1 minute read

One use of const_cast is to avoid duplicating code when implementing getters in a class. If you want users of your class to have write-access to a member object, then in order to provide write-access and read-only access to the member you might do something like:

class Texture
{
    // lots of stuff
}

class Player
{
    Texture _texture;

public:
    
    const Texture& texture() const { return _texture; }
    Texture& texture() { return _texture; }
};

But we would like to avoid the code duplication.

We have two options. We can implement the const version and do a cast in the non-const version like so:

Option 1

const Texture& texture() const 
{ 
	return _texture; 
}

Texture& texture() 
{
    return const_cast<Texture&>(static_cast<const Player&>(*this).texture());
}

Or we can do the opposite, we can implement the non-const version and cast in the const version. Something like:

Option 2

const Texture& texture() const
{ 
    return (const_cast<Person&>(*this)).texture();
}

Texture& texture() 
{ 
	return _texture; 
}

When I first saw this, my gut reaction was to go with Option 2 mostly because casting away the const felt wrong. However, what I didn’t immediately realize is that Option 2 is calling a non-const function within a const context, which is even worse!. For example, imagine someone came along later and changed the non-const version of texture() to modify some internal state of the object.

Because of this Option 1 is preferred.

Note: This topic is discussed in Item 3 in Effective C++, Third Edition by Scott Meyers

Tags:

Categories:

Updated: