TITLE: a case for protected virtual functions

(Source: comp.lang.c++.moderated, 12 Oct 2000)

KANZE: James Kanze <James.Kanze@dresdner-bank.de>

>>>> That's interesting, because in new code, all of my virtual
>>>> functions are protected.  A virtual function should never be
>>>> public, of course.

MORENO: Carlos Moreno

>>> WHAT?!!!  So much for virtual destructors, for instance...

KANZE:

>> You've hit upon the exception, of course:-).  I don't tend to
>> think of destructors as functions, but of course, the destructor
>> normally will be public, and normally will be virtual.

DIMOV: Peter Dimov

> Another exception: interfaces.
>
> class abstract_base
> {
> public:
>     virtual void method1() = 0;
> };

KANZE:

No exception here.  In most cases, the interface should be more or less:

    class Interface
    {
    public:
        void                fnc()
        {
            assert( anyPreconditions ) ;
            doFnc() ;
            assert( anyPostconditions ) ;
        }
    protected:
        virtual void        doFnc() = 0 ;
    } ;

At an interface between subsystems, it is also not uncommon to insert
logging code in the non-virtual public member.

There are exceptions to this, however, as I have posted in response to
a posting from Gerhard Menzl.



_______________________________________________
cpptips mailing list
http://cpptips.hyperformix.com