} with the template argument list of \tcode{A}\iref{temp.dep.type} having a member typedef \tcode{type} designating a template specialization with the template argument list of \tcode{A} but with \tcode{C} as the template. \begin{note} Equivalently, the template parameter list of the specialization is that of \tcode{C}, the template argument list of the specialization is \tcode{B}, and the member typedef names \tcode{C} with the template argument list of \tcode{C}. \end{note} \pnum \begin{example} \begin{codeblock} template struct B { B(T); }; template struct C : public B { using B::B; }; template struct D : public B {}; C c(42); // OK, deduces \tcode{C} D d(42); // error: deduction failed, no inherited deduction guides B(int) -> B; C c2(42); // OK, deduces \tcode{C} template struct E : public B { using B::B; }; E e(42); // error: deduction failed, arguments of \tcode{E} cannot be deduced from introduced guides template struct F { F(T, U, V); }; template struct G : F { using G::F::F; } G g(true, 'a', 1); // OK, deduces \tcode{G} template struct H { T array[N]; }; template struct I { volatile T array[N]; }; template struct J { unsigned char array[N]; }; H h = { "abc" }; // OK, deduces \tcode{H} (not \tcode{T = const char}) I i = { "def" }; // OK, deduces \tcode{I} J j = { "ghi" }; // error: cannot bind reference to array of \tcode{unsigned char} to array of \tcode{char} in deduction \end{codeblock} \end{example} \pnum When resolving a placeholder for a deduced class type where the \grammarterm{template-name} designates a type template template parameter \tcode{P}, the type template template argument for \tcode{P} shall be a deducible template. Let \tcode{A} be an alias template whose template parameter list is that of \tcode{P} and whose \grammarterm{defining-type-id} is a \grammarterm{simple-template-id} whose \grammarterm{template-name} designates the type template template argument and whose \grammarterm{template-argument-list} is the template argument list of \tcode{P}. The alias template \tcode{A} is then used instead of the original \grammarterm{template-name} to resolve the placeholder. \pnum When resolving a placeholder for a deduced class type\iref{dcl.type.simple} where the \grammarterm{template-name} or \grammarterm{splice-type-specifier} designates an alias template \tcode{A}, the \grammarterm{defining-type-id} of \tcode{A} must be of the form \begin{ncsimplebnf} \opt{\keyword{typename}} \opt{nested-name-specifier} \opt{\keyword{template}} simple-template-id \end{ncsimplebnf} as specified in \ref{dcl.type.simple}. The guides of \tcode{A} are the set of functions or function templates formed as follows. For each function or function template \tcode{f} in the guides of the template named by the \grammarterm{simple-template-id} of the \grammarterm{defining-type-id}, the template arguments of the return type of \tcode{f} are deduced from the \grammarterm{defining-type-id} of \tcode{A} according to the process in \ref{temp.deduct.type} with the exception that deduction does not fail if not all template arguments are deduced. If deduction fails for another reason, proceed with an empty set of deduced template arguments. Let \tcode{g} denote the result of substituting these deductions into \tcode{f}. If substitution succeeds, form a function or function template \tcode{f'} with the following properties and add it to the set of guides of \tcode{A}: \begin{itemize} \item The function type of \tcode{f'} is the function type of \tcode{g}. \item If \tcode{f} is a function template, \tcode{f'} is a function template whose template parameter list consists of all the template parameters of \tcode{A} (including their default template arguments) that appear in the above deductions or (recursively) in their default template arguments, followed by the template parameters of \tcode{f} that were not deduced (including their default template arguments), otherwise \tcode{f'} is not a function template. \item The associated constraints\iref{temp.constr.decl} are the conjunction of the associated constraints of \tcode{g} and a constraint that is satisfied if and only if the arguments of \tcode{A} are deducible (see below) from the return type. \item If \tcode{f} is a copy deduction candidate, then \tcode{f'} is considered to be so as well. \item If \tcode{f} was generated from a \grammarterm{deduction-guide}\iref{temp.deduct.guide}, then \tcode{f'} is considered to be so as well. \item The \grammarterm{explicit-specifier} of \tcode{f'} is the \grammarterm{explicit-specifier} of \tcode{g} (if any). \end{itemize} \indextext{template!deducible arguments of}% \pnum The arguments of a template \tcode{A} are said to be deducible from a type \tcode{T} if, given a class template \begin{codeblock} template class AA; \end{codeblock} with a single partial specialization whose template parameter list is that of \tcode{A} and whose template argument list is a specialization of \tcode{A} with the template argument list of \tcode{A}\iref{temp.dep.type}, \tcode{AA} matches the partial specialization. \pnum Initialization and overload resolution are performed as described in \ref{dcl.init} and \ref{over.match.ctor}, \ref{over.match.copy}, or \ref{over.match.list} (as appropriate for the type of initialization performed) for an object of a hypothetical class type, where the guides of the template named by the placeholder are considered to be the constructors of that class type for the purpose of forming an overload set, and the initializer is provided by the context in which class template argument deduction was performed. The following exceptions apply: \begin{itemize} \item The first phase in \ref{over.match.list} (considering initializer-list constructors) is omitted if the initializer list consists of a single expression of type \cv{}~\tcode{U}, where \tcode{U} is, or is derived from, a specialization of the class template directly or indirectly named by the placeholder. \item During template argument deduction for the aggregate deduction candidate, the number of elements in a trailing parameter pack is only deduced from the number of remaining function arguments if it is not otherwise deduced. \end{itemize} If the function or function template was generated from a constructor or \grammarterm{deduction-guide} that had an \grammarterm{explicit-specifier}, each such notional constructor is considered to have that same \grammarterm{explicit-specifier}. All such notional constructors are considered to be public members of the hypothetical class type. \pnum \begin{example} \begin{codeblock} template struct A { explicit A(const T&, ...) noexcept; // \#1 A(T&&, ...); // \#2 }; int i; A a1 = { i, i }; // error: explicit constructor \#1 selected in copy-list-initialization during deduction, // cannot deduce from non-forwarding rvalue reference in \#2 A a2{i, i}; // OK, \#1 deduces to \tcode{A} and also initializes A a3{0, i}; // OK, \#2 deduces to \tcode{A} and also initializes A a4 = {0, i}; // OK, \#2 deduces to \tcode{A} and also initializes template A(const T&, const T&) -> A; // \#3 template explicit A(T&&, T&&) -> A; // \#4 A a5 = {0, 1}; // error: explicit deduction guide \#4 selected in copy-list-initialization during deduction A a6{0,1}; // OK, \#4 deduces to \tcode{A} and \#2 initializes A a7 = {0, i}; // error: \#3 deduces to \tcode{A}, \#1 and \#2 declare same constructor A a8{0,i}; // error: \#3 deduces to \tcode{A}, \#1 and \#2 declare same constructor template struct B { template using TA = T; template B(U, TA); }; B b{(int*)0, (char*)0}; // OK, deduces \tcode{B} template struct S { T x; T y; }; template struct C { S s; T t; }; template struct D { S s; T t; }; C c1 = {1, 2}; // error: deduction failed C c2 = {1, 2, 3}; // error: deduction failed C c3 = {{1u, 2u}, 3}; // OK, deduces \tcode{C} D d1 = {1, 2}; // error: deduction failed D d2 = {1, 2, 3}; // OK, braces elided, deduces \tcode{D} template struct E { T t; decltype(t) t2; }; E e1 = {1, 2}; // OK, deduces \tcode{E} template struct Types {}; template struct F : Types, T... {}; struct X {}; struct Y {}; struct Z {}; struct W { operator Y(); }; F f1 = {Types{}, {}, {}}; // OK, \tcode{F} deduced F f2 = {Types{}, X{}, Y{}}; // OK, \tcode{F} deduced F f3 = {Types{}, X{}, W{}}; // error: conflicting types deduced; \tcode{\keyword{operator} Y} not considered \end{codeblock} \end{example} \pnum \begin{example} \begin{codeblock} template struct C { C(T, U); // \#1 }; template C(T, U) -> C>; // \#2 template using A = C; template using B = A; int i{}; double d{}; A a1(&i, &i); // deduces \tcode{A} A a2(i, i); // error: cannot deduce \tcode{V *} from \tcode{i} A a3(&i, &d); // error: \#1: cannot deduce \tcode{(V*, V*)} from \tcode{(int *, double *)} // \#2: cannot deduce \tcode{A} from \tcode{C} B b1(&i, &i); // deduces \tcode{B} B b2(&d, &d); // error: cannot deduce \tcode{B} from \tcode{C} \end{codeblock} Possible exposition-only implementation of the above procedure: \begin{codeblock} // The following concept ensures a specialization of \tcode{A} is deduced. template class AA; template class AA> { }; template concept deduces_A = requires { sizeof(AA); }; // \tcode{f1} is formed from the constructor \#1 of \tcode{C}, generating the following function template template auto f1(T, U) -> C; // Deducing arguments for \tcode{C} from \tcode{C} deduces \tcode{T} as \tcode{V *} and \tcode{U} as \tcode{V *}; // \tcode{f1'} is obtained by transforming \tcode{f1} as described by the above procedure. template requires deduces_A> auto f1_prime(V *, V*) -> C; // \tcode{f2} is formed from the deduction-guide \#2 of \tcode{C} template auto f2(T, U) -> C>; // Deducing arguments for \tcode{C>} from \tcode{C} deduces \tcode{T} as \tcode{V *}; // \tcode{f2'} is obtained by transforming \tcode{f2} as described by the above procedure. template requires deduces_A>> auto f2_prime(V *, U) -> C>; // The following concept ensures a specialization of \tcode{B} is deduced. template class BB; template class BB> { }; template concept deduces_B = requires { sizeof(BB); }; // The guides for \tcode{B} derived from the above \tcode{f1'} and \tcode{f2'} for \tcode{A} are as follows: template requires deduces_A> && deduces_B> auto f1_prime_for_B(W *, W *) -> C; template requires deduces_A>> && deduces_B>> auto f2_prime_for_B(W *, U) -> C>; \end{codeblock} \end{example} \indextext{overloading!argument lists|)}% \indextext{overloading!candidate functions|)} \pnum \begin{example} \begin{codeblock} template struct Y { Y(); Y(Ts ...); }; template class X> void f() { X x; // OK, deduces \tcode{Y} X x0{}; // OK, deduces \tcode{Y} X x1{1}; // OK, deduces \tcode{Y} X x2{1, 2}; // error: cannot deduce \tcode{X} from \tcode{Y} } template void f(); \end{codeblock} \end{example} \rSec2[over.match.viable]{Viable functions}% \indextext{overloading!resolution!viable functions|(} \pnum From the set of candidate functions constructed for a given context\iref{over.match.funcs}, a set of viable functions is chosen, from which the best function will be selected by comparing argument conversion sequences and associated constraints\iref{temp.constr.decl} for the best fit\iref{over.match.best}. The selection of viable functions considers associated constraints, if any, and relationships between arguments and function parameters other than the ranking of conversion sequences. \pnum \indextext{ellipsis!overload resolution and}% \indextext{default argument!overload resolution and}% First, to be a viable function, a candidate function shall have enough parameters to agree in number with the arguments in the list. \begin{itemize} \item If there are $m$ arguments in the list, all candidate functions having exactly $m$ parameters are viable. \item A candidate