Modern C++
Programming
8. Object-Oriented
Programming II
Polymorphism and Operator Overloading
Federico Busato
2024-11-05
Table of Contents
3 Operator Overloading
Overview
Comparison Operator operator<
Spaceship Operator operator<=>
Subscript Operator operator[]
Multidimensional Subscript Operator operator[]
Function Call Operator operator()
static operator() and static operator[]
Conversion Operator operator T()
Return Type Overloading Resolution
⋆
2/66
Polymorphism
Polymorphism
Polymorphism (meaning “having multiple forms”) is the capability of an entity of
mutating its behavior in accordance with the specific usage context
Polymorphism dispatch can be implemented at
• Compile-time (static polymorphism): when the called instance is known
before the program start
• Run-time (dynamic polymorphism): when the called instance is known only
during the execution, i.e. depends on run-time values
In C++, the term polymorphic is strongly associated with dynamic polymorphism
(overriding)
4/66
Function Binding
Connecting the function call to the function body is called Binding
• In Early Binding or Static Binding or Compile-time Binding, the compiler identifies
the type of object at compile-time
- the program can jump directly to the function address
• In Late Binding or Dynamic Binding or Run-time binding, the run-time identifies
the type of object at execution-time and then matches the function call with the
correct function definition
- the program has to read the address held in the pointer and then jump to that
address (less efficient since it involves an extra level of indirection)
C++ achieves late binding by declaring a virtual function
5/66
Polymorphism Forms
• A-hoc polymorphism: when it involves to a set of individually specified types,
e.g. function overloading
void f(int);
void f(double);
• Parametric polymorphism: when it involves generic types, e.g. templates
template<typename T>
void f(T);
• Subtyping: when it operates on elements of subtypes, e.g. virtual functions
// B : A
void f(A*); // also works for B if the called function are virtual
6/66
C++ Mechanisms for Polymorphism 1/2
• Preprocessing
# define ADD(x, y) x + y // ADD(3, 4) or ADD(3.0, 4.0)
• Function/Operator overloading
void f(int);
void f(double);
• Templates
template<typename T>
void f(T); // f(3) or f(3.0)
• Virtual functions (see next slides)
7/66
C++ Mechanisms for Polymorphism 2/2
Mechanism Implementation Form
Preprocessing static Parametric
Function/Operator overloading static Ad-hoc
Template static Parametric
Virtual function dynamic Subtyping
8/66
Dynamic Polymorphism in C++
• At run-time, objects of a base class behaves as objects of a derived class
• A Base class may define and implement polymorphic methods, and derived
classes can override them, which means they provide their own implementations,
invoked at run-time depending on the context
struct A {
void f() { cout << "A"; }
};
struct B : A {
void f() { cout << "B"; }
};
void g(A& a) { a.f(); } // accepts A and B
// note: g(B&) would only accept B
A a; B b;
g(a); // print "A"
g(b); // print "A" not "B"!!!
9/66
Polymorphism - virtual method
struct A {
virtual void f() { cout << "A"; }
}; // now "f()" is virtual, evaluated at run-time
struct B : A {
virtual void f() { cout << "B"; }
}; // now "B::f()" overrides "A::f()", evaluated at run-time
void g(A& a) { a.f(); } // accepts A and B
A a;
B b;
g(a); // print "A"
g(b); // NOW, print "B"!!!
