pure virtual

[vasanth kumar]

I am writing a small example below. I want to know, is this allowed as per
the standard C++. We make an abstract class by making any membor method as
pure virtual. when we make it pure virtual, what is the meaning of providing
definition as shown below. are the code lines in bold are good practice.

class Vehicle {
public:
virtual void startEngine() = 0;
virtual ~Vehicle() {};
};
void Vehicle::startEngine()
{
}
class Car : public Vehicle {
public:
void startEngine(){
Vehicle::startEngine();
}
};

void main()
{
Car objCar=new Car();
objCar.Vehicle::startEngine();
}

Thanks,
Vasanth

 

[Sharad Kala]

I am writing a small example below. I want to know, is this allowed as per
the standard C++. We make an abstract class by making any membor method as
pure virtual. when we make it pure virtual, what is the meaning of providing
definition as shown below. are the code lines in bold are good practice.

Providing definition for pure virtual functions is legal. It's not a
question of good/bad practice but more to do with your requirements.
Consider this code -
struct base
{
// ...
virtual ~base() = 0; //Definition required
};

struct derived : base
{

};

int main()
{
derived d;
}; // Linker error

void main()

Always, int main ()

Sharad

 

[vasanth kumar]

thanks for the reply.

yes, destructor case I understand. provides default destructor. can anyone
think of a requirement for providing this definition for other member
functions.

-Vasanth
-------------

 

[Sharad Kala]

thanks for the reply.

You are welcome, please don't top-post.

yes, destructor case I understand. provides default destructor. can anyone
think of a requirement for providing this definition for other member
functions.

Sure, Herb Sutter gives 3 cases in which pure virtual functions could have a
body - http://www.gotw.ca/gotw/031.htm. He also says that the destructor
example is the most common.

Sharad