function having fewer than $m$ parameters is viable only if it has an ellipsis in its parameter list\iref{dcl.fct}. For the purposes of overload resolution, any argument for which there is no corresponding parameter is considered to ``match the ellipsis''\iref{over.ics.ellipsis}. \item A candidate function \tcode{C} having more than $m$ parameters is viable only if the set of scopes $G$, as defined below, is not empty. $G$ consists of every scope $X$ that satisfies all of the following: \begin{itemize} \item There is a declaration of \tcode{C}, whose host scope is $X$, considered by the overload resolution. \item For every $k^\textrm{th}$ parameter $P$ where $k$ > $m$, there is a reachable declaration, whose host scope is $X$, that specifies a default argument\iref{dcl.fct.default} for $P$. \end{itemize} If \tcode{C} is selected as the best viable function\iref{over.match.best}: \begin{itemize} \item $G$ shall contain exactly one scope (call it $S$). \item If the candidates are denoted by a \grammarterm{splice-expression}, then $S$ shall not be a block scope. \item The default arguments used in the call to \tcode{C} are the default arguments specified by the reachable declarations whose host scope is $S$. \end{itemize} For the purposes of overload resolution, the parameter list is truncated on the right, so that there are exactly $m$ parameters. \end{itemize} \begin{example} \begin{codeblock} namespace A { extern "C" void f(int, int = 5); extern "C" void f(int = 6, int); } namespace B { extern "C" void f(int, int = 7); } void use() { [:^^A::f:](3, 4); // OK, default argument was not used for viability [:^^A::f:](3); // error: default argument provided by declarations from two scopes [:^^A::f:](); // OK, default arguments provided by declarations in the scope of \tcode{A} using A::f; using B::f; f(3, 4); // OK, default argument was not used for viability f(3); // error: default argument provided by declaration from two scopes f(); // OK, default arguments provided by declarations in the scope of \tcode{A} void g(int = 8); g(); // OK [:^^g:](); // error: host scope is block scope } void h(int = 7); constexpr std::meta::info r = ^^h; void poison() { void h(int = 8); h(); // OK, calls \tcode{h(8)} [:^^h:](); // error: default argument provided by declarations from two scopes } void call_h() { [:^^h:](); // error: default argument provided by declarations from two scopes [:r:](); // error: default argument provided by declarations from two scopes } template int k(int = 3, Ts...); int i = k(); // error: no default argument for the second parameter int j = k<>(); // OK \end{codeblock} \end{example} \pnum Second, for a function to be viable, if it has associated constraints\iref{temp.constr.decl}, those constraints shall be satisfied\iref{temp.constr.constr}. \pnum Third, for \tcode{F} to be a viable function, there shall exist for each argument an implicit conversion sequence\iref{over.best.ics} that converts that argument to the corresponding parameter of \tcode{F}. If the parameter has reference type, the implicit conversion sequence includes the operation of binding the reference, and the fact that an lvalue reference to non-\keyword{const} cannot bind to an rvalue and that an rvalue reference cannot bind to an lvalue can affect the viability of the function (see~\ref{over.ics.ref}). \rSec2[over.match.best]{Best viable function}% \rSec3[over.match.best.general]{General}% \indextext{overloading!resolution!best viable function|(} \pnum \indextext{conversion!overload resolution and}% Define $\text{ICS}^i(\tcode{F})$ as the implicit conversion sequence that converts the $i^\text{th}$ argument in the list to the type of the $i^\text{th}$ parameter of viable function \tcode{F}. \ref{over.best.ics} defines the implicit conversion sequences and \ref{over.ics.rank} defines what it means for one implicit conversion sequence to be a better conversion sequence or worse conversion sequence than another. \pnum Given these definitions, a viable function $\tcode{F}_1$ is defined to be a \defnx{better}{overloading!resolution!better viable function} function than another viable function $\tcode{F}_2$ if for all arguments $i$, $\text{ICS}^i(\tcode{F}_1)$ is not a worse conversion sequence than $\text{ICS}^i(\tcode{F}_2)$, and then \begin{itemize} \item for some argument $j$, $\text{ICS}^j(\tcode{F}_1)$ is a better conversion sequence than $\text{ICS}^j(\tcode{F}_2)$, or, if not that, \item the context is an initialization by user-defined conversion (see~\ref{dcl.init}, \ref{over.match.conv}, and~\ref{over.match.ref}) and the standard conversion sequence from the result of $\tcode{F}_1$ to the destination type (i.e., the type of the entity being initialized) is a better conversion sequence than the standard conversion sequence from the result of $\tcode{F}_2$ to the destination type \begin{example} \begin{codeblock} struct A { A(); operator int(); operator double(); } a; int i = a; // \tcode{a.\keyword{operator} int()} followed by no conversion is better than // \tcode{a.\keyword{operator} double()} followed by a conversion to \tcode{int} float x = a; // ambiguous: both possibilities require conversions, // and neither is better than the other \end{codeblock} \end{example} or, if not that, \item the context is an initialization by conversion function for direct reference binding\iref{over.match.ref} of a reference to function type, the return type of $\tcode{F}_1$ is the same kind of reference (lvalue or rvalue) as the reference being initialized, and the return type of $\tcode{F}_2$ is not \begin{example} \begin{codeblock} template struct A { operator T&(); // \#1 operator T&&(); // \#2 }; typedef int Fn(); A a; Fn& lf = a; // calls \#1 Fn&& rf = a; // calls \#2 \end{codeblock} \end{example} or, if not that, \item $\tcode{F}_1$ is not a function template specialization and $\tcode{F}_2$ is a function template specialization, or, if not that, \item $\tcode{F}_1$ and $\tcode{F}_2$ are function template specializations, and the function template for $\tcode{F}_1$ is more specialized than the template for $\tcode{F}_2$ according to the partial ordering rules described in~\ref{temp.func.order}, or, if not that, \item $\tcode{F}_1$ and $\tcode{F}_2$ are non-template functions and $\tcode{F}_1$ is more partial-ordering-constrained than $\tcode{F}_2$\iref{temp.constr.order} \begin{example} \begin{codeblock} template struct S { constexpr void f(); // \#1 constexpr void f(this S&) requires true; // \#2 }; void test() { S<> s; s.f(); // calls \#2 } \end{codeblock} \end{example} or, if not that, \item $\tcode{F}_1$ is a constructor for a class \tcode{D}, $\tcode{F}_2$ is a constructor for a base class \tcode{B} of \tcode{D}, and for all arguments the corresponding parameters of $\tcode{F}_1$ and $\tcode{F}_2$ have the same type \begin{example} \begin{codeblock} struct A { A(int = 0); }; struct B: A { using A::A; B(); }; int main() { B b; // OK, \tcode{B::B()} } \end{codeblock} \end{example} or, if not that, \item $\tcode{F}_2$ is a rewritten candidate\iref{over.match.oper} and $\tcode{F}_1$ is not \begin{example} \begin{codeblock} struct S { friend auto operator<=>(const S&, const S&) = default; // \#1 friend bool operator<(const S&, const S&); // \#2 }; bool b = S() < S(); // calls \#2 \end{codeblock} \end{example} or, if not that, \item $\tcode{F}_1$ and $\tcode{F}_2$ are rewritten candidates, and $\tcode{F}_2$ is a synthesized candidate with reversed order of parameters and $\tcode{F}_1$ is not \begin{example} \begin{codeblock} struct S { friend std::weak_ordering operator<=>(const S&, int); // \#1 friend std::weak_ordering operator<=>(int, const S&); // \#2 }; bool b = 1 < S(); // calls \#2 \end{codeblock} \end{example} or, if not that, \item $\tcode{F}_1$ and $\tcode{F}_2$ are generated from class template argument deduction\iref{over.match.class.deduct} for a class \tcode{D}, and $\tcode{F}_2$ is generated from inheriting constructors from a base class of \tcode{D} while $\tcode{F}_1$ is not, and for each explicit function argument, the corresponding parameters of $\tcode{F}_1$ and $\tcode{F}_2$ are either both ellipses or have the same type, or, if not that, \item $\tcode{F}_1$ is generated from a \grammarterm{deduction-guide}\iref{over.match.class.deduct} and $\tcode{F}_2$ is not, or, if not that, \item $\tcode{F}_1$ is the copy deduction candidate\iref{over.match.class.deduct} and $\tcode{F}_2$ is not, or, if not that, \item $\tcode{F}_1$ is generated from a non-template constructor and $\tcode{F}_2$ is generated from a constructor template. \begin{example} \begin{codeblock} template struct A { using value_type = T; A(value_type); // \#1 A(const A&); // \#2 A(T, T, int); // \#3 template A(int, T, U); // \#4 // \#5 is the copy deduction candidate, \tcode{A(A)} }; A x(1, 2, 3); // uses \#3, generated from a non-template constructor template A(T) -> A; // \#6, less specialized than \#5 A a(42); // uses \#6 to deduce \tcode{A} and \#1 to initialize A b = a; // uses \#5 to deduce \tcode{A} and \#2 to initialize template A(A) -> A>; // \#7, as specialized as \#5 A b2 = a; // uses \#7 to deduce \tcode{A>} and \#1 to initialize \end{codeblock} \end{example} \end{itemize} \pnum If there is exactly one viable function that is a better function than all other viable functions, then it is the one selected by overload resolution; otherwise the call is ill-formed. \begin{footnote} The algorithm for selecting the best viable function is linear in the number of viable functions. Run a simple tournament to find a function \tcode{W} that is not worse than any opponent it faced. Although it is possible that another function \tcode{F} that \tcode{W} did not face is at least as good as \tcode{W}, \tcode{F} cannot be the best function because at some point in the tournament \tcode{F} encountered another function \tcode{G} such that \tcode{F} was not better than \tcode{G}. Hence, either \tcode{W} is the best function or there is no best function. So, make a second pass over the viable functions to verify that \tcode{W} is better than all other functions. \end{footnote} \begin{example} \begin{codeblock} void Fcn(const int*, short); void Fcn(int*, int); int i; short s = 0; void f() { Fcn(&i, s); // is ambiguous because \tcode{\&i} $\to$ \tcode{int*} is better than \tcode{\&i} $\to$ \tcode{const int*} // but \tcode{s} $\to$ \tcode{short} is also better than \tcode{s} $\to$ \tcode{int} Fcn(&i, 1L); // calls \tcode{Fcn(int*, int)}, because \tcode{\&i} $\to$ \tcode{int*} is better than \tcode{\&i} $\to$ \tcode{const int*} // and \tcode{1L} $\to$ \tcode{short} and \tcode{1L} $\to$ \tcode{int} are indistinguishable Fcn(&i, 'c'); // calls \tcode{Fcn(int*, int)}, because \tcode{\&i} $\to$ \tcode{int*} is better than \tcode{\&i} $\to$ \tcode{const int*} // and \tcode{'c'} $\to$ \tcode{int} is better than \tcode{'c'} $\to$ \tcode{short} } \end{codeblock} \end{example} \pnum \begin{note} If the best viable function was made viable by one or more default arguments, additional requirements apply\iref{over.match.viable}. \end{note} \rSec3[over.best.ics]{Implicit conversion sequences}% \rSec4[over.best.ics.general]{General}% \indextext{overloading!resolution!implicit conversions and|(} \indextext{implicit conversion sequence|see{conversion sequence, implicit}} \pnum An \defnadj{implicit}{conversion sequence} is a sequence of conversions used to convert an argument in a function call to the type of the corresponding parameter of the function being called. The sequence of conversions is an implicit conversion as defined in \ref{conv}, which means it is governed by the rules for initialization of an object or reference by a single expression\iref{dcl.init,dcl.init.ref}. \pnum Implicit conversion