The virtual keyword is not necessary in derived classes but it improves the readability and
clarity the intention
10/66
When virtual works
struct A {
virtual void f() { cout << "A"; }
};
struct B : A {
void f() { cout << "B"; }
};
void f(A& a) { a.f(); } // ok, print "B"
void g(A* a) { a->f(); } // ok, print "B"
void h(A a) { a.f(); } // does not work with pass-by value!! print "A"
B b;
f(b); // print "B"
g(&b); // print "B"
h(b); // print "A" (cast to A)
11/66
Polymorphism Dynamic Behavior
struct A {
virtual void f() { cout << "A"; }
};
struct B : A {
void f() { cout << "B"; }
};
A* get_object(bool selectA) {
return (selectA) ? new A() : new B();
}
get_object(true)->f(); // print "A"
get_object(false)->f(); // print "B"
12/66
Virtual Table 1/2
vtable
The virtual table (vtable) is a lookup table of functions used to resolve function
calls and support dynamic dispatch (late binding)
A virtual table contains one entry for each virtual function that can be called by
objects of the class. Each entry in this table is simply a function pointer that points to
the most-derived function accessible by that class
The compiler adds a hidden pointer to the base class which points to the virtual table
for that class ( sizeof considers the vtable pointer)
13/66
Virtual Table 2/2
struct A {
virtual void f();
virtual void g();
};
struct B : A {
void f();
};
14/66
Does the vtable really exist? (answer: YES)
struct A {
int x = 3;
virtual void f() { cout << "abc"; }
};
A* a1 = new A;
A* a2 = (A*) malloc(sizeof(A));
cout << a1->x; // print "3"
cout << a2->x; // undefined value!!
a1->f(); // print "abc"
a2->f(); // segmentation fault
A
Lesson learned: Never use malloc in C++
15/66
Virtual Method Notes
virtual classes allocate one extra pointer (hidden)
struct A {
virtual void f1();
virtual void f2();
};
class B : A {};
cout << sizeof(A); // 8 bytes (vtable pointer)
cout << sizeof(B); // 8 bytes (vtable pointer)
16/66
override Keyword 1/2
override Keyword (C++11)
The override keyword ensures that the function is virtual and is overriding a
virtual function from a base class
It forces the compiler to check the base class to see if there is a virtual function
with this exact signature
override implies virtual ( virtual should be omitted)
17/66
override Keyword 2/2
struct A {
virtual void f(int a); // a "float" value is casted to "int"
}; // ∗ ∗ ∗
struct B : A {
void f(int a) override; // ok
void f(float a); // (still) very dangerous!!
// ∗ ∗ ∗
// void f(float a) override; // compile error not safe
// void f(int a) const override; // compile error not safe
};
// ∗ ∗ ∗ f(3.3f) has a different behavior between A and B
18/66
final Keyword
final Keyword (C++11)
The final keyword prevents inheriting from classes or overriding methods in
derived classes
struct A {
virtual void f(int a) final; // "final" method
};
struct B : A {
// void f(int a); // compile error f(int) is "final"
void f(float a); // dangerous (still possible)
}; // "override" prevents these errors
struct C final { // cannot be extended
};
// struct D : C { // compile error C is "final"
// };
19/66
Virtual Methods (Common Error 1)
All classes with at least one virtual method should declare a virtual
destructor
struct A {
∼A() { cout << "A"; } // <-- here the problem (not virtual)
virtual void f(int a) {}
};
struct B : A {
int* array;
B() { array = new int[1000000]; }
∼B() { delete[] array; }
};
//----------------------------------------------------------------------
void destroy(A* a) {
delete a; // call ∼A()
}
B* b = new B;
destroy(b); // without virtual, ∼B() is not called
// destroy() prints only "A" -> huge memory leak!!
20/66
Virtual Methods (Common Error 2)
Do not call virtual methods in constructor and destructor
• Constructor : The derived class is not ready until constructor is completed
• Destructor : The derived class is already destroyed
struct A {
A() { f(); } // what instance is called? "B" is not ready
// it calls A::f(), even though A::f() is virtual
virtual void f() { cout << "Explosion"; }
};
struct B : A {
B() = default; // call A(). Note: A() may be also implicit
void f() override { cout << "Safe"; }
};
B b; // call B(), print "Explosion", not "Safe"!!
