class IBase {
public:
virtual void func1() = 0;
virtual void func2() const = 0;
virtual void func3() = 0;
};
class Derived : public IBase {
public:
void func1() { ... }
void func2() const { ... }
void func3() { ... }
void func4() { ... }
};
IBase *b = new Derived();
b->func1(); // ok
b->func2(); // ok
b->func3(); // ok
b->func4(); // compiling error, why ?
delete b;
Inheritance question C++
Author
My first problem is:
My second question, I have heard that avoid multiple inheritance is a good practice, but is a good practice do this for almost all classes ?
class A : public ErrorHandling<A>, public ClassInformation<A>, public InterfaceA {};
ErrorHandling is a class like the one described in effective C++, ClassInformation is a class that have methods to count instances, get class size, get this pointer and such of things.
thnx
1. b is a pointer to IBase. Imagine you had:
Since calling func4 only makes sense for Derived objects it can only be called for objects which are definitely Derived objects, not IBase objects which may not have a func4 member function.
By the way you should have a virtual destructor for IBase if you intend to delete derived objects via a pointer-to-base.
Enigma
class OtherDerived : public IBase{ public: void func1(); void func2() const; void func3(); // no func4};IBase* b = new OtherDerived();b->func1(); // okb->func2(); // okb->func3(); // okb->func4(); // of course it's an error OtherDerived has no func4!delete b;Since calling func4 only makes sense for Derived objects it can only be called for objects which are definitely Derived objects, not IBase objects which may not have a func4 member function.
By the way you should have a virtual destructor for IBase if you intend to delete derived objects via a pointer-to-base.
Enigma
The derived instance HAS all 4 functions, but the compiler is not allowed to KNOW that it has the func4 when using a base class pointer ...
to bring this point home, consider this function:
void TestFunc(IBase *object)
{
object->func3(); // fine
// it should be fairly obvious that this next cannot compile
object->func4();
// because the compile, using STATIC, COMPILE-TIME type checking
// is expecting an object of type IBase, or ANY derived class, not
// a specific derived class
}
and yes, as mentioned before, you NEED a virtual destructor.
since you are doing this:
IBase *object = new Derived();
you have just guaranteed yourself an error if you do not have a virtual destructor.
You can fix it two ways:
Derived *derived = new Derived();
IBase *base = derived;
// do stuff here
delete derived;
OR
add a virtual destructor to IBase
to bring this point home, consider this function:
void TestFunc(IBase *object)
{
object->func3(); // fine
// it should be fairly obvious that this next cannot compile
object->func4();
// because the compile, using STATIC, COMPILE-TIME type checking
// is expecting an object of type IBase, or ANY derived class, not
// a specific derived class
}
and yes, as mentioned before, you NEED a virtual destructor.
since you are doing this:
IBase *object = new Derived();
you have just guaranteed yourself an error if you do not have a virtual destructor.
You can fix it two ways:
Derived *derived = new Derived();
IBase *base = derived;
// do stuff here
delete derived;
OR
add a virtual destructor to IBase
Quote:Original post by Silly_con
My second question, I have heard that avoid multiple inheritance is a good practice, but is a good practice do this for almost all classes ?
No. Multiple inheritance has its issues (even in C++). Problems are diamond hierarchies virtual inheritance and such.
Also you need to carefully decide what you are trying to model.
The class information from your example is rather aggregation (A contains class information and is not class information).
The public interfaceA can also be modelled by the PIMPL idiom a.s.o. In short: there is no single technique that is best suited for all problems. Get familiar with many approaches (functional, policy-based, generic programming, ...) and select the one that is best suited to solve your problem efficiently.
Regards,
Pat.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