sequences are concerned only with the type, cv-qualification, and value category of the argument and how these are converted to match the corresponding properties of the parameter. \begin{note} Other properties, such as the lifetime, storage duration, linkage, alignment, accessibility of the argument, whether the argument is a bit-field, and whether a function is deleted\iref{dcl.fct.def.delete}, are ignored. So, although an implicit conversion sequence can be defined for a given argument-parameter pair, the conversion from the argument to the parameter might still be ill-formed in the final analysis. \end{note} \pnum A well-formed implicit conversion sequence is one of the following forms: \begin{itemize} \item a standard conversion sequence\iref{over.ics.scs}, \item a user-defined conversion sequence\iref{over.ics.user}, or \item an ellipsis conversion sequence\iref{over.ics.ellipsis}. \end{itemize} \pnum However, if the target is \begin{itemize} \item the first parameter of a constructor or \item the object parameter of a user-defined conversion function \end{itemize} and the constructor or user-defined conversion function is a candidate by \begin{itemize} \item \ref{over.match.ctor}, when the argument is the temporary in the second step of a class copy-initialization, \item \ref{over.match.copy}, \ref{over.match.conv}, or \ref{over.match.ref} (in all cases), or \item the second phase of \ref{over.match.list} when the initializer list has exactly one element that is itself an initializer list, and the target is the first parameter of a constructor of class \tcode{X}, and the conversion is to \tcode{X} or reference to \cv{}~\tcode{X}, \end{itemize} user-defined conversion sequences are not considered. \begin{note} These rules prevent more than one user-defined conversion from being applied during overload resolution, thereby avoiding infinite recursion. \end{note} \begin{example} \begin{codeblock} struct Y { Y(int); }; struct A { operator int(); }; Y y1 = A(); // error: \tcode{A::\keyword{operator} int()} is not a candidate struct X { X(); }; struct B { operator X(); }; B b; X x{{b}}; // error: \tcode{B::\keyword{operator} X()} is not a candidate \end{codeblock} \end{example} \pnum For the case where the parameter type is a reference, see~\ref{over.ics.ref}. \pnum When the parameter type is not a reference, the implicit conversion sequence models a copy-initialization of the parameter from the argument expression. The implicit conversion sequence is the one required to convert the argument expression to a prvalue of the type of the parameter. \begin{note} When the parameter has a class type, this is a conceptual conversion defined for the purposes of \ref{over}; the actual initialization is defined in terms of constructors and is not a conversion. \end{note} \pnum When the cv-unqualified version of the type of the argument expression is the same as the parameter type, the implicit conversion sequence is an identity conversion. When the parameter has a class type and the argument expression has a (possibly cv-qualified) derived class type, the implicit conversion sequence is a derived-to-base \indextext{conversion!derived-to-base}% conversion from the derived class to the base class. A derived-to-base conversion has Conversion rank\iref{over.ics.scs}. \begin{note} There is no such standard conversion; this derived-to-base conversion exists only in the description of implicit conversion sequences. \end{note} \begin{example} An implicit conversion sequence from an argument of type \tcode{const A} to a parameter of type \tcode{A} can be formed, even if overload resolution for copy-initialization of \tcode{A} from the argument would not find a viable function\iref{over.match.ctor,over.match.viable}. The implicit conversion sequence for that case is the identity sequence; it contains no ``conversion'' from \tcode{const A} to \tcode{A}. \end{example} \pnum When the parameter is the implicit object parameter of a static member function, the implicit conversion sequence is a standard conversion sequence that is neither better nor worse than any other standard conversion sequence. \pnum In all contexts, when converting to the implicit object parameter or when converting to the left operand of an assignment operation only standard conversion sequences are allowed. \begin{note} When a conversion to the explicit object parameter occurs, it can include user-defined conversion sequences. \end{note} \pnum If no conversions are required to match an argument to a parameter type, the implicit conversion sequence is the standard conversion sequence consisting of the identity conversion\iref{over.ics.scs}. \pnum If no sequence of conversions can be found to convert an argument to a parameter type, an implicit conversion sequence cannot be formed. \pnum If there are multiple well-formed implicit conversion sequences converting the argument to the parameter type, the implicit conversion sequence associated with the parameter is defined to be the unique conversion sequence designated the \defnadj{ambiguous}{conversion sequence}. For the purpose of ranking implicit conversion sequences as described in~\ref{over.ics.rank}, the ambiguous conversion sequence is treated as a user-defined conversion sequence that is indistinguishable from any other user-defined conversion sequence. \begin{note} This rule prevents a function from becoming non-viable because of an ambiguous conversion sequence for one of its parameters. \begin{example} \begin{codeblock} class B; class A { A (B&);}; class B { operator A (); }; class C { C (B&); }; void f(A) { } void f(C) { } B b; f(b); // error: ambiguous because there is a conversion \tcode{b} $\to$ \tcode{C} (via constructor) // and an (ambiguous) conversion \tcode{b} $\to$ \tcode{A} (via constructor or conversion function) void f(B) { } f(b); // OK, unambiguous \end{codeblock} \end{example} \end{note} If a function that uses the ambiguous conversion sequence is selected as the best viable function, the call will be ill-formed because the conversion of one of the arguments in the call is ambiguous. \pnum The three forms of implicit conversion sequences mentioned above are defined in the following subclauses. \rSec4[over.ics.scs]{Standard conversion sequences} \pnum \tref{over.ics.scs} summarizes the conversions defined in \ref{conv} and partitions them into four disjoint categories: Lvalue Transformation, Qualification Adjustment, Promotion, and Conversion. \begin{note} These categories are orthogonal with respect to value category, cv-qualification, and data representation: the Lvalue Transformations do not change the cv-qualification or data representation of the type; the Qualification Adjustments do not change the value category or data representation of the type; and the Promotions and Conversions do not change the value category or cv-qualification of the type. \end{note} \pnum \begin{note} As described in \ref{conv}, a standard conversion sequence either is the Identity conversion by itself (that is, no conversion) or consists of one to three conversions from the other four categories. If there are two or more conversions in the sequence, the conversions are applied in the canonical order: \textbf{Lvalue Transformation}, \textbf{Promotion} or \textbf{Conversion}, \textbf{Qualification Adjustment}. \end{note} \pnum \indextext{conversion rank}% Each conversion in \tref{over.ics.scs} also has an associated rank (Exact Match, Promotion, or Conversion). These are used to rank standard conversion sequences\iref{over.ics.rank}. The rank of a conversion sequence is determined by considering the rank of each conversion in the sequence and the rank of any reference binding\iref{over.ics.ref}. If any of those has Conversion rank, the sequence has Conversion rank; otherwise, if any of those has Promotion rank, the sequence has Promotion rank; otherwise, the sequence has Exact Match rank. \begin{floattable}{Conversions}{over.ics.scs}{l|c|c|c} \topline \hdstyle{Conversion} & \hdstyle{Category} & \hdstyle{Rank} & \hdstyle{Subclause} \\ \capsep No conversions required & Identity & & \\ \cline{1-2}\cline{4-4} Lvalue-to-rvalue conversion & & & \ref{conv.lval} \\ \cline{1-1}\cline{4-4} Array-to-pointer conversion & Lvalue Transformation & & \ref{conv.array} \\ \cline{1-1}\cline{4-4} Function-to-pointer conversion & & \rb{Exact Match}& \ref{conv.func} \\ \cline{1-2}\cline{4-4} Qualification conversions & & & \ref{conv.qual} \\ \cline{1-1}\cline{4-4} Function pointer conversion & \rb{Qualification Adjustment} & & \ref{conv.fctptr} \\ \hline Integral promotions & & & \ref{conv.prom} \\ \cline{1-1}\cline{4-4} Floating-point promotion & \rb{Promotion} & \rb{Promotion} & \ref{conv.fpprom} \\ \hline Integral conversions & & & \ref{conv.integral} \\ \cline{1-1}\cline{4-4} Floating-point conversions & & & \ref{conv.double} \\ \cline{1-1}\cline{4-4} Floating-integral conversions & & & \ref{conv.fpint} \\ \cline{1-1}\cline{4-4} Pointer conversions & \rb{Conversion} & \rb{Conversion} & \ref{conv.ptr} \\ \cline{1-1}\cline{4-4} Pointer-to-member conversions & & & \ref{conv.mem} \\ \cline{1-1}\cline{4-4} Boolean conversions & & & \ref{conv.bool} \\ \end{floattable} \rSec4[over.ics.user]{User-defined conversion sequences} \pnum A \defnadj{user-defined}{conversion sequence} consists of an initial standard conversion sequence followed by a user-defined conversion\iref{class.conv} followed by a second standard conversion sequence. If the user-defined conversion is specified by a constructor\iref{class.conv.ctor}, the initial standard conversion sequence converts the source type to the type of the first parameter of that constructor. If the user-defined conversion is specified by a conversion function\iref{class.conv.fct}, the initial standard conversion sequence converts the source type to the type of the object parameter of that conversion function. \pnum The second standard conversion sequence converts the result of the user-defined conversion to the target type for the sequence; any reference binding is included in the second standard conversion sequence. Since an implicit conversion sequence is an initialization, the special rules for initialization by user-defined conversion apply when selecting the best user-defined conversion for a user-defined conversion sequence (see~\ref{over.match.best} and~\ref{over.best.ics}). \pnum If the user-defined conversion is specified by a specialization of a conversion function template, the second standard conversion sequence shall have Exact Match rank. \pnum A conversion of an expression of class type to the same class type is given Exact Match rank, and a conversion of an expression of class type to a base class of that type is given Conversion rank, in spite of the fact that a constructor (i.e., a user-defined conversion function) is called for those cases. \rSec4[over.ics.ellipsis]{Ellipsis conversion sequences} \pnum \indextext{ellipsis!conversion sequence}% An ellipsis conversion sequence occurs when an argument in a function call is matched with the ellipsis parameter specification of the function called (see~\ref{expr.call}). \rSec4[over.ics.ref]{Reference binding} \pnum When a parameter of type ``reference to \cv~\tcode{T}'' binds directly\iref{dcl.init.ref} to an argument expression: \begin{itemize} \item If the argument expression has a type \tcode{D} that is a derived class of \tcode{T}, the implicit conversion sequence is a derived-to-base conversion from \tcode{D} to \tcode{T}\iref{over.best.ics}. \item Otherwise, if the type of the argument is possibly cv-qualified \tcode{T}, or if \tcode{T} is an array type of unknown bound with element type \tcode{U} and the argument has an array type of