21/66
Virtual Methods (Common Error 3)
Do not use default parameters in virtual methods
Default parameters are not inherited
struct A {
virtual void f(int i = 5) { cout << "A::" << i << "\n"; }
virtual void g(int i = 5) { cout << "A::" << i << "\n"; }
};
struct B : A {
void f(int i = 3) override { cout << "B::" << i << "\n"; }
void g(int i) override { cout << "B::" << i << "\n"; }
};
A a; B b;
a.f(); // ok, print "A::5"
b.f(); // ok, print "B::3"
A& ab = b;
ab.f(); // !!! print "B::5" // the virtual table of A
// contains f(int i = 5) and
ab.g(); // !!! print "B::5" // g(int i = 5) but it points
// to B implementations
22/66
Pure Virtual Method 1/2
Pure Virtual Method
A pure virtual method is a function that must be implemented in derived classes
(concrete implementation)
Pure virtual functions can have or not have a body
struct A {
virtual void f() = 0; // pure virtual without body
virtual void g() = 0; // pure virtual with body
};
void A::g() {} // pure virtual implementation (body) for g()
struct B : A {
void f() override {} // must be implemented
void g() override {} // must be implemented
};
23/66
Pure Virtual Method 2/2
A class with one pure virtual function cannot be instantiated
struct A {
virtual void f() = 0;
};
struct B1 : A {
// virtual void f() = 0; // implicitly declared
};
struct B2 : A {
void f() override {}
};
// A a; // "A" has a pure virtual method
// B1 b1; // "B1" has a pure virtual method
B2 b2; // ok
24/66
Abstract Class and Interface
• A class is interface if it has only pure virtual functions and optionally (suggested)
a virtual destructor. Interfaces do not have implementation or data
• A class is abstract if it has at least one pure virtual function
struct A { // INTERFACE
virtual ∼A(); // to implement
virtual void f() = 0;
};
struct B { // ABSTRACT CLASS
B() {} // abstract classes may have a contructor
virtual void g() = 0; // at least one pure virtual
protected:
int x; // additional data
};
25/66
Hierarchy Casting
Class-casting allows implicit or explicit conversion of a class into another one across
its hierarchy
26/66
Hierarchy Casting
Upcasting Conversion between a derived class reference or pointer to a base class
- It can be implicit or explicit
- It is safe
- static cast or dynamic cast // see next slides
Downcasting Conversion between a base class reference or pointer to a derived class
- It is only explicit
- It can be dangerous
- static cast or dynamic cast
Sidecasting (Cross-cast) Conversion between a class reference or pointer to another
class of the same hierarchy level
- It is only explicit
- It can be dangerous
- dynamic cast
27/66
Upcasting and Downcasting Example
struct A {
virtual void f() { cout << "A"; }
};
struct B : A {
int var = 3;
void f() override { cout << "B"; }
};
A a;
B b;
A& a1 = b; // implicit cast upcasting
static_cast<A&>(b).f(); // print "B" upcasting
static_cast<B&>(a).f(); // print "A" downcasting
cout << b.var; // print 3 (no cast)
cout << static_cast<B&>(a).var; // potential segfault!!! downcasting
// "var" does not exist in "A"
28/66
Sidecasting Example
struct A {
virtual void f() { cout << "A"; }
};
struct B1 : A {
void f() override { cout << "B1"; }
};
struct B2 : A {
void f() override { cout << "B2"; }
};
B1 b1;
B2 b2;
dynamic_cast<B2&>(b1).f(); // sidecasting, throw std::bad cast
dynamic_cast<B1&>(b2).f(); // sidecasting, throw std::bad cast
// static_cast<B1&>(b2).f(); // compile error
29/66
Run-time Type Identification
RTTI
Run-Time Type Information (RTTI) is a mechanism that allows the type of object
to be determined at runtime
C++ expresses RTTI through three features:
• dynamic cast keyword: conversion of polymorphic types
• typeid keyword: identifying the exact type of object
• type info class: type information returned by the typeid operator
RTTI is available only for classes that are polymorphic, which means they have at least
one virtual method
30/66
type info and typeid
type info class has the method name() which returns the name of the type
struct A {
virtual void f() {}
};
struct B : A {};
A a;
B b;
A& a1 = b; // implicit upcasting
cout << typeid(a).name(); // print "1A"
cout << typeid(b).name(); // print "1B"
cout << typeid(a1).name(); // print "1B"
31/66
dynamic cast
dynamic cast , differently from static cast , uses RTTI for deducing the
correctness of the output type
This operation happens at run-time and it is expensive
dynamic cast<New>(Obj) has the following properties:
• Convert between a derived class Obj to a base class New → upcasting.
