Re: Un-ringing the bell: making parent methods unavailable to children
From: Ronald Laeremans [MSFT] (ronaldl_at_online.microsoft.com)
Date: 12/04/04
- Next message: Peteroid: "Why don't Managed Classes support: const, virtual, and static?"
- Previous message: Altman: "Re: Mysteriously changing variable"
- In reply to: Peteroid: "Re: Un-ringing the bell: making parent methods unavailable to children"
- Messages sorted by: [ date ] [ thread ]
Date: Fri, 3 Dec 2004 16:44:55 -0800
Your example did not have the method as virtual.
My assumption was that you were asking for a method that would make it
impossible for an instance of a derived class to call it, not just less
convenient. And the former cannot be done by any method I can think of. Of
course making the method private in the derived class will prevent it from
being called directly.
Thanks.
Ronald
"Peteroid" <peter_oliphant@msn.com> wrote in message
news:ubXNJeN2EHA.2608@TK2MSFTNGP10.phx.gbl...
>I see your point, but it does require going out of your way by virtue of
> 'tricking' the compiler into treating it as an instance of its parent
> (i.e.,
> you have to know you are doing it purposefully, you can't 'accidentally'
> use
> Add() from an instance of the child class). I tend to avoid pointers and
> typecasting in my code anyway, they can get you into this kind of trouble
> (i.e., shoot yourself in the foot by forcing the compiler to treat a
> pointer
> to one type of data structure as a pointer to another type via overriding
> typecasts)...
>
> So, based on your example, technically a child instance still can't use
> Add() unless it is type case as its parent class. And even then, won't the
> virtual function table still use the child's version of Add() via
> polymorphism? If not, the trick you mentioned would also override
> polymorphism in general. Note that in any case that Add() is polymorphed
> into a NOP for safety, and I could put an 'assert(false) inside it to make
> sure it told me at run time if it ever tried to use it (as a last ditch
> protection, having failed discovering problem at compile time) .
>
> Do you have a better way of making a parent public method unavailable to a
> child instance? Love to hear it... :)
>
> [==Peteroid==]
>
>
> "Ronald Laeremans [MSFT]" <ronaldl@online.microsoft.com> wrote in message
> news:%23jejl$x1EHA.3816@TK2MSFTNGP09.phx.gbl...
>> This doesn't work. See below.
>>
>> class Parent_Class
>> {
>> public:
>> void Add( ) {/* do something */};
>> } ;
>>
>> class Child_Class : public Parent_Class
>> {
>> private:
>> void Add( ) {}; // now child instance can't call Add( ) publicly
>> } ;
>>
>> int main()
>> {
>> Child_Class ck;
>> ((Parent_Class*)&ck)->Add();
>> }
>>
>> Ronald Laeremans
>> Visual C++ team
>> "Peteroid" <peter_oliphant@msn.com> wrote in message
>> news:%23RILLCY1EHA.1152@TK2MSFTNGP14.phx.gbl...
>> > Is it possible to make a public parent class method unavailable (i.e.,
>> > generate an error at compile time) to a particular child class?
>> >
>> > For example, say a parent class has a public method Add( ). I want to
>> > create
>> > a child class of this parent class that does not have an Add( ) method
>> > (while possibly another child class does).
>> >
>> > I think I figured it out while writing this, so tell me if this is the
>> > 'standard method'. Make the parent class public method virtual, and
>> > then
>> > polymorph it in the child class as private. That is:
>> >
>> > class Parent_Class
>> > {
>> > public:
>> > Add( ) { // do something // }
>> > } ;
>> >
>> > class Child_Class : public Parent_Class
>> > {
>> > private:
>> > Add( ) {} // now child instance can't call Add( ) publically
>> > } ;
>> >
>> > [==Peteroid==]
>> >
>> >
>>
>>
>
>
- Next message: Peteroid: "Why don't Managed Classes support: const, virtual, and static?"
- Previous message: Altman: "Re: Mysteriously changing variable"
- In reply to: Peteroid: "Re: Un-ringing the bell: making parent methods unavailable to children"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
