Home | Projects | Notes > C++ Programming > Constructors & Destructors

Constructors & Destructors

Constructors

• Special member method

• Invoked during object creation

• Useful for initialization (to a known/safe/stable state)

• Has same name as the class

• No return type is specified

• Can be overloaded

Examples

• Player class

  xxxxxxxxxx
  12
  1
  class Player
  2
  {
  3
  private:
  4
      std::string name;
  5
      int health;
  6
      int xp;
  7
  public:
  8
      // Overloaded constructors
  9
      Player();
  10
      Player(std::string name);
  11
      Player(std::string name, int health, int xp);
  12
  };
  

• Account class

  xxxxxxxxxx
  12
  1
  class Account
  2
  {
  3
  private:
  4
      std::string name;
  5
      double balance;
  6
  public:
  7
      // Constructors
  8
      Account();
  9
      Account(std::string name, double balance);
  10
      Account(std::string name);
  11
      Account(double balance);
  12
  };
  

   

Destructors

• Special member method

• Has same name as the class except that it is proceeded with a tilde (~)

• Invoked automatically when an object is destroyed (When an object goes out of its scope or a pointer pointing to an object gets deleted)

  • It is a great place to release memory, close files and free up any other resources.

• No return type and no parameters

• Only 1 destructor is allowed per class - CANNOT be overloaded!

  • Doesn't make sense to overload a destructor since it is called automatically by C++.

• Useful to release memory and other resources

Example

• Player class

  xxxxxxxxxx
  14
  1
  class Player
  2
  {
  3
  private:
  4
      std::string name;
  5
      int health;
  6
      int xp;
  7
  public:
  8
      // Overloaded constructors
  9
      Player();
  10
      Player(std::string name);
  11
      Player(std::string name, int health, int xp);
  12
      // Destructor
  13
      ~Player();
  14
  };
  

  xxxxxxxxxx
  9
  1
  {
  2
      // The following 3 objects are all local and get created on stack memory
  3
      Player jack;
  4
      Player sunny {"Sunny", 100, 4};
  5
      Player yena {"Yena"};
  6
  } // 3 destructors called
  7
  
  8
  Player *kj = new Player("KJ", 1000, 0);
  9
  delete enemy;   // Destructor called
  

• Account class

  xxxxxxxxxx
  14
  1
  class Account
  2
  {
  3
  private:
  4
      std::string name;
  5
      double balance;
  6
  public:
  7
      // Constructors
  8
      Account();
  9
      Account(std::string name, double balance);
  10
      Account(std::string name);
  11
      Account(double balance);
  12
      // Destructor
  13
      ~Account();
  14
  };
  

   

Default Constructor

• A constructor that does not expect any arguments (Also, called the no-args constructor)

• If you do NOT provide any constructors at all for a class, C++ will generate a default constructor that does nothing.

  • Since the compiler generated default constructor does nothing, the class member attributes could contain garbage value because they haven't been initialized.

• Called when you instantiate a new object with no arguments

• It's a best practice to provide a user-defined constructor that ensures objects are instantiated in a known, stable (safe) state.

  • Once the user provides a user-defined constructor for a class, C++ will not generate the no-args default constructor automatically. If you still need it, then you must explicitly define it yourself.

Examples

• Player class

  xxxxxxxxxx
  2
  1
  Player jack;                    // Creates an object on the stack
  2
  Palyer *sunny = new Player;     // Creates an object on the heap
  

• Account class

  With default constructor

  xxxxxxxxxx
  15
  1
  class Account
  2
  {
  3
  private:
  4
      std::string name;
  5
      double balance;
  6
  public:
  7
      // Default constructor
  8
      Account()
  9
      {
  10
          name = "None";
  11
          balance = 0.0;
  12
      }
  13
      bool withdraw(double amount);
  14
      bool deposit(double amount);
  15
  };
  

  xxxxxxxxxx
  5
  1
  Account jack_account;
  2
  Account yena_account;
  3
  
  4
  Account *sunny_account = new Account;
  5
  delete sunny_account;
  

  With a user-defined constructor (No default constructor)

  xxxxxxxxxx
  15
  1
  class Account
  2
  {
  3
  private:
  4
      std::string name;
  5
      double balance;
  6
  public:
  7
      // No default constructor
  8
      Account(std::string name_val, double bal)
  9
      {
  10
          name = name_val;
  11
          balance bal;
  12
      }
  13
      bool withdraw(double amount);
  14
      bool deposit(double amount);
  15
  };
  

  xxxxxxxxxx
  7
  1
  Account jack_account;                   // Error
  2
  Account yena_account;                   // Error
  3
  
  4
  Account *sunny_account = new Account;   // Error
  5
  delete sunny_account;
  6
  
  7
  Account kj_account {"KJ", 15000.0};     // OK
  

  Errors because the default constructor is no longer automatically generated by the compiler.