New/Obj are both pointers or references
• Throw std::bad cast if New/Obj are references and New/Obj cannot be
converted
• Returns NULL if New/Obj are pointers and New/Obj cannot be converted
32/66
dynamic cast Example 1
struct A {
virtual void f() { cout << "A"; }
};
struct B : A {
void f() override { cout << "B"; }
};
A a;
B b;
dynamic_cast<A&>(b).f(); // print "B" upcasting
// dynamic_cast<B&>(a).f(); // throw std::bad cast
// wrong downcasting
dynamic_cast<B*>(&a); // returns nullptr
// wrong downcasting
33/66
dynamic cast Example 2
struct A {
virtual void f() { cout << "A"; }
};
struct B : A {
void f() override { cout << "B"; }
};
A* get_object(bool selectA) {
return (selectA) ? new A() : new B();
}
void g(bool value) {
A* a = get_object(value);
B* b = dynamic_cast<B*>(a); // downcasting + check
if (b != nullptr)
b->f(); // exectuted only when it is safe
}
34/66
Operator Overloading
Operator Overloading
Operator overloading is a special case of polymorphism in which some operators
are treated as polymorphic functions and have different behaviors depending on the
type of its arguments
struct Point {
int x, y;
Point operator+(const Point& p) const {
return {x + p.x, y + p.y};
}
};
Point a{1, 2};
Point b{5, 3};
Point c = a + b; // "c" is (6, 5)
35/66
Operator Overloading
Category Operators
Arithmetic + - * / % ++ --
Comparison == != < <= > >= <=>
Bitwise | & ˆ ∼ << >>
Logical ! && ||
Compound Assignment Arithmetic += -= *= /= %=
Compound Assignment Bitwise >>= <<= |= &= ˆ=
Subscript []
Function call ()
Address-of, Reference, Dereferencing & -> ->* *
Memory new new[] delete delete[]
Comma ,
• Categories not in bold are rarely used in practice
• Operators that cannot be overloaded: ? . .* :: sizeof typeof
36/66
Comparison Operator operator<
Relational and comparison operators operator<, <=, ==, >= > are used for
comparing two objects
In particular, the operator< is used to determine the ordering of a set of objects
(e.g. sort)
# include <algorithm>
struct A {
int x;
bool operator<(A a) const {
return x * x < a.x * a.x;
}
};
A array[] = {5, -1, 4, -7};
std::sort(array, array + 4);
// array: {-1, 4, 5, -7}
37/66
Spaceship Operator operator<=> 1/4
C++20 allows overloading the spaceship operator <=> for replacing all comparison
operators operator<, <=, ==, >= >
struct A {
bool operator==(const A&) const;
bool operator!=(const A&) const;
bool operator<(const A&) const;
bool operator<=(const A&) const;
bool operator>(const A&) const;
bool operator>=(const A&) const;
};
// replaced by
struct B {
int operator<=>(const B&) const;
};
38/66
Spaceship Operator operator<=> 2/4
# include <compare>
struct Obj {
int x;
auto operator<=>(const Obj& other) {
return x - other.x; // or even better "x <=> other.x"
}
};
Obj a{3};
Obj b{5};
a < b; // true, even if the operator< is not defined
a == b; // false
a <=> b < 0; // true
39/66
Spaceship Operator operator<=> 3/4
The compiler can also generate the code for the spaceship operator = default , even
for multiple fields and arrays, by using the default comparison semantic of its members
# include <compare>
struct Obj {
int x;
char y;
short z[2];
auto operator<=>(const Obj&) const = default;
// if x == other.x, then compare y
// if y == other.y, then compare z
// if z[0] == other.z[0], then compare z[1]
};
40/66
Spaceship Operator operator<=> 4/4
The spaceship operator can use one of the following ordering:
strong ordering • if a is equivalent to b , f(a) is also equivalent to f(b)
• exactly one of < , == , or > must be true
O integral types, e.g. int , char
weak ordering • if a is equivalent to b , f(a) may not be equivalent to f(b)
• exactly one of < , == , or > must be true
O rectangles, e.g. R{2, 5} == R{5, 2}