known bound whose element type is possibly cv-qualified \tcode{U}, the implicit conversion sequence is the identity conversion. \item Otherwise, if \tcode{T} is a function type, the implicit conversion sequence is a function pointer conversion. \item Otherwise, the implicit conversion sequence is a qualification conversion. \end{itemize} \begin{example} \begin{codeblock} struct A {}; struct B : public A {} b; int f(A&); int f(B&); int i = f(b); // calls \tcode{f(B\&)}, an exact match, rather than \tcode{f(A\&)}, a conversion void g() noexcept; int h(void (&)() noexcept); // \#1 int h(void (&)()); // \#2 int j = h(g); // calls \#1, an exact match, rather than \#2, a function pointer conversion \end{codeblock} \end{example} If the parameter binds directly to the result of applying a conversion function to the argument expression, the implicit conversion sequence is a user-defined conversion sequence\iref{over.ics.user} whose second standard conversion sequence is determined by the above rules. \pnum When a parameter of reference type is not bound directly to an argument expression, the conversion sequence is the one required to convert the argument expression to the referenced type according to~\ref{over.best.ics}. Conceptually, this conversion sequence corresponds to copy-initializing a temporary of the referenced type with the argument expression. Any difference in top-level cv-qualification is subsumed by the initialization itself and does not constitute a conversion. \pnum Except for an implicit object parameter, for which see~\ref{over.match.funcs}, an implicit conversion sequence cannot be formed if it requires binding an lvalue reference other than a reference to a non-volatile \keyword{const} type to an rvalue or binding an rvalue reference to an lvalue of object type. \begin{note} This means, for example, that a candidate function cannot be a viable function if it has a non-\keyword{const} lvalue reference parameter (other than the implicit object parameter) and the corresponding argument would require a temporary to be created to initialize the lvalue reference (see~\ref{dcl.init.ref}). \end{note} \pnum Other restrictions on binding a reference to a particular argument that are not based on the types of the reference and the argument do not affect the formation of an implicit conversion sequence, however. \begin{example} A function with an ``lvalue reference to \tcode{int}'' parameter can be a viable candidate even if the corresponding argument is an \tcode{int} bit-field. The formation of implicit conversion sequences treats the \tcode{int} bit-field as an \tcode{int} lvalue and finds an exact match with the parameter. If the function is selected by overload resolution, the call will nonetheless be ill-formed because of the prohibition on binding a non-\keyword{const} lvalue reference to a bit-field\iref{dcl.init.ref}. \end{example} \rSec4[over.ics.list]{List-initialization sequence} \pnum When an argument is an initializer list\iref{dcl.init.list}, it is not an expression and special rules apply for converting it to a parameter type. \pnum If the initializer list is a \grammarterm{designated-initializer-list} and the parameter is not a reference, a conversion is only possible if the parameter has an aggregate type that can be initialized from the initializer list according to the rules for aggregate initialization\iref{dcl.init.aggr}, in which case the implicit conversion sequence is a user-defined conversion sequence whose second standard conversion sequence is an identity conversion. \begin{note} Aggregate initialization does not require that the members are declared in designation order. If, after overload resolution, the order does not match for the selected overload, the initialization of the parameter will be ill-formed\iref{dcl.init.list}. \begin{example} \begin{codeblock} struct A { int x, y; }; struct B { int y, x; }; void f(A a, int); // \#1 void f(B b, ...); // \#2 void g(A a); // \#3 void g(B b); // \#4 void h() { f({.x = 1, .y = 2}, 0); // OK; calls \#1 f({.y = 2, .x = 1}, 0); // error: selects \#1, initialization of \tcode{a} fails // due to non-matching member order\iref{dcl.init.list} g({.x = 1, .y = 2}); // error: ambiguous between \#3 and \#4 } \end{codeblock} \end{example} \end{note} \pnum Otherwise, if the parameter type is an aggregate class \tcode{X} and the initializer list has a single element of type \cv{}~\tcode{U}, where \tcode{U} is \tcode{X} or a class derived from \tcode{X}, the implicit conversion sequence is the one required to convert the element to the parameter type. \pnum Otherwise, if the parameter type is a character array \begin{footnote} Since there are no parameters of array type, this will only occur as the referenced type of a reference parameter. \end{footnote} and the initializer list has a single element that is an appropriately-typed \grammarterm{string-literal}\iref{dcl.init.string}, the implicit conversion sequence is the identity conversion. \pnum Otherwise, if the parameter type is \tcode{std::initializer_list} and all the elements of the initializer list can be implicitly converted to \tcode{X}, the implicit conversion sequence is the worst conversion necessary to convert an element of the list to \tcode{X}, or if the initializer list has no elements, the identity conversion. This conversion can be a user-defined conversion even in the context of a call to an initializer-list constructor. \begin{example} \begin{codeblock} void f(std::initializer_list); f( {} ); // OK, \tcode{f(initializer_list)} identity conversion f( {1,2,3} ); // OK, \tcode{f(initializer_list)} identity conversion f( {'a','b'} ); // OK, \tcode{f(initializer_list)} integral promotion f( {1.0} ); // error: narrowing struct A { A(std::initializer_list); // \#1 A(std::initializer_list>); // \#2 A(std::initializer_list); // \#3 }; A a{ 1.0,2.0 }; // OK, uses \#1 void g(A); g({ "foo", "bar" }); // OK, uses \#3 typedef int IA[3]; void h(const IA&); h({ 1, 2, 3 }); // OK, identity conversion \end{codeblock} \end{example} \pnum Otherwise, if the parameter type is ``array of \tcode{N} \tcode{X}'' or ``array of unknown bound of \tcode{X}'', if there exists an implicit conversion sequence from each element of the initializer list (and from \tcode{\{\}} in the former case if \tcode{N} exceeds the number of elements in the initializer list) to \tcode{X}, the implicit conversion sequence is the worst such implicit conversion sequence. \pnum Otherwise, if the parameter is a non-aggregate class \tcode{X} and overload resolution per~\ref{over.match.list} chooses a single best constructor \tcode{C} of \tcode{X} to perform the initialization of an object of type \tcode{X} from the argument initializer list: \begin{itemize} \item If \tcode{C} is not an initializer-list constructor and the initializer list has a single element of type \cv{}~\tcode{U}, where \tcode{U} is \tcode{X} or a class derived from \tcode{X}, the implicit conversion sequence has Exact Match rank if \tcode{U} is \tcode{X}, or Conversion rank if \tcode{U} is derived from \tcode{X}. \item Otherwise, the implicit conversion sequence is a user-defined conversion sequence whose second standard conversion sequence is an identity conversion. \end{itemize} If multiple constructors are viable but none is better than the others, the implicit conversion sequence is the ambiguous conversion sequence. User-defined conversions are allowed for conversion of the initializer list elements to the constructor parameter types except as noted in~\ref{over.best.ics}. \begin{example} \begin{codeblock} struct A { A(std::initializer_list); }; void f(A); f( {'a', 'b'} ); // OK, \tcode{f(A(std::initializer_list))} user-defined conversion struct B { B(int, double); }; void g(B); g( {'a', 'b'} ); // OK, \tcode{g(B(int, double))} user-defined conversion g( {1.0, 1.0} ); // error: narrowing void f(B); f( {'a', 'b'} ); // error: ambiguous \tcode{f(A)} or \tcode{f(B)} struct C { C(std::string); }; void h(C); h({"foo"}); // OK, \tcode{h(C(std::string("foo")))} struct D { D(A, C); }; void i(D); i({ {1,2}, {"bar"} }); // OK, \tcode{i(D(A(std::initializer_list\{1,2\}), C(std::string("bar"))))} \end{codeblock} \end{example} \pnum Otherwise, if the parameter has an aggregate type which can be initialized from the initializer list according to the rules for aggregate initialization\iref{dcl.init.aggr}, the implicit conversion sequence is a user-defined conversion sequence whose second standard conversion sequence is an identity conversion. \begin{example} \begin{codeblock} struct A { int m1; double m2; }; void f(A); f( {'a', 'b'} ); // OK, \tcode{f(A(int,double))} user-defined conversion f( {1.0} ); // error: narrowing \end{codeblock} \end{example} \pnum Otherwise, if the parameter is a reference, see~\ref{over.ics.ref}. \begin{note} The rules in this subclause will apply for initializing the underlying temporary for the reference. \end{note} \begin{example} \begin{codeblock} struct A { int m1; double m2; }; void f(const A&); f( {'a', 'b'} ); // OK, \tcode{f(A(int,double))} user-defined conversion f( {1.0} ); // error: narrowing void g(const double &); g({1}); // same conversion as \tcode{int} to \tcode{double} \end{codeblock} \end{example} \pnum Otherwise, if the parameter type is not a class: \begin{itemize} \item if the initializer list has one element that is not itself an initializer list, the implicit conversion sequence is the one required to convert the element to the parameter type; \begin{example} \begin{codeblock} void f(int); f( {'a'} ); // OK, same conversion as \tcode{char} to \tcode{int} f( {1.0} ); // error: narrowing \end{codeblock} \end{example} \item if the initializer list has no elements, the implicit conversion sequence is the identity conversion. \begin{example} \begin{codeblock} void f(int); f( { } ); // OK, identity conversion \end{codeblock} \end{example} \end{itemize} \pnum In all cases other than those enumerated above, no conversion is possible. \rSec3[over.ics.rank]{Ranking implicit conversion sequences} \pnum This subclause defines a partial ordering of implicit conversion sequences based on the relationships \defnadj{better}{conversion sequence} and \defnadj{better}{conversion}. If an implicit conversion sequence S1 is defined by these rules to be a better conversion sequence than S2, then it is also the case that S2 is a \defnadj{worse}{conversion sequence} than S1. If conversion sequence S1 is neither better than nor worse than conversion sequence S2, S1 and S2 are said to be \defnx{indistinguishable conversion sequences}{conversion sequence!indistinguishable}. \pnum When comparing the basic forms of implicit conversion sequences (as defined in~\ref{over.best.ics}) \begin{itemize} \item a standard conversion sequence\iref{over.ics.scs} is a better conversion sequence than a user-defined conversion sequence or an ellipsis conversion sequence, and \item a user-defined conversion sequence\iref{over.ics.user} is a better conversion sequence than an ellipsis conversion sequence\iref{over.ics.ellipsis}. \end{itemize} \pnum Two implicit conversion sequences of the same form are indistinguishable conversion sequences unless one of the following rules applies: \begin{itemize} \item List-initialization sequence \tcode{L1} is a better conversion sequence than list-initialization sequence \tcode{L2} if \begin{itemize} \item \tcode{L1} converts to \tcode{std::initializer_list} for some \tcode{X} and \tcode{L2} does not, or, if not that, \item \tcode{L1} and \tcode{L2} convert to arrays of the same element type, and either the number of elements $n_1$ initialized by \tcode{L1} is less than the number of elements $n_2$ initialized by \tcode{L2}, or $n_1 = n_2$ and \tcode{L2} converts to an array of unknown bound and \tcode{L1} does not, \end{itemize} even if one of the other rules in this paragraph would otherwise apply. \begin{example} \begin{codeblock} void f1(int); // \#1 void f1(std::initializer_list); // \#2 void g1() { f1({42}); } // chooses \#2 void f2(std::pair); // \#3 void f2(std::initializer_list); // \#4 void g2() { f2({"foo","bar"}); } // chooses \#4 \end{codeblock} \end{example} \begin{example} \begin{codeblock} void f(int (&&)[] ); // \#1 void f(double (&&)[] ); // \#2 void f(int (&&)[2]); // \#3 f( {1} ); // Calls \#1: Better than \#2 due to conversion, better than \#3 due to bounds f( {1.0} ); // Calls \#2: Identity conversion is better than floating-integral conversion f( {1.0, 2.0} ); // Calls \#2: Identity conversion is better than floating-integral conversion f( {1, 2} ); // Calls \#3: Converting to array of known bound is better than to unknown bound, // and an identity conversion is better than floating-integral conversion \end{codeblock} \end{example} \item Standard conversion sequence \tcode{S1} is a better conversion sequence than standard conversion sequence \tcode{S2} if \begin{itemize} \item \indextext{subsequence rule!overloading}% \tcode{S1} is a proper subsequence of \tcode{S2} (comparing the conversion sequences in the canonical form defined by~\ref{over.ics.scs}, excluding any Lvalue Transformation; the identity conversion sequence is considered to be a subsequence of any non-identity conversion sequence) or, if not that, \item the rank of \tcode{S1} is better than the rank of \tcode{S2}, or \tcode{S1} and \tcode{S2} have the same rank and are distinguishable by the rules in the paragraph below, or, if not that, \item \tcode{S1} and \tcode{S2} include reference bindings\iref{dcl.init.ref} and neither refers to an implicit object parameter of a non-static member function declared without a \grammarterm{ref-qualifier}, and \tcode{S1} binds an rvalue reference to an rvalue and \tcode{S2} binds an lvalue reference \begin{example} \begin{codeblock} int i; int f1(); int&& f2(); int g(const int&); int g(const int&&); int j = g(i); // calls \tcode{g(const int\&)} int k = g(f1()); // calls \tcode{g(const int\&\&)} int l = g(f2()); // calls \tcode{g(const int\&\&)} struct A { A& operator<<(int); void p() &; void p() &&; }; A& operator<<(A&&, char); A() << 1; // calls \tcode{A::\keyword{operator}<<(int)} A() << 'c'; // calls \tcode{\keyword{operator}<<(A\&\&, char)} A a; a << 1; // calls \tcode{A::\keyword{operator}<<(int)} a << 'c'; // calls \tcode{A::\keyword{operator}<<(int)} A().p(); // calls \tcode{A::p()\&\&} a.p(); // calls \tcode{A::p()\&} \end{codeblock} \end{example} or, if not that, \item \tcode{S1} and \tcode{S2} include reference bindings\iref{dcl.init.ref} and \tcode{S1} binds an lvalue reference to an lvalue of function type and \tcode{S2} binds an rvalue reference to an lvalue of function type \begin{example} \begin{codeblock} int f(void(&)()); // \#1 int f(void(&&)()); // \#2 void g(); int i1 = f(g); // calls \#1 \end{codeblock} \end{example} or, if not that, \item \tcode{S1} and \tcode{S2} differ only in their qualification conversion\iref{conv.qual} and yield similar types \tcode{T1} and \tcode{T2}, respectively (where a standard conversion sequence that is a reference binding is considered to yield the cv-unqualified referenced type), where \tcode{T1} and \tcode{T2} are not the same type, and \tcode{const T2} is reference-compatible with \tcode{T1}\iref{dcl.init.ref} \begin{example} \begin{codeblock} int f(const volatile int *); int f(const int *); int i; int j = f(&i); // calls \tcode{f(const int*)} int g(const int*); int g(const volatile int* const&); int* p; int k = g(p); // calls \tcode{g(const int*)} \end{codeblock} \end{example} or, if not that, \item \tcode{S1} and \tcode{S2} bind ``reference to \tcode{T1}'' and ``reference to \tcode{T2}'', respectively\iref{dcl.init.ref}, where \tcode{T1} and \tcode{T2} are not the same type, and \tcode{T2} is reference-compatible with \tcode{T1} \begin{example} \begin{codeblock} int f(const int &); int f(int &); int g(const int &); int g(int); int i; int j = f(i); // calls \tcode{f(int \&)} int k = g(i); // ambiguous struct X { void f() const; void f(); }; void g(const X& a, X b) { a.f(); // calls \tcode{X::f() const} b.f(); // calls \tcode{X::f()} } int h(int (&)[]); int h(int (&)[1]); void g2() { int a[1]; h(a); // calls \tcode{h(int (\&)[1])} } \end{codeblock} \end{example} or, if not that, \item \tcode{S1} and \tcode{S2} bind the same reference type ``reference to \tcode{T}'' and have source types \tcode{V1} and \tcode{V2}, respectively, where the standard conversion sequence from \tcode{V1*} to \tcode{T*} is better than the standard conversion sequence from \tcode{V2*} to \tcode{T*}. \begin{example} \begin{codeblock} struct Z {}; struct A { operator Z&(); operator const Z&(); // \#1 }; struct B { operator Z(); operator const Z&&(); // \#2 }; const Z& r1 = A(); // OK, uses \#1 const Z&& r2 = B(); // OK, uses \#2 \end{codeblock} \end{example} \end{itemize} \item User-defined conversion sequence \tcode{U1} is a better conversion sequence than another user-defined conversion sequence \tcode{U2} if they contain the same user-defined conversion function or constructor or they initialize the same class in an aggregate initialization and in either case the second standard conversion sequence of \tcode{U1} is better than the second standard conversion sequence of \tcode{U2}. \begin{example} \begin{codeblock} struct A { operator short(); } a; int f(int); int f(float); int i = f(a); // calls \tcode{f(int)}, because \tcode{short} $\to$ \tcode{int} is // better than \tcode{short} $\to$ \tcode{float}. \end{codeblock} \end{example} \end{itemize} \pnum Standard conversion sequences are ordered by their ranks: an Exact Match is a better conversion than a Promotion, which is a better conversion than a Conversion. Two conversion sequences with the same rank are indistinguishable unless one of the following rules applies: \begin{itemize} \item A conversion that does not convert a pointer or a pointer to member to \tcode{bool} is better than one that does. \item A conversion that promotes an enumeration whose underlying type is fixed to its underlying type is better than one that promotes to the promoted underlying type, if the two are different. \item A conversion in either direction between floating-point type \tcode{FP1} and floating-point type \tcode{FP2} is better than a conversion in the same direction between \tcode{FP1} and arithmetic type \tcode{T3} if \begin{itemize} \item the floating-point conversion rank\iref{conv.rank} of \tcode{FP1} is equal to the rank of \tcode{FP2}, and \item \tcode{T3} is not a floating-point type, or \tcode{T3} is a floating-point type whose rank is not equal to the rank of \tcode{FP1}, or the floating-point conversion subrank\iref{conv.rank} of \tcode{FP2} is greater than the subrank of \tcode{T3}. \begin{example} \begin{codeblock} int f(std::float32_t); int f(std::float64_t); int f(long long); float x; std::float16_t y; int i = f(x); // calls \tcode{f(std::float32_t)} on implementations where // \tcode{float} and \tcode{std::float32_t} have equal conversion ranks int j = f(y); // error: ambiguous, no equal conversion rank \end{codeblock} \end{example} \end{itemize} \item If class \tcode{B} is derived directly or indirectly from class \tcode{A}, conversion of \tcode{B*} to \tcode{A*} is better than conversion of \tcode{B*} to \tcode{\keyword{void}*}, and conversion of \tcode{A*} to \tcode{\keyword{void}*} is better than conversion of \tcode{B*} to \tcode{\keyword{void}*}. \item If class \tcode{B} is derived directly or indirectly from class \tcode{A} and class \tcode{C} is derived directly or indirectly from \tcode{B}, \begin{itemize} \item conversion of \tcode{C*} to \tcode{B*} is better than conversion of \tcode{C*} to \tcode{A*}, \begin{example} \begin{codeblock} struct A {}; struct B : public A {}; struct C : public B {}; C* pc; int f(A*); int f(B*); int i = f(pc); // calls \tcode{f(B*)} \end{codeblock} \end{example} \item conversion of \tcode{A::*} to \tcode{B::*} is better than conversion of \tcode{A::*} to \tcode{C::*}, \item conversion of \tcode{C} to \tcode{B} is better than conversion of \tcode{C} to \tcode{A}, \item conversion of \tcode{B*} to \tcode{A*} is better than conversion of \tcode{C*} to \tcode{A*}, \item conversion of \tcode{B::*} to \tcode{C::*} is better than conversion of \tcode{A::*} to \tcode{C::*}, and \item conversion of \tcode{B} to \tcode{A} is better than conversion of \tcode{C} to \tcode{A}. \end{itemize} \begin{note} Compared conversion sequences will have different source types only in the context of comparing the second standard conversion sequence of an initialization by user-defined conversion (see~\ref{over.match.best}); in all other contexts, the source types will be the same and the target types will be different. \end{note} \end{itemize}% \indextext{overloading!resolution!implicit conversions and|)}% \indextext{overloading!resolution|)} \rSec1[over.over]{Address of an overload set}% \indextext{overloading!address of overloaded function}% \indextext{overloaded function!address of} \pnum An expression that designates an overload set $S$ and that appears without arguments is resolved to a function, a pointer to function, or a pointer to member function for a specific function that is chosen from a set of functions selected from $S$ determined based on the target type required in the context (if any), as described below. The target can be \begin{itemize} \item an object or reference being initialized\iref{dcl.init,dcl.init.ref,dcl.init.list}, \item the left side of an assignment\iref{expr.assign}, \item a parameter of a function\iref{expr.call}, \item a parameter of a user-defined operator\iref{over.oper}, \item the return value of a function, operator function, or conversion\iref{stmt.return}, \item an explicit type conversion\iref{expr.type.conv,expr.static.cast,expr.cast}, or \item a constant template parameter\iref{temp.arg.nontype}. \end{itemize} If the target type contains a placeholder type, placeholder type deduction is performed\iref{dcl.type.auto.deduct}, and the remainder of this subclause uses the target type so deduced. The expression can be preceded by the \tcode{\&} operator. \begin{note} Any redundant set of parentheses surrounding the function name is ignored\iref{expr.prim.paren}. \end{note} \pnum If there is no target, all non-template functions named are selected. Otherwise, a non-template function with type \tcode{F} is selected for the function type \tcode{FT} of the target type if \tcode{F} (after possibly applying the function pointer conversion\iref{conv.fctptr}) is identical to \tcode{FT}. \begin{note} That is, the class of which the function is a member is ignored when matching a pointer-to-member-function type. \end{note} \pnum The specialization, if any, generated by template argument deduction\iref{temp.over,temp.deduct.funcaddr,temp.arg.explicit} for each function template named is added to the set of selected functions considered. \pnum Non-member functions, static member functions, and explicit object member functions match targets of function pointer type or reference to function type. Implicit object member functions match targets of pointer-to-member-function type. \begin{note} %% FIXME: Should this only apply after the eliminations in the next paragraph? If an implicit object member function is chosen, the result can be used only to form a pointer to member\iref{expr.unary.op}. \end{note} \pnum All functions with associated constraints that are not satisfied\iref{temp.constr.decl} are eliminated from the set of selected functions. If more than one function in the set remains, all function template specializations in the set are eliminated if the set also contains a function that is not a function template specialization. Any given non-template function \tcode{F0} is eliminated if the set contains a second non-template function that is more partial-ordering-constrained than \tcode{F0}\iref{temp.constr.order}. Any given function template specialization \tcode{F1} is eliminated if the set contains a second function template specialization whose function template is more specialized than the function template of \tcode{F1} according to the partial ordering rules of~\ref{temp.func.order}. After such eliminations, if any, there shall remain exactly one selected function. \pnum \begin{example} \begin{codeblock} int f(double); int f(int); int (*pfd)(double) = &f; // selects \tcode{f(double)} int (*pfi)(int) = &f; // selects \tcode{f(int)} int (*pfe)(...) = &f; // error: type mismatch int (&rfi)(int) = f; // selects \tcode{f(int)} int (&rfd)(double) = f; // selects \tcode{f(double)} void g() { (int (*)(int))&f; // cast expression as selector } \end{codeblock} The initialization of \tcode{pfe} is ill-formed because no \tcode{f()} with type \tcode{int(...)