Overloading Constructor

• Classes can have as many constructors as necessary

• Each must have a unique signature

• Default constructor is no longer compiler-generated once another constructor is declared

• It is important that we initialize our objects to some state that we know. We don't want garbage data in any of the attributes after they are created.

Examples

• Player class

  xxxxxxxxxx
  33
  1
  class player
  2
  {
  3
  private:
  4
      std::string name;
  5
      int health;
  6
      int xp;
  7
  public:
  8
      // Overloaded constructors
  9
      Player();
  10
      Player(std::string name_val);
  11
      Player(std::string name_val, int health_val, int xp_val);
  12
  };
  13
  
  14
  Player::Player()
  15
  {
  16
      name = "None";
  17
      health = 0;
  18
      xp = 0;
  19
  }
  20
  
  21
  Player::Player(std::string name_val)
  22
  {
  23
      name = name_val;
  24
      health = 0;
  25
      xp = 0;
  26
  }
  27
  
  28
  Player::Player(std::string name_val, int health_val, int xp_val)
  29
  {
  30
      name = name_val;
  31
      health = health_val;
  32
      xp = xp_val;
  33
  }
  

  L16: For the record, by the time that we get to this statement, the name string object has already been constructed and initialized to an empty string. So we're really just assigning a new string to it, in this case, to an existing object. This is, in reality, different from initialization.

  $\to$$\to$ See "Constructor initialization lists" section.

  xxxxxxxxxx
  6
  1
  Player empty;               // None, 0, 0
  2
  Player hero("Hero");        // Hero, 0, 0
  3
  Player villain("Villain");  // Villain, 0, 0
  4
  Player frank {"Frank", 100, 4};                 // Frank, 100, 4
  5
  Player *enemy = new Player("Enemy", 1000, 0);   // Enemy, 1000, 0
  6
  delete enemy;
  

Constructor Initialization Lists

• So far, all data member values have been set in the constructor body. While this works, it technically isn't initialization because by the time the constructor body is executed these member attributes have already been created. So, we're really just assigning values to already created attributes.

• Constructor initialization lists help initialize the member data values to specified values BEFORE the constructor body executes. (True initialization!)

• Constructor initialization lists

  • Are more efficient

  • Immediately follows the parameter list

  • Initializes the data members as the object is created

  • Order of initialization is the order of their declarations in the class

    • The order specified in the initialization list does not matter!

Examples

• Player class

  xxxxxxxxxx
  24
  1
  class player
  2
  {
  3
  private:
  4
      std::string name;
  5
      int health;
  6
      int xp;
  7
  public:
  8
      // Overloaded constructors
  9
      Player();
  10
      Player(std::string name_val);
  11
      Player(std::string name_val, int health_val, int xp_val);
  12
  };
  13
  
  14
  Player::Player()
  15
      : name{"None"}, health{0}, sp{0}                    // Initialization list
  16
  { }
  17
  
  18
  Player::Player::Player(std::string name_val)()
  19
      : name{name_val}, health{0}, sp{0}                  // Initialization list
  20
  { }
  21
  
  22
  Player::Player(std::string name_val, int health_val, int xp_val)
  23
      : name{name_val}, health{health_val}, xp{xp_val};   // Initialization list
  24
  { }
  

  Using constructor initialization lists, we can be sure that our data members have been initialized to our own values before any code in the constructor body is executed. $\to$$\to$ True initialization!

Delegating Constructors

• Often the code for constructors is very similar. Many times only the initialization values of the data members is what chages.

• Duplicated code can lead to errors, so we want to minimize the amount of duplicated code in our programs.

• C++ allows delegating constructors

  • Code for one constructor can call another in the initialization list

  • Avoids duplicating code

Examples

• Player class

  xxxxxxxxxx
  24
  1
  class player
  2
  {
  3
  private:
  4
      std::string name;
  5
      int health;
  6
      int xp;
  7
  public:
  8
      // Overloaded constructors
  9
      Player();
  10
      Player(std::string name_val);
  11
      Player(std::string name_val, int health_val, int xp_val);
  12
  };
  13
  
  14
  Player::Player(std::string name_val, int health_val, int xp_val)
  15
      : name{name_val}, health{health_val}, xp{xp_val};   // Initialization list
  16
  { }
  17
  
  18
  Player::Player()
  19
      : Player{"None", 0, 0}      // Delegating constructor
  20
  { }
  21
  
  22
  Player::Player::Player(std::string name_val)()
  23
      : Player{name_val, 0, 0}    // Delegating constructor
  24
  { }
  

  In this case, the no-args constructor is the delegating constructor, since it delegates object initialization to another constructor.