partial ordering • if a is equivalent to b , f(a) may not be equivalent to f(b)
• < , == , or > may all be false
O floating-point float , e.g. NaN
41/66
Subscript Operator operator[]
The array subscript operator[] allows accessing to an object in an array-like fashion
The operator accepts everything as parameter, not just integers
struct A {
char permutation[] {'c', 'b', 'd', 'a', 'h', 'y'};
char& operator[](char c) { // read/write
return permutation[c - 'a'];
}
char operator[](char c) const { // read only
return permutation[c - 'a'];
}
};
A a;
a['d'] = 't';
42/66
Multidimensional Subscript Operator operator[]
C++23 introduces the multidimensional subscript operator and replaces the standard
behavior of the comma operator
struct A {
int operator[](int x) { return x; }
};
struct B {
int operator[](int x, int y) { return x * y; } // not allowed before C++23
};
int main() {
A a;
cout << a[3, 4]; // return 4 (bug)
B b;
cout << b[3, 4]; // return 12, C++23
}
43/66
Function Call Operator operator()
The function call operator operator() is generally overloaded to create objects
which behave like functions, or for classes that have a primary operation (see Basic
Concepts IV lecture)
# include <numeric> // for std::accumulate
struct Multiply {
int operator()(int a, int b) const {
return a * b;
}
};
int array[] = { 2, 3, 4 };
int factorial = std::accumulate(array, array + 3, 1, Multiply{});
cout << factorial; // 24
44/66
static operator() and static operator[]
C++23 introduces the static version of the function call operator operator()
and the subscript operator operator[] to avoid passing the this pointer
# include <numeric> // for std::accumulate
struct Multiply {
// int operator()(int a, int b); // declaration only
static int operator()(int a, int b); // best efficiency, no need to access
}; // internal data members
struct MyArray {
// int operator[](int x);
static int operator[](int x); // best efficiency
};
int array[] = { 2, 3, 4 };
int factorial = std::accumulate(array, array + 3, 1, Multiply{});
45/66
Conversion Operator operator T() 1/2
The conversion operator operator T() allows objects to be either implicitly or
explicitly (casting) converted to another type
class MyBool {
int x;
public:
MyBool(int x1) : x{x1} {}
operator bool() const { // implicit return type
return x == 0;
}
};
MyBool my_bool{3};
bool b = my_bool; // b = false, call operator bool()
46/66
Conversion Operator operator T() 2/2
C++11 Conversion operators can be marked explicit to prevent implicit
conversions. It is a good practice as for class constructors
struct A {
operator bool() { return true; }
};
struct B {
explicit operator bool() { return true; }
};
A a;
B b;
bool c1 = a;
// bool c2 = b; // compile error: explicit
bool c3 = static_cast<bool>(b);
47/66
Return Type Overloading Resolution
⋆
struct A {
operator float() { return 3.0f; }
operator int() { return 2; }
};
auto f() {
return A{};
}
float x = f();
int y = f();
cout << x << " " << y; // x=3.0f, y=2
48/66
Increment and Decrement Operators operator++/--
The increment and decrement operators operator++, operator-- are used to update
the value of a variable by one unit
struct A {
int* ptr;
int pos;
A& operator++() { // Prefix notation (++var):
++ptr; // returns the new copy of the object by-reference
++pos;
return *this;
}
A operator++(int a) { // Postfix notation (var++):
A tmp = *this; // returns the old copy of the object by-value
++ptr;
++pos;
return tmp;
}
};
49/66
Assignment Operator operator= 1/3
The assignment operator operator= is used to copy values from one object to
another already existing object
# include <algorithm> //std::fill, std::copy
struct Array {
char* array;
int size;