} has been declared, and not because of any ambiguity. \end{example} \begin{example} \begin{codeblock} struct X { int f(int); static int f(long); }; int (X::*p1)(int) = &X::f; // OK int (*p2)(int) = &X::f; // error: mismatch int (*p3)(long) = &X::f; // OK int (X::*p4)(long) = &X::f; // error: mismatch int (X::*p5)(int) = &(X::f); // error: wrong syntax for // pointer to member int (*p6)(long) = &(X::f); // OK \end{codeblock} \end{example} \begin{example} \begin{codeblock} template struct X { void f(short) requires B; void f(long); template void g(short) requires B; template void g(long); }; void test() { &X::f; // error: ambiguous; constraints are not considered &X::g; // error: ambiguous; constraints are not considered } \end{codeblock} \end{example} \pnum \begin{note} If \tcode{f} and \tcode{g} are both overload sets, the Cartesian product of possibilities is considered to resolve \tcode{f(\&g)}, or the equivalent expression \tcode{f(g)}. \end{note} \pnum \indextext{conversion!overload resolution and pointer}% \begin{note} Even if \tcode{B} is a public base of \tcode{D}, we have \begin{codeblock} D* f(); B* (*p1)() = &f; // error void g(D*); void (*p2)(B*) = &g; // error \end{codeblock} \end{note} \rSec1[over.oper]{Overloaded operators}% \rSec2[over.oper.general]{General}% \indextext{overloading!operator|(}% \indextext{overloaded operator|see{overloading, operator}}% \indextext{operator overloading|see{overloading, operator}} \pnum \indextext{operator!overloaded}% A declaration whose \grammarterm{declarator-id} is an \grammarterm{operator-function-id} shall declare a function or function template or an explicit instantiation or specialization of a function template. A function so declared is an \defnadj{operator}{function}. A function template so declared is an \defnx{operator function template}{function!operator!template}. A specialization of an operator function template is also an operator function. An operator function is said to \defnx{implement}{operator!implementation} the operator named in its \grammarterm{operator-function-id}. \begin{bnf} \nontermdef{operator-function-id}\br \keyword{operator} operator \end{bnf} \begin{bnf} %% Ed. note: character protrusion would misalign various operators. \microtypesetup{protrusion=false} \nontermdef{operator} \textnormal{one of}\br \terminal{\keyword{new} \ \ \ \ \ \keyword{delete} \ \ \keyword{new}[] \ \ \ \keyword{delete}[] \keyword{co_await} (\rlap{\,)} \ \ \ \ \ \ \ [\rlap{\,]} \ \ \ \ \ \ \ -> \ \ \ \ \ \ ->*}\br \terminal{\~ \ \ \ \ \ \ \ ! \ \ \ \ \ \ \ + \ \ \ \ \ \ \ - \ \ \ \ \ \ \ * \ \ \ \ \ \ \ / \ \ \ \ \ \ \ \% \ \ \ \ \ \ \ \caret{} \ \ \ \ \ \ \ \&}\br \terminal{| \ \ \ \ \ \ \ = \ \ \ \ \ \ \ += \ \ \ \ \ \ -= \ \ \ \ \ \ *= \ \ \ \ \ \ /= \ \ \ \ \ \ \%= \ \ \ \ \ \ \caret{}= \ \ \ \ \ \ \&=}\br \terminal{|= \ \ \ \ \ \ == \ \ \ \ \ \ != \ \ \ \ \ \ < \ \ \ \ \ \ \ > \ \ \ \ \ \ \ <= \ \ \ \ \ \ >= \ \ \ \ \ \ <=> \ \ \ \ \ \&\&}\br \terminal{|| \ \ \ \ \ \ << \ \ \ \ \ \ >> \ \ \ \ \ \ <<= \ \ \ \ \ >>= \ \ \ \ \ ++ \ \ \ \ \ \ -- \ \ \ \ \ \ ,} \end{bnf} \begin{note} The operators \tcode{new[]}, \tcode{delete[]}, \tcode{()}, and \tcode{[]} are formed from more than one token. The latter two operators are function call\iref{expr.call} and subscripting\iref{expr.sub}. \end{note} \indextext{operator!subscripting}% \indextext{operator!function call}% \pnum Both the unary and binary forms of \begin{ncsimplebnf} \terminal{+ \ \ \ \ \ - \ \ \ \ \ * \ \ \ \ \ \&} \end{ncsimplebnf} can be overloaded. \pnum \begin{note} \indextext{restriction!operator overloading}% The following operators cannot be overloaded: \begin{ncsimplebnf} \terminal{. \ \ \ \ \ .* \ \ \ \ :: \ \ \ \ ?:} \end{ncsimplebnf} nor can the preprocessing symbols \tcode{\#}\iref{cpp.stringize} and \tcode{\#\#}\iref{cpp.concat}. \end{note} \pnum \indextext{call!operator function}% Operator functions are usually not called directly; instead they are invoked to evaluate the operators they implement~(\ref{over.unary} -- \ref{over.inc}). They can be explicitly called, however, using the \grammarterm{operator-function-id} as the name of the function in the function call syntax\iref{expr.call}. \begin{example} \begin{codeblock} complex z = a.operator+(b); // \tcode{complex z = a+b;} void* p = operator new(sizeof(int)*n); \end{codeblock} \end{example} \pnum The allocation and deallocation functions, \keyword{operator} \keyword{new}, \keyword{operator} \tcode{\keyword{new}[]}, \keyword{operator} \keyword{delete}, and \keyword{operator} \tcode{\keyword{delete}[]}, are described completely in~\ref{basic.stc.dynamic}. The attributes and restrictions found in the rest of \ref{over.oper} do not apply to them unless explicitly stated in~\ref{basic.stc.dynamic}. \pnum The \keyword{co_await} operator is described completely in~\ref{expr.await}. The attributes and restrictions found in the rest of \ref{over.oper} do not apply to it unless explicitly stated in~\ref{expr.await}. \pnum \indextext{restriction!overloading}% An operator function shall have at least one function parameter or implicit object parameter whose type is a class, a reference to a class, an enumeration, or a reference to an enumeration. It is not possible to change the precedence, grouping, or number of operands of operators. The meaning of the operators \tcode{=}, (unary) \tcode{\&}, and \tcode{,} (comma), predefined for each type, can be changed for specific class types by defining operator functions that implement these operators. Likewise, the meaning of the operators (unary) \tcode{\&} and \tcode{,} (comma) can be changed for specific enumeration types. \indextext{overloaded operator!inheritance of}% Operator functions are inherited in the same manner as other base class functions. \pnum An operator function shall be a prefix unary, binary, function call, subscripting, class member access, increment, or decrement operator function. \pnum \indextext{operator}% \begin{note} The identities among certain predefined operators applied to fundamental types (for example, \tcode{++a} $\equiv$ \tcode{a+=1}) need not hold for operator functions. Some predefined operators, such as \tcode{+=}, require an operand to be an lvalue when applied to fundamental types; this is not required by operator functions. \end{note} \pnum \indextext{argument!overloaded operator and default}% An operator function cannot have default arguments\iref{dcl.fct.default}, except where explicitly stated below. Operator functions cannot have more or fewer parameters than the number required for the corresponding operator, as described in the rest of \ref{over.oper}. \pnum Operators not mentioned explicitly in subclauses~\ref{over.assign} through~\ref{over.inc} act as ordinary unary and binary operators obeying the rules of~\ref{over.unary} or~\ref{over.binary}.% \indextext{overloading!resolution!best viable function|)}% \indextext{overloading!resolution!viable functions|)} \rSec2[over.unary]{Unary operators}% \indextext{unary operator!overloaded}% \indextext{overloading!unary operator} \pnum A \defnadj{prefix unary}{operator function} is a function named \tcode{\keyword{operator}@} for a prefix \grammarterm{unary-operator} \tcode{@}\iref{expr.unary.op} that is either a non-static member function\iref{class.mfct} with no non-object parameters or a non-member function with one parameter. \indextext{unary operator!interpretation of}% For a \grammarterm{unary-expression} of the form \tcode{@ \grammarterm{cast-expression}}, the operator function is selected by overload resolution\iref{over.match.oper}. If a member function is selected, the expression is interpreted as \begin{ncsimplebnf} cast-expression \terminal{.} \keyword{operator} \terminal{@} \terminal{(}\terminal{)} \end{ncsimplebnf} Otherwise, if a non-member function is selected, the expression is interpreted as \begin{ncsimplebnf} \keyword{operator} \terminal{@} \terminal{(} cast-expression \terminal{)} \end{ncsimplebnf} \begin{note} The operators \tcode{++} and \tcode{--}\iref{expr.pre.incr} are described in~\ref{over.inc}. \end{note} \pnum \begin{note} The unary and binary forms of the same operator have the same name. Consequently, a unary operator can hide a binary operator from an enclosing scope, and vice versa. \end{note} \rSec2[over.binary]{Binary operators}% \rSec3[over.binary.general]{General}% \indextext{binary operator!overloaded}% \indextext{overloading!binary operator} \pnum A \defnadj{binary}{operator function} is a function named \tcode{\keyword{operator}@} for a binary operator \tcode{@} that is either a non-static member function\iref{class.mfct} with one non-object parameter or a non-member function with two parameters. \indextext{binary operator!interpretation of}% For an expression \tcode{$x$ @ $y$} with subexpressions $x$ and $y$, the operator function is selected by overload resolution\iref{over.match.oper}. If a member function is selected, the expression is interpreted as \begin{ncsimplebnf} $x$ \terminal{.} \keyword{operator} \terminal{@} \terminal{(} $y$ \terminal{)} \end{ncsimplebnf} Otherwise, if a non-member function is selected, the expression is interpreted as \begin{ncsimplebnf} \keyword{operator} \terminal{@} \terminal{(} $x$ \terminal{,} $y$ \terminal{)} \end{ncsimplebnf} \pnum An \defnadj{equality}{operator function} is an operator function for an equality operator\iref{expr.eq}. A \defnadj{relational}{operator function} is an operator function for a relational operator\iref{expr.rel}. A \defnadj{three-way comparison}{operator function} is an operator function for the three-way comparison operator\iref{expr.spaceship}. A \defnadj{comparison}{operator function} is an equality operator function, a relational operator function, or a three-way comparison operator function. \rSec3[over.assign]{Simple assignment} \indextext{assignment operator!overloaded}% \indextext{overloading!assignment operator} \pnum A \defnadj{simple assignment}{operator function} is a binary operator function named \tcode{\keyword{operator}=}. A simple assignment operator function shall be a non-static member function. \begin{note} Because only standard conversion sequences are considered when converting to the left operand of an assignment operation\iref{over.best.ics}, an expression \tcode{$x$ = $y$} with a subexpression $x$ of class type is always interpreted as \tcode{$x$.