Skip to main content

Multiple Inheritance

In C++,  a class can inherit from more than one class. This is called multiple inheritance.

Some other OOPS languages like Java, do not support multiple inheritance.


Syntax for multiple inheritance is

class cls-name:public base-cls1,public base-cls2,public base-cls3
{
/*code*/
}

We are specifying multiple classes as base classes here. The base class list is comma separated and has access specifier for each base class. Each of these base classes must have been declared earlier.

Derived class objects contain sub-objects of each of these base classes.
class LandAnimal
{
    int legs;
 public:
 void walk()
 {
 cout<<"walks";
 }
};
class WaterAnimal
{
   int fins;
 public:
 void swim()
 {
 cout<<"swims";
 }
};
class Amphibian:public LandAnimal,public WaterAnimal
{
};
int main()
{
 Amphibian frog;
 frog.walk();
 frog.swim();
} 

Output
walks
swims

In the above example, class Amphibian inherits from 2 classes - LandAnimal and WaterAnimal.

The derived class has members of both these base classes. It can call both walk() and swim() functions. We construct an object of this class and it has both data members legs and fins and it has functions swim and walk.

 Size of a Ambphibian object = size of LandAnimal+size of WaterAnimal. 

 Constructors 

The derived class constructors and destructors in multiple inheritance will call all of the base class constructors and destructors.  The order of constructor calls is same as declaration order of base classes.

Amphibian constructor in the earlier example, will call LandAnimal constructor first, then WaterAnimal constructor and finally Amphibian constructor. And order of destructor calls is reverse of order of constructor calls. 

class A
{
 int n1; int n2;
public: 
 A(int a,int b):n1(a),n2(b)
 {cout<<"A ctor ";}
};
class B
{
 int n3; int n4;
public: 
 B(int a,int b):n3(a),n4(b)
 {cout<<"B ctor ";}
};
class MI:public B, public A
{
 int m; 
public:
 MI(int num=0):m(num),A(num,num),B(num,num)
 {cout<<"Derived ctor";}
};
int main(int argc, char **argv)
{
 MI obj1(10); 
}

output of program
B ctor A ctor Derived ctor

Here MI class declaration uses B followed by A in base class specification. So B ctor is called before A ctor when MI object is created even though in member initialization list, A class ctor is called first.

Dreaded Diamond problem

When there are multilevel and multiple inheritances together, derived class may contain more than one instance of an ancestor class. In such a situation, using a member from this ancestor class throws an ambiguity error.

class A
{
 int m;
public:
 void setm(int a){m = a;};
 int getm(){return m;}
 void print(){cout<<"m"<<m<<endl;}
};
class B: public A
{
public:
 B(){setm(10);};
};
class C:public A
{
public:
 C() {setm(20);};
};
class D:public B,public C
{
 /*multiple inherited class*/
};
int main()
{
 D obj;
 cout<<obj.getm();/*error*/
 obj.print();/*error*/
}
When we compile this program obj.print() and obj.getm() throw ambiguity error.

obj has members of A class duplicated one via B class and another via C class.

When calling print() or getm() , compiler is unable to decide whether to call these functions of B class or  of C class.

This is called dreaded diamond problem because  shape of inheritance diagram in this case looks like a diamond.

To avoid this error, virtual inheritance is used. 

Virtual Inheritance


In virtual inheritance, when there are multiple paths to a base class, this base class object is included only once. 

Virtual inheritance is implemented using virtual keyword before base class name in derived class definition.
class A
{
 int m;
public:
 int getm(){return m;}
 void setm(int a){m = a;}
 void print(){cout<<"A print";}
};
class B:virtual public A
{
/***code**/
};
class C:public virtual A
{
/***code**/
};
class D:public B,public C
{
/**this has only one sub object of A*/
};
int main()
{
 D obj;
 obj.print();/*No error*/
}

Both B and C classes use virtual inheritance.  By using virtual keyword, we are ensuring that  only one copy of a A class's member variables are inherited by grandchild derived classes.

So D which inherits from both B and C, it is inheriting only one sub-object of A class. So there is only one data member called m in D class and only one print() function. There will not be any ambiguity when any member of A is being used from D class object because D object will have only one sub-object of A class.

Also you can observe that obj  causes  A class constructor to be called only once and A destructor to be called only once.

Comments

Popular posts from this blog

Ten questions in C/C++

Let us see some questions in C and C++ Write printf statement in C to print - I got 98% in Maths Can you execute a function before main() in C? If yes, how is it done? Can you write a program to find if a number is even without using modulo operator? How do you define a data member which is common to all objects of a class in C++? Can we have a single constructor for a class, but still create objects from the class passing zero/one and two parameters? What problems might occur if a class has no default constructor? Is the following statement correct? fprintf(stdout,"Hello world");   Why do we use the following statement in C++ program? using namespace std;  Can you write a single statement to check if the number is a power of 2?  What does the following statement mean in C/C++?4 if(a) b++;    So we have 10 questions. How many of these can you answer?   Do you need more questions in C and C++?    You can find t...

It is a constant

In good old days, C programmers would use preprocessor directive to define constants. e.g. #define s 10 But we know now that, as compiler never gets to see these, preprocessor statements are error prone. Hence we have const s. A constant - defined with keyword const promises that this entity is never going to change. And if we accidentally modify a const, compiler throws an error. Let us look at an example. #include<iostream> using namespace std; int main () { int a = 10 ; const int b = 12 ; a ++ ; b = 18 ; } When we compile this program, compiler tells us that default.cpp: In function ‘int main()’: default.cpp:8:7: error: assignment of read-only variable ‘b’      b = 18; So it is catching the error that we are trying to modify a const. Whenever a local variable or parameter need not be modified, declare it as a const.  Yes, we can make even parameters as constant. We can make objects constant or even me...

Abstract class

 If we can not create any objects of a class, then it is called an abstract class. A class is made abstract by adding at least one pure virtual function to it. Pure virtual function A function is said to be a pure virtual function , if it has no definition but has only declaration. Pure virtual function is defined with the keyword virtual and followed by return type, function name and "=0". class Shape { public: virtual void printarea() =0 ; }; int main () { Shape obj1; //error } Here printarea() function of Shape class is a pure virtual function as it has no body. To make a function as pure virtual function, you should use =0 at the end of virtual function declaration Abstract class When a class has at least one pure virtual function, it is incomplete and no objects can be created from that class. Such a class is called an abstract class . In the earlier example class Shape is an abstract class, and objects can not be created from t...