Array(int size1, char value) : size{size1} {
array = new char[size];
std::fill(array, array + size, value);
}
∼Array() { delete[] array; }
Array& operator=(const Array& x) { .... } // --> see next slide
};
Array a{5, 'o'}; // ["ooooo"]
Array b{3, 'b'}; // ["bbb"]
a = b; // a = ["bbb"] <-- goal
50/66
Assignment Operator operator= 2/3
• First option:
Array& operator=(const Array& x) {
if (this == &x) // (1) Check for self assignment
return *this;
delete[] array; // (2) Release class resources
size = x.size; // (3) Re-initialize class resources
array = new int[x.size];
std::copy(x.array, x.array + size, array); // (4) deep copy
return *this;
}
• Second option (less intuitive):
Array& operator=(Array x) { // pass by-value
swap(*this, x); // now we need a swap function for A
return *this; // x is destroyed at the end
} // --> see next slide
51/66
Assignment Operator operator=
⋆
3/3
swap method:
friend void swap(A& x, A& y) {
using std::swap;
swap(x.size, y.size);
swap(x.array, y.array);
}
• why using std::swap ? if swap(x, y) finds a better match, it will use that
instead of std::swap
• why friend ? it allows the function to be used from outside the structure/class
scope
stackoverflow.com/questions/3279543
stackoverflow.com/questions/5695548
52/66
Stream Operator operator<<
The stream operation operator<< can be overloaded to perform input and output
for user-defined types
# include <iostream>
struct Point {
int x, y;
friend std::ostream& operator<<(std::ostream& stream,
const Point& point) {
stream << "(" << point.x << "," << point.y << ")";
return stream;
}
// operator<< is a member of std::ostream -> need friend
}; // implementation and definition can be splitted (not suggested for operator<<)
Point point{1, 2};
std::cout << point; // print "(1, 2)"
53/66
Operators Precedence
Operators preserve precedence and short-circuit properties
struct MyInt {
int x;
int operatorˆ(int exp) { // exponential
int ret = 1;
for (int i = 0; i < exp; i++)
ret *= x;
return ret;
}
};
MyInt x{3};
int y = xˆ2;
cout << y; // 9
int z = xˆ2 + 2;
cout << z; // 81 !!!
54/66
Binary Operators Note
Binary operators should be implemented as friend methods
struct A {}; struct C {};
struct B : A {
bool operator==(const A& x) { return true; }
};
struct D : C {
friend bool operator==(const C& x, const C& y) { return true; } // inline
};
// bool operator==(const C& x, const C& y) { return true; } // out-of-line
A a; B b; C c; D d;
b == a; // ok
// a == b; // compile error // "A" does not have == operator
c == d; // ok, use operator==(const C&, const C&)
d == c; // ok, use operator==(const C&, const C&)
55/66
Overview
The term layout refers to how an object is arranged in memory
C++ defines four types of layouts:
• aggregate
• trivial copyable
• standard layout
• plain-old data (POD)
Such layouts are important to understand how the C++ objects interact with pure C
API and for optimization purposes, e.g. pass in registers, memcpy , and serialization
56/66
Aggregate 1/3
Aggregate
An aggregate is an array, struct, or class which supports aggregate
initialization (form of list-initialization) through curly braces syntax {}
• No user-provided constructors
• No private / protected non- static data members and base class
• No virtual functions
* No base classes, until C++17
* No brace-or-equal-initializers for non-static data members, until C++14
R Apply recursively to base classes non- static data members
No restrictions:
• Non- static uninitialized (until C++14) data and function members
• static data and function members
stackoverflow.com/questions/4178175
57/66
Aggregate - Examples 2/3
struct Aggregate {
int x; // ok, public member
int y[3]; // ok, arrays are also fine