\keyword{operator}=($y$)}. \end{note} \pnum \begin{note} Since a copy assignment operator is implicitly declared for a class if not declared by the user\iref{class.copy.assign}, a base class assignment operator function is always hidden by the copy assignment operator function of the derived class. \end{note} \pnum \begin{note} Any assignment operator function, even the copy and move assignment operators, can be virtual. For a derived class \tcode{D} with a base class \tcode{B} for which a virtual copy/move assignment has been declared, the copy/move assignment operator in \tcode{D} does not override \tcode{B}'s virtual copy/move assignment operator. \begin{example} \begin{codeblock} struct B { virtual int operator= (int); virtual B& operator= (const B&); }; struct D : B { virtual int operator= (int); virtual D& operator= (const B&); }; D dobj1; D dobj2; B* bptr = &dobj1; void f() { bptr->operator=(99); // calls \tcode{D::\keyword{operator}=(int)} *bptr = 99; // ditto bptr->operator=(dobj2); // calls \tcode{D::\keyword{operator}=(const B\&)} *bptr = dobj2; // ditto dobj1 = dobj2; // calls implicitly-declared \tcode{D::\keyword{operator}=(const D\&)} } \end{codeblock} \end{example} \end{note} \rSec2[over.call]{Function call}% \indextext{function call operator!overloaded}% \indextext{overloading!function call operator} \pnum A \defnadj{function call}{operator function} is a function named \tcode{\keyword{operator}()} that is a member function with an arbitrary number of parameters. It may have default arguments. For an expression of the form \begin{ncsimplebnf} postfix-expression \terminal{(} \opt{expression-list} \terminal{)} \end{ncsimplebnf} where the \grammarterm{postfix-expression} is of class type, the operator function is selected by overload resolution\iref{over.call.object}. If a surrogate call function is selected, let $e$ be the result of invoking the corresponding conversion operator function on the \grammarterm{postfix-expression}; the expression is interpreted as \begin{ncsimplebnf} $e$ \terminal{(} \opt{expression-list} \terminal{)} \end{ncsimplebnf} Otherwise, the expression is interpreted as \begin{ncsimplebnf} postfix-expression \terminal{.} \keyword{operator} \terminal{(}\terminal{)} \terminal{(} \opt{expression-list} \terminal{)} \end{ncsimplebnf} \rSec2[over.sub]{Subscripting}% \indextext{subscripting operator!overloaded}% \indextext{overloading!subscripting operator} \pnum A \defnadj{subscripting}{operator function} is a member function named \tcode{\keyword{operator}[]} with an arbitrary number of parameters. It may have default arguments. For an expression of the form \begin{ncsimplebnf} postfix-expression \terminal{[} \opt{expression-list} \terminal{]} \end{ncsimplebnf} the operator function is selected by overload resolution\iref{over.match.oper}. If a member function is selected, the expression is interpreted as \begin{ncsimplebnf} postfix-expression . \keyword{operator} \terminal{[}\terminal{]} \terminal{(} \opt{expression-list} \terminal{)} \end{ncsimplebnf} \pnum \begin{example} \begin{codeblock} struct X { Z operator[](std::initializer_list); Z operator[](auto...); }; X x; x[{1,2,3}] = 7; // OK, meaning \tcode{x.\keyword{operator}[](\{1,2,3\})} x[1,2,3] = 7; // OK, meaning \tcode{x.\keyword{operator}[](1,2,3)} int a[10]; a[{1,2,3}] = 7; // error: built-in subscript operator a[1,2,3] = 7; // error: built-in subscript operator \end{codeblock} \end{example} \rSec2[over.ref]{Class member access} \indextext{member access operator!overloaded}% \indextext{overloading!member access operator} \pnum A \defnadj{class member access}{operator function} is a function named \tcode{\keyword{operator}->} that is a non-static member function taking no non-object parameters. For an expression of the form \begin{ncsimplebnf} postfix-expression \terminal{->} \opt{\keyword{template}} id-expression \end{ncsimplebnf} the operator function is selected by overload resolution\iref{over.match.oper}, and the expression is interpreted as \begin{ncsimplebnf} \terminal{(} postfix-expression . \keyword{operator} \terminal{->} \terminal{(}\terminal{)} \terminal{)} \terminal{->} \opt{\keyword{template}} id-expression \end{ncsimplebnf} Analogously, for an expression of the form \begin{ncsimplebnf} postfix-expression \terminal{->} splice-expression \end{ncsimplebnf} the operator function is selected by overload resolution, and the expression is interpreted as \begin{ncsimplebnf} \terminal{(} postfix-expression . \keyword{operator} \terminal{->} \terminal{(}\terminal{)} \terminal{)} \terminal{->} splice-expression \end{ncsimplebnf} \rSec2[over.inc]{Increment and decrement} \indextext{increment operator!overloaded|see{overloading, increment operator}}% \indextext{decrement operator!overloaded|see{overloading, decrement operator}}% \indextext{prefix ++ and -{-} overloading@prefix \tcode{++} and \tcode{--}!overloading}% \indextext{postfix ++ and -{-} overloading@postfix \tcode{++} and \tcode{--}!overloading}% \pnum \indextext{overloading!increment operator}% An \defnadj{increment}{operator function} is a function named \tcode{\keyword{operator}++}. If this function is a non-static member function with no non-object parameters, or a non-member function with one parameter, it defines the prefix increment operator \tcode{++} for objects of that type. If the function is a non-static member function with one non-object parameter (which shall be of type \tcode{int}) or a non-member function with two parameters (the second of which shall be of type \tcode{int}), it defines the postfix increment operator \tcode{++} for objects of that type. When the postfix increment is called as a result of using the \tcode{++} operator, the \tcode{int} argument will have value zero. \begin{footnote} Calling \tcode{\keyword{operator}++} explicitly, as in expressions like \tcode{a.\keyword{operator}++(2)}, has no special properties: The argument to \tcode{\keyword{operator}++} is \tcode{2}. \end{footnote} \begin{example} \begin{codeblock} struct X { X& operator++(); // prefix \tcode{++a} X operator++(int); // postfix \tcode{a++} }; struct Y { }; Y& operator++(Y&); // prefix \tcode{++b} Y operator++(Y&, int); // postfix \tcode{b++} void f(X a, Y b) { ++a; // \tcode{a.\keyword{operator}++();} a++; // \tcode{a.\keyword{operator}++(0);} ++b; // \tcode{\keyword{operator}++(b);} b++; // \tcode{\keyword{operator}++(b, 0);} a.operator++(); // explicit call: like \tcode{++a;} a.operator++(0); // explicit call: like \tcode{a++;} operator++(b); // explicit call: like \tcode{++b;} operator++(b, 0); // explicit call: like \tcode{b++;} } \end{codeblock} \end{example} \pnum \indextext{overloading!decrement operator}% A \defnadj{decrement}{operator function} is a function named \tcode{\keyword{operator}--} and is handled analogously to an increment operator function. \indextext{overloading!operator|)} \rSec1[over.built]{Built-in operators}% \indextext{overloading!built-in operators and} \pnum The candidate operator functions that represent the built-in operators defined in \ref{expr.compound} are specified in this subclause. These candidate functions participate in the operator overload resolution process as described in~\ref{over.match.oper} and are used for no other purpose. \begin{note} Because built-in operators take only operands with non-class type, and operator overload resolution occurs only when an operand expression originally has class or enumeration type, operator overload resolution can resolve to a built-in operator only when an operand has a class type that has a user-defined conversion to a non-class type appropriate for the operator, or when an operand has an enumeration type that can be converted to a type appropriate for the operator. Also note that some of the candidate operator functions given in this subclause are more permissive than the built-in operators themselves. As described in~\ref{over.match.oper}, after a built-in operator is selected by overload resolution the expression is subject to the requirements for the built-in operator given in \ref{expr.compound}, and therefore to any additional semantic constraints given there. In some cases, user-written candidates with the same name and parameter types as a built-in candidate operator function cause the built-in operator function to not be included in the set of candidate functions. \end{note} \pnum \indextext{type!integral!promoted}% \indextext{type!arithmetic!promoted}% In this subclause, the term \defn{promoted integral type} is used to refer to those cv-unqualified integral types which are preserved by integral promotion\iref{conv.prom} (including e.g. \tcode{int} and \tcode{long} but excluding e.g. \tcode{char}). \begin{note} In all cases where a promoted integral type is required, an operand of unscoped enumeration type will be acceptable by way of the integral promotions. \end{note} \pnum In the remainder of this subclause, \cvqual{vq} represents either \tcode{volatile} or no cv-qualifier. \pnum For every pair (\tcode{\placeholder{T}}, \cvqual{vq}), where \tcode{\placeholder{T}} is a cv-unqualified arithmetic type other than \tcode{bool} or a cv-unqualified pointer to (possibly cv-qualified) object type, there exist candidate operator functions of the form \begin{codeblock} @\cvqual{vq} \placeholder{T}@& operator++(@\cvqual{vq} \placeholder{T}@&); @\placeholder{T}@ operator++(@\cvqual{vq} \placeholder{T}@&, int); @\cvqual{vq} \placeholder{T}@& operator--(@\cvqual{vq} \placeholder{T}@&); @\placeholder{T}@ operator--(@\cvqual{vq} \placeholder{T}@&, int); \end{codeblock} \pnum For every (possibly cv-qualified) object type \tcode{\placeholder{T}} and for every function type \tcode{\placeholder{T}} that has neither \grammarterm{cv-qualifier}s nor a \grammarterm{ref-qualifier}, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@& operator*(@\placeholder{T}@*); \end{codeblock} \pnum For every type \tcode{\placeholder{T}} there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@* operator+(@\placeholder{T}@*); \end{codeblock} \pnum For every cv-unqualified floating-point or promoted integral type \tcode{\placeholder{T}}, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@ operator+(@\placeholder{T}@); @\placeholder{T}@ operator-(@\placeholder{T}@); \end{codeblock} \pnum For every promoted integral type \tcode{\placeholder{T}}, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@ operator~(@\placeholder{T}@); \end{codeblock} \pnum For every quintuple (\tcode{\placeholder{C1}}, \tcode{\placeholder{C2}}, \tcode{\placeholder{T}}, \cvqual{cv1}, \cvqual{cv2}), where \tcode{\placeholder{C2}} is a class type, \tcode{\placeholder{C1}} is the same type as \tcode{\placeholder{C2}} or is a derived class of \tcode{\placeholder{C2}}, and \tcode{\placeholder{T}} is an object type or a function type, there exist candidate operator functions of the form \begin{codeblock} @\cvqual{cv12} \placeholder{T}@& operator->*(@\cvqual{cv1} \placeholder{C1}@*, @\cvqual{cv2} \placeholder{T C2}@::*); \end{codeblock} where \cvqual{cv12} is the union of \cvqual{cv1} and \cvqual{cv2}. The return type is shown for exposition only; see~\ref{expr.mptr.oper} for the determination of the operator's result type. \pnum For every pair of types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}, where each of \tcode{\placeholder{L}} and \tcode{\placeholder{R}} is a cv-unqualified floating-point or promoted integral type, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{LR}@ operator*(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{LR}@ operator/(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{LR}@ operator+(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{LR}@ operator-(@\placeholder{L}@, @\placeholder{R}@); bool operator==(@\placeholder{L}@, @\placeholder{R}@); bool operator!