Default Constructor Parameters

• Can often simplify our code and reduce the number of overloaded constructors by providing default constructor parameters.

• Same rules apply as we learned with non-member functions

Examples

• Player class

  We can simply have one constructor that does everything that we did with the three constructors in the previous example.

  xxxxxxxxxx
  21
  1
  class player
  2
  {
  3
  private:
  4
      std::string name;
  5
      int health;
  6
      int xp;
  7
  public:
  8
      // Constructor with default parameter values
  9
      Player(std::string name_val = "None",
  10
             int health_val = 0,
  11
             int xp_val = 0);
  12
  };
  13
  
  14
  Player::Player(std::string name_val, int health_val, int xp_val)
  15
      : name{name_val}, health{health_val}, xp{xp_val}
  16
  { }
  17
  
  18
  Player empty;                   // None, 0, 0
  19
  Player jack{"Jack"};            // Jack, 0, 0
  20
  Player yena{"Yena", 100, 55};   // Yena, 100, 55
  21
  Player hero{"Hero", 100};       // Hero, 100, 0
  

  Notice that you don't supply the default parameters here. You can simply use an initializer list to initialize the class attributes, and the compiler will take care of providing the default values as needed.

  This is very handy and results in less code to write, which means less code to test, which means less code that could have errors.

  Since the single constructor is doing the job of several constructors, we have to be careful not to declare ambiguous constructors. For example, what would happen if you declared a no-args constructor? The compiler will generate an error since it doesn't know which one to call if no arguments were provided when creating an object.

Copy Constructor

• When objects are copied C++ must create a new object from an existing object, and it uses "copy constructor" to do this. (Copy constructors are called whenever the compiler needs to make a copy of an object.)

• When is a copy of an object made?

  • Passing object by value as a parameter

  • Returning an object from a function by value

  • Constructing one object based on another of the same class

• C++ must have a way of accomplishing this so it provides a compiler-defined copy constructor if you don't. (If you don't provide your own way of copying objects by value then the compiler provides a default way of copying objects.)

• Copies the values of each data member to the new object

  • If you do not provide a user-defined copy constructor, C++ will generate a default copy constructor that will do member-wise copy.

• Perfectly fine in many cases, but beware if you have a pointer data member,the pointer will be copied not what it is pointing to.

  • Shallow copy vs. Deep copy

• Best practices

  • Provide a copy constructor when your class has raw pointer members

  • Provide the copy constructor with a const reference parameter

  • Use STL classes as member attributes as they already provide copy constructors

  • Avoid using raw pointer data members if possible or use smart pointers.

• Copy constructor signature

  xxxxxxxxxx
  4
  1
  Type::Type(const Type &source);
  2
  
  3
  Player::Player(const Player &source);
  4
  Account::Account(const Account &source);
  

  Should pass the source object by constant reference!

  • Why pass by reference? $\to$$\to$ If we pass it in by value, then the passed object will get copied along the way, which is the whole point of the copy constructor. We would defeat the purpose and end up with never ending recursive calls.

  • Why constant? $\to$$\to$ Since we are copying the source without modifying it.

Examples

• Pass object by value

  xxxxxxxxxx
  10
  1
  Player hero{"Hero", 100, 20};
  2
  
  3
  void display_player(Player p)
  4
  {
  5
      // p is a COPY of hero in this example (copy constructor does the copy)
  6
      // Use p
  7
      // Destructor for p will be called
  8
  }
  9
  
  10
  display_player(hero);
  

• Return object by value

  xxxxxxxxxx
  9
  1
  Player enemy;
  2
  
  3
  Player create_super_enemy()
  4
  {
  5
      Player an_enemy{"Super Enemy", 100, 100};
  6
      return an_enemy;    // A COPY of an_enemy is returned (copy constructor does the copy)
  7
  }
  8
  
  9
  enemy = create_super_enemy();
  

  Here, a copy of an_enemy object is made by the copy constructor.