int z { 3 }; // only C++14
Aggregate() = default; // ok, defaulted constructor
Aggregate& operator=(const& Aggregate); // ok, function
private: // copy-assignment
void f() {} // ok, private function
};
struct NotAggregate1 {
NotAggregate1(); // !! user-provided constructor
virtual void f(); // !! virtual function
};
class NotAggregate2 : NotAggregate1 { // !! the base class is not an aggregate
int x; // !! x is private
NotAggregate1 y; // !! y is not an aggregate (recursive property)
};
58/66
Aggregate - Examples 3/3
struct Aggregate1 {
int x;
struct Aggregate2 {
int a;
int b[3];
} y;
};
int array1[3] = {1, 2, 3};
int array2[3] {1, 2, 3};
Aggregate1 agg1 = {1, {2, {3, 4, 5}}};
Aggregate1 agg2 {1, {2, {3, 4, 5}}};
Aggregate1 agg3 = {1, 2, 3, 4, 5};
59/66
Trivial Class 1/2
Trivial Class
A Trivial Class is a class trivial copyable (supports memcpy)
Trivial copyable:
• No user-provided copy/move/default constructors, destructor, and copy/move
assignment operators
• No virtual functions
R Apply recursively to base classes and non- static data members
No restrictions:
• User-declared constructors different from copy/move/default
• Functions or static ,non- static data members initialization
• protected / private members
60/66
Trivial Class - Examples 2/2
struct NonTrivial {
NonTrivial(); // !! user-provided constructor
virtual void f(); // !! virtual function
};
struct Trivial1 {
Trivial1() = default; // ok, defaulted constructor
Trivial1(int) {} // ok, user-default constructor
static int x; // ok, static member
void f(); // ok, function
private:
int z { 3 } // ok, private and initialized
};
struct Trivial2 : Trivial1 { // ok, base class is trivial
int Trivial1[3]; // ok, array of trivials is trivial
};
61/66
Standard-Layout Class 1/2
Standard-Layout
A standard-layout class is a class with the same memory layout of the
equivalent C struct or union (useful for communicating with other languages)
• No virtual functions
• Only one control access ( public / protected / private ) for all non- static
data members
• No base classes with non - static data members
• No base classes of the same type as the first non- static data member
R Apply recursively to base classes and non- static data members
62/66
Standard-Layout Class (examples) 2/2
struct StandardLayout1 {
StandardLayout1(); // ok, user-provided contructor
void f(); // ok, non-virtual function
};
class StandardLayout2 : StandardLayout1 {
int x, y; // ok, both are private
StandardLayout1 y; // ok, 'y' is not the first data member
};
struct StandardLayout4 : StandardLayout1, StandardLayout2 {
// ok, can use multiple inheritance as long as only
// one class in the hierarchy has non-static data members
};
63/66
Plain Old Data (POD)
Plain Old Data (POD): Trivial copyable (T) + Standard-Layout (S)
(T) No user-provided copy/move/default constructors, destructor, and copy/move
assignment operators
(S) Only one control access ( public / protected / private ) for all non- static
data members
(S) No base classes with non - static data members
(S) No base classes of the same type as the first non- static data member
(T, S) No virtual functions
R Apply recursively to base classes and non- static data members
64/66
C++ std Utilities
C++11 provides three utilities to check if a type is POD, Trivial Copyable,
Standard-Layout
• std::is pod checks for POD, deprecated in C++20
• std::is trivially copyable checks for trivial copyable
• std::is standard layout checks for standard-layout
# include <type_traits>
struct A {
int x;
private:
int y;
};
cout << std::is_trivially_copyable_v<A>; // true
cout << std::is_standard_layout_v<A>; // false
cout << std::is_pod_v<A>; // false
65/66
Object Layout Hierarchy
66/66