=(@\placeholder{L}@, @\placeholder{R}@); bool operator<(@\placeholder{L}@, @\placeholder{R}@); bool operator>(@\placeholder{L}@, @\placeholder{R}@); bool operator<=(@\placeholder{L}@, @\placeholder{R}@); bool operator>=(@\placeholder{L}@, @\placeholder{R}@); \end{codeblock} where \tcode{\placeholder{LR}} is the result of the usual arithmetic conversions\iref{expr.arith.conv} between types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}. \pnum For every cv-unqualified integral type \tcode{\placeholder{T}} there exists a candidate operator function of the form \begin{codeblock} std::strong_ordering operator<=>(@\placeholder{T}@, @\placeholder{T}@); \end{codeblock} \pnum For every pair of cv-unqualified floating-point types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}, there exists a candidate operator function of the form \begin{codeblock} std::partial_ordering operator<=>(@\placeholder{L}@, @\placeholder{R}@); \end{codeblock} \pnum For every cv-qualified or cv-unqualified object type \tcode{\placeholder{T}} there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@* operator+(@\placeholder{T}@*, std::ptrdiff_t); @\placeholder{T}@& operator[](@\placeholder{T}@*, std::ptrdiff_t); @\placeholder{T}@* operator-(@\placeholder{T}@*, std::ptrdiff_t); @\placeholder{T}@* operator+(std::ptrdiff_t, @\placeholder{T}@*); @\placeholder{T}@& operator[](std::ptrdiff_t, @\placeholder{T}@*); \end{codeblock} \pnum For every \tcode{\placeholder{T}}, where \tcode{\placeholder{T}} is a pointer to object type, there exist candidate operator functions of the form \begin{codeblock} std::ptrdiff_t operator-(@\placeholder{T}@, @\placeholder{T}@); \end{codeblock} \pnum For every \tcode{\placeholder{T}}, where \tcode{\placeholder{T}} is an enumeration type or a pointer type, there exist candidate operator functions of the form \begin{codeblock} bool operator==(@\placeholder{T}@, @\placeholder{T}@); bool operator!=(@\placeholder{T}@, @\placeholder{T}@); bool operator<(@\placeholder{T}@, @\placeholder{T}@); bool operator>(@\placeholder{T}@, @\placeholder{T}@); bool operator<=(@\placeholder{T}@, @\placeholder{T}@); bool operator>=(@\placeholder{T}@, @\placeholder{T}@); @\placeholdernc{R}@ operator<=>(@\placeholder{T}@, @\placeholder{T}@); \end{codeblock} where \tcode{\placeholder{R}} is the result type specified in \ref{expr.spaceship}. \pnum For every \tcode{\placeholder{T}}, where \tcode{\placeholder{T}} is a pointer-to-member type, \tcode{std::meta::info}, or \tcode{std::nullptr_t}, there exist candidate operator functions of the form \begin{codeblock} bool operator==(@\placeholder{T}@, @\placeholder{T}@); bool operator!=(@\placeholder{T}@, @\placeholder{T}@); \end{codeblock} \pnum For every pair of promoted integral types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{LR}@ operator%(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{LR}@ operator&(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{LR}@ operator^(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{LR}@ operator|(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{L}@ operator<<(@\placeholder{L}@, @\placeholder{R}@); @\placeholder{L}@ operator>>(@\placeholder{L}@, @\placeholder{R}@); \end{codeblock} where \tcode{\placeholder{LR}} is the result of the usual arithmetic conversions\iref{expr.arith.conv} between types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}. \pnum For every triple (\tcode{\placeholder{L}}, \cvqual{vq}, \tcode{\placeholder{R}}), where \tcode{\placeholder{L}} is a cv-unqualified arithmetic type and \tcode{\placeholder{R}} is a cv-unqualified floating-point or promoted integral type, there exist candidate operator functions of the form \begin{codeblock} @\cvqual{vq} \placeholder{L}@& operator=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator*=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator/=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator+=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator-=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); \end{codeblock} \pnum For every pair (\tcode{\placeholder{T}}, \cvqual{vq}), where \tcode{\placeholder{T}} is any type, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@*@\cvqual{vq}@& operator=(@\placeholder{T}@*@\cvqual{vq}@&, @\placeholder{T}@*); \end{codeblock} \pnum For every pair (\tcode{\placeholder{T}}, \cvqual{vq}), where \tcode{\placeholder{T}} is an enumeration or pointer-to-member type, there exist candidate operator functions of the form \begin{codeblock} @\cvqual{vq} \placeholder{T}@& operator=(@\cvqual{vq} \placeholder{T}@&, @\placeholder{T}@); \end{codeblock} \pnum For every pair (\tcode{\placeholder{T}}, \cvqual{vq}), where \tcode{\placeholder{T}} is a cv-qualified or cv-unqualified object type, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@*@\cvqual{vq}@& operator+=(@\placeholder{T}@*@\cvqual{vq}@&, std::ptrdiff_t); @\placeholder{T}@*@\cvqual{vq}@& operator-=(@\placeholder{T}@*@\cvqual{vq}@&, std::ptrdiff_t); \end{codeblock} \pnum For every triple (\tcode{\placeholder{L}}, \cvqual{vq}, \tcode{\placeholder{R}}), where \tcode{\placeholder{L}} is a cv-unqualified integral type and \tcode{\placeholder{R}} is a promoted integral type, there exist candidate operator functions of the form \begin{codeblock} @\cvqual{vq} \placeholder{L}@& operator%=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator<<=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator>>=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator&=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator^=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); @\cvqual{vq} \placeholder{L}@& operator|=(@\cvqual{vq} \placeholder{L}@&, @\placeholder{R}@); \end{codeblock} \pnum There also exist candidate operator functions of the form \begin{codeblock} bool operator!(bool); bool operator&&(bool, bool); bool operator||(bool, bool); \end{codeblock} \pnum For every pair of types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}, where each of \tcode{\placeholder{L}} and \tcode{\placeholder{R}} is a cv-unqualified floating-point or promoted integral type, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{LR}@ operator?:(bool, @\placeholder{L}@, @\placeholder{R}@); \end{codeblock} where \tcode{\placeholder{LR}} is the result of the usual arithmetic conversions\iref{expr.arith.conv} between types \tcode{\placeholder{L}} and \tcode{\placeholder{R}}. \begin{note} As with all these descriptions of candidate functions, this declaration serves only to describe the built-in operator for purposes of overload resolution. The operator ``\tcode{?:}'' cannot be overloaded. \end{note} \pnum For every type \tcode{\placeholder{T}}, where \tcode{\placeholder{T}} is a pointer, pointer-to-member, or scoped enumeration type, there exist candidate operator functions of the form \begin{codeblock} @\placeholder{T}@ operator?:(bool, @\placeholder{T}@, @\placeholder{T}@); \end{codeblock}% \indextext{overloading|)} \rSec1[over.literal]{User-defined literals}% \indextext{user-defined literal!overloaded}% \indextext{overloading!user-defined literal} \begin{bnf} \nontermdef{literal-operator-id}\br \keyword{operator} unevaluated-string identifier\br \keyword{operator} user-defined-string-literal \end{bnf} \pnum The \grammarterm{user-defined-string-literal} in a \grammarterm{literal-operator-id} shall have no \grammarterm{encoding-prefix}. The \grammarterm{unevaluated-string} or \grammarterm{user-defined-string-literal} shall be empty. The \grammarterm{ud-suffix} of the \grammarterm{user-defined-string-literal} or the \grammarterm{identifier} in a \grammarterm{literal-operator-id} is called a \defnx{literal suffix identifier}{literal!suffix identifier}. The first form of \grammarterm{literal-operator-id} is deprecated\iref{depr.lit}. Some literal suffix identifiers are reserved for future standardization; see~\ref{usrlit.suffix}. A declaration whose \grammarterm{literal-operator-id} uses such a literal suffix identifier is ill-formed, no diagnostic required. \pnum A declaration whose \grammarterm{declarator-id} is a \grammarterm{literal-operator-id} shall declare a function or function template that belongs to a namespace (it could be a friend function\iref{class.friend}) or an explicit instantiation or specialization of a function template. A function declared with a \grammarterm{literal-operator-id} is a \defnx{literal operator}{literal!operator}. A function template declared with a \grammarterm{literal-operator-id} is a \defnx{literal operator template}{literal!operator!template}. \pnum The declaration of a literal operator shall have a \grammarterm{parameter-declaration-clause} equivalent to one of the following: \begin{codeblock} const char* unsigned long long int long double char wchar_t char8_t char16_t char32_t const char*, std::size_t const wchar_t*, std::size_t const char8_t*, std::size_t const char16_t*, std::size_t const char32_t*, std::size_t \end{codeblock} If a parameter has a default argument\iref{dcl.fct.default}, the program is ill-formed. \pnum A \defnx{raw literal operator}{literal!operator!raw} is a literal operator with a single parameter whose type is \tcode{const char*}. \pnum A \defnx{numeric literal operator template}{literal!operator!template numeric} is a literal operator template whose \grammarterm{template-parameter-list} has a single \grammarterm{template-parameter} that is a constant template parameter pack\iref{temp.variadic} with element type \tcode{char}. A \defnx{string literal operator template}{literal!operator!template string} is a literal operator template whose \grammarterm{template-parameter-list} comprises a single \grammarterm{parameter-declaration} that declares a constant template parameter of class type. The declaration of a literal operator template shall have an empty \grammarterm{parameter-declaration-clause} and shall declare either a numeric literal operator template or a string literal operator template. \pnum Literal operators and literal operator templates shall not have C language linkage. \pnum \begin{note} Literal operators and literal operator templates are usually invoked implicitly through user-defined literals\iref{lex.ext}. However, except for the constraints described above, they are ordinary namespace-scope functions and function templates. In particular, they are looked up like ordinary functions and function templates and they follow the same overload resolution rules. Also, they can be declared \keyword{inline} or \keyword{constexpr}, they can have internal, module, or external linkage, they can be called explicitly, their addresses can be taken, etc. \end{note} \pnum \begin{example} \begin{codeblock} void operator ""_km(long double); // OK string operator "" _i18n(const char*, std::size_t); // OK, deprecated template double operator ""_\u03C0(); // OK, UCN for lowercase pi float operator ""_e(const char*); // OK float operator ""E(const char*); // ill-formed, no diagnostic required: // reserved literal suffix\iref{usrlit.suffix,lex.ext} double operator""_Bq(long double); // OK, does not use the reserved \grammarterm{identifier} \tcode{_Bq}\iref{lex.name} double operator"" _Bq(long double); // ill-formed, no diagnostic required: // uses the reserved \grammarterm{identifier} \tcode{_Bq}\iref{lex.name} float operator " "B(const char*); // error: non-empty \grammarterm{string-literal} string operator ""5X(const char*, std::size_t); // error: invalid literal suffix identifier double operator ""_miles(double); // error: invalid \grammarterm{parameter-declaration-clause} template int operator ""_j(const char*); // error: invalid \grammarterm{parameter-declaration-clause} extern "C" void operator ""_m(long double); // error: C language linkage \end{codeblock} \end{example}