• Construct one object based on another

  xxxxxxxxxx
  2
  1
  Player hero{"Hero", 100, 100};
  2
  Player another_hero{hero};          // A COPY of hero is made using the copy constructor
  

• Declaring and implementing a copy constructor

  xxxxxxxxxx
  8
  1
  // Declaration
  2
  Type::Type(const Type &source);
  3
  
  4
  // Implementation
  5
  Type::Type(const Type &source)
  6
  {
  7
      // Code or initialization list to copy the object
  8
  }
  

  xxxxxxxxxx
  12
  1
  // Declaration
  2
  Player::Player(const Player &source);
  3
  Account::Account(const Account &source);
  4
  
  5
  // Implementation
  6
  Player::Player(const Player &source)
  7
      : name{source.name}, health{source.health}, xp{source.xp}
  8
  { }
  9
  
  10
  Account::Account(const Account &source)
  11
      : name{source.name}, balance{source.balance}
  12
  { }
  

   

Move Constructor

• One of C++11's most pervasive features was probably move semantics. And, to really understand move semantics, it's important to be able to tell whether an expression is an l-value or an r-value.

  Rule of thumb:

  • L-value

    When you can refer to an object by name or you can follow a pointer to get to an object, then that object is addressable and it's an l-value.

  • R-value

    Everything else

• In context of move semantics r-values refer to temporary objects that are created by the compiler, and objects returned from methods.

• Sometimes when we execute code the compiler creates unnamed temporary values:

  xxxxxxxxxx
  2
  1
  int total{0};
  2
  total = 100 + 200;
  

  1. 100 + 200 is evaluated and 300 stored in an unnamed temp value (This value is not addressable so it's an r-value.)

  2. The 300 (r-value) is then stored in the variable total (l-value).

  3. Then the temp value is discarded

  The same happens with objects as well. However, with objects there can be a great amount of overhead if copy constructors are called over and over again making copies of these temporary objects. And, when we have raw pointers and we have to do deep copies then the overhead is even greater.

  This is where move semantics and the move constructor comes into picture.

• Sometimes copy constructors are called many times automatically due to the copy semantics of C++.

• Copy constructors doing deep copying can have a significant performance bottleneck.

  • Deep copy must be implemented if a class contains a raw poiner.

  • This is computationally expensive since we have to allocate space for the copy and then copy the data over.

• C++11 introduced move semantics and the move constructor (Extremely efficient!)

• Move constructor moves an object rather than copy it.

• Move constructors are optional but recommended when you have a raw pointer.

  • If you don't provide them, then the copy constructors will be called.

• Copy elision - C++ may optimize copying away completely (RVO - Return Value Optimization)

  • You may not even see the copy constructors being called. If you experience this, it's probably due to something called copy elision.

  • Copy elision is a compiler optimization technique that eliminates unnecessary copying. Compilers are really smart with their optimizations now. And one of the common techniques is called "Return Value Optimization". That's when the compiler generates code that doesn't create a copy every return value from a function making the code much more efficient.

What Does Move Constructor Do?

• Instead of making a deep copy of the resource

  • It simply moves the resource on the heap

  • Simply copies the address of the resource from source to the current object

  • And, nulls out the pointer in the source pointer.

• Very efficient since not doing a copy at all.

R-value References (References to R-values)

• In the context of move semantics, think of r-value references as references to temporary objects.

• Used in moving semantics and perfect forwarding

• Move semantics is all about r-value references

• Used by move constructor and move assignment operator to efficiently move an object rather than copy it.

• R-value reference operator (&&)

  • This is unlike the l-value reference operator (&)

Examples

• R-value references

  xxxxxxxxxx
  5
  1
  // Syntax
  2
  Type::Type(Type &&source);
  3
  
  4
  Player::Player(Player &&source);
  5
  Move::Move(Move &&source);
  

  Difference from copy constructors?

  • There's no const qualifier for the parameter source because we need to modify it in order to null out its pointer.

