Pure virtual function problem with G++

Darwish · 07-05-2005, 05:19 PM

I have a problem when defining a pure virtual function in G++
here's a sample code:

virtual void print() = 0
{ cout << "Test" << endl; }

the code works without a glitch in VC++..

, but when compiling under GCC it prints the error message:

error: Function definition with a pure virtual function..

Is pure virtual function definitions in G++ are prohibited?!!
and if it is prohibited is there another approach

Thanks for any help..

carl.waldbieser · 07-06-2005, 12:09 AM

I am using gcc version 3.3.5 (Debian 1:3.3.5-8ubuntu2). The following compiles cleanly for me:

Code:

#include <iostream>

class Foo
{
public:
	virtual void bar()=0;
};

void Foo::bar()
{
	std::cout << "Whee!\n";
}

class Baz : public Foo
{
public:
	virtual void bar() { std::cout << "Frotz!\n"; }
};

int main() 
{
	Foo* pfoo = new Baz();
	pfoo->bar();
	delete pfoo;

	return 0;
}

Output is "Frotz!".

Not sure what the syntax for declaring the body inline is

In case anyone is wondering, this example is somewhat nonsensical, as Foo::bar() can never be called because class Foo can never be instantiated. The body is somewhat meaningless. Scott Meyers demonstrates a use for a pure virtual function body in Item 36 of "Effective C++" (2nd Ed.).

Darwish · 07-06-2005, 07:37 AM

Thanks alot..
I understood from ur code to define the pure virtual function outside the class and this method worked exactly as I need....

ta0kira · 07-08-2005, 07:04 AM

Also you can't define a pure virtual function; if you use "= 0" then you can't give it a definition, you can only declare it. Not only did carl.waldbieser put the function in the class definition, he also took out the "= 0".
ta0kira

PS Nevermind. Didn't look at it close enough the first time. Sorry!

lowpro2k3 · 07-08-2005, 08:19 PM

I wasn't aware you could provide a body for a pure virtual function. I'm trying to visualize an example of using one, Carl I have the book "Effective C++: Second Edition", is this the book you're talking about? Besides the limited polymorphism I've used pure virtual functions for I have barely used them before. You guys might have inspired me to write some code tonite

carl.waldbieser · 07-08-2005, 09:03 PM

Yes, that is the book I referred to. It is somewhat surprising and counterintuitive that the standard actually *does* allow you to define a body for a pure virtual function. Typically, pure virtual functions are used to define protocol classes (synonymous with "interfaces" in other languages), so the idea of having a body is somewhat of an oddity. Meyers demonstrates how you can use them to prevent pollution of a class namespace. I am not sure what kinds of practical applicability pure virtual function bodies have-- to me they seem about as obscure as the comma operator.

ta0kira · 07-09-2005, 08:06 AM

I can't see where the defined pure virtual function would ever get called unless you were to take a member pointer to it and explicitly call that, or otherwise fully qualify the function name when making the call. You can't even instantiate a derivative of the class without redefining it, thereby covering it up. Maybe this is how it would be used:

Code:

struct Base
{
	virtual
		void
	Function() = 0;
};

void Base::Function() {}

struct Derived :
	public Base
{
		void
	Function();
};

void Derived::Function()
{ Base::Function(); }

int main()
{
	//Base MyBase;
	Derived MyDerived;
}

ta0kira