  • Parameter is an r-value reference

  xxxxxxxxxx
  9
  1
  int x{100};
  2
  
  3
  int &l_ref = x;     // L-value reference (since 'x' is an l-value, addressable, got a name)
  4
  l_ref = 10;         // Change x to 10
  5
  
  6
  int &&r_ref = 200;  // R-value reference
  7
  r_ref = 300;        // Change r_ref to 300 (just changed the temporary variable value!)
  8
  
  9
  int &&x_ref = x;    // Compiler error (Cannot assign an l-value to an r-value reference)
  

• R-value reference parameters

  xxxxxxxxxx
  8
  1
  int x{100};             // x is an l-value
  2
  
  3
  void func(int &&num);   // B
  4
  
  5
  func(200);              // Calls B (200 is an r-value)
  6
  func(x);                // Error (x is an l-value)
  7
                          // - Cannot bind r-value reference of type 'int&&' to 
  8
                          //   an l-value of type 'int'
  

• L-value reference parameters

  xxxxxxxxxx
  8
  1
  int x{100};             // x is an l-value
  2
  
  3
  void func(int &num);    // A
  4
  
  5
  func(x);                // Calls A (x is an l-value)
  6
  func(200);              // Error (200 is an r-value)
  7
                          // - Cannot bind non-const l-value reference of type 'int&' to 
  8
                          //   an r-value of type 'int'
  

• L-value and r-value reference parameters

  xxxxxxxxxx
  7
  1
  int x{100};             // x is an l-value
  2
  
  3
  void func(int &num);    // A
  4
  void func(int &&num);   // B
  5
  
  6
  func(x);                // Calls A (x is an l-value)
  7
  func(200);              // Calls B (200 is an r-value)
  

• Move class WITHOUT move constructor

  xxxxxxxxxx
  18
  1
  class Move
  2
  {
  3
  private:
  4
      int *data;
  5
  public:
  6
      void set_data_value(int d)  { *data = d; }
  7
      int get_data_value()        { return *data }
  8
      Move(int d);                // Constructor
  9
      Move(const Move &source);   // Copy constructor
  10
      ~Move();                    // Destructor
  11
  };
  12
  
  13
  // Copy constructor (Deep copy)
  14
  Move::Move(const Move &source)
  15
  {
  16
      data = new int;
  17
      *data = *source.data;
  18
  }
  

  The move constructor 'steals' the data and nulls out the source pointer.

  Inefficient copying (Copy constructors will be called to copy the temps)

  xxxxxxxxxx
  4
  1
  vector<Move> vec;
  2
  
  3
  vec.push_back(Move{10});
  4
  vec.push_back(Move{20});
  

  Move{10} and Move{20} create temporary objects which are unnamed and r-values. The compiler is going to use the copy constructor to make copies of these.

  Sample output:

  xxxxxxxxxx
  11
  1
  Constructor for: 10
  2
  Constructor for: 10
  3
  Copy constructor (deep copy) for: 10
  4
  Destructor freeing data for: 10
  5
  Constructor for: 20
  6
  Constructor for: 20
  7
  Copy constructor (deep copy) for: 20
  8
  Constructor for: 10
  9
  Copy constructor (deep copy) for: 10
  10
  Destructor freeing data for: 10
  11
  Destructor freeing data for: 20
  

  Copy constructor does several deep copies, which is very inefficient.

• Move class WITH move constructor

  xxxxxxxxxx
  26
  1
  class Move
  2
  {
  3
  private:
  4
      int *data;
  5
  public:
  6
      void set_data_value(int d)  { *data = d; }
  7
      int get_data_value()        { return *data }
  8
      Move(int d);                // Constructor
  9
      Move(const Move &source);   // Copy constructor
  10
      Move(Move &&source);        // Move constructor
  11
      ~Move();                    // Destructor
  12
  };
  13
  
  14
  // Copy constructor (Deep copy)
  15
  Move::Move(const Move &source)
  16
  {
  17
      data = new int;
  18
      *data = *source.data;
  19
  }
  20
  
  21
  // Move constructor
  22
  Move::Move(Move &&source)
  23
      : data{source.data}
  24
  {
  25
      source.data = nullptr;  // Null out source's pointer
  26
  }
  

  Efficient copying (Move constructors will be called for the temp r-values)

  xxxxxxxxxx
  4
  1
  vector<Move> vec;
  2
  
  3
  vec.push_back(Move{10});
  4
  vec.push_back(Move{20});
  

  xxxxxxxxxx
  10
  1
  Constructor for: 10
  2
  Move constructor (moving resource): 10
  3
  Destructor freeing data for nullptr
  4
  Constructor for: 20
  5
  Move constructor (moving resource): 20
  6
  Move constructor (moving resource): 10
  7
  Destructor freeing data for nullptr
  8
  Destructor freeing data for nullptr
  9
  Destructor freeing data for: 10
  10
  Destructor freeing data for: 20
  

  No copy constructor is called at all. Instead, move constructor is being called.

  Also, note that the destructors are called for nullptr. That's destroying the object that we just moved and set its data pointer to nullptr.