14 replies to this topic

[lride]

[source lang="cpp"]class Menu{ private: Menu * subMenu;.....};[/source]

is this possible?

[saejox]

Menu in your example does not contain another instance of Menu. It holds a pointer to a Menu object. Therefore possible.

[Álvaro]

Moreover, this kind of thing is common when you program some data structures, most directly a linked list.

[D.Chhetri]

[source lang="cpp"]class Menu{private: Menu * subMenu;.....};[/source]

is this possible?

Did you try to compile this? Simple question like this could just be tested. Learn to be more independent.

[lride]

I of course tried this, but I asked this because I didn't know what to put in the constructor
[source lang="cpp"]Menu::Menu(){subMenu=new Menu()//??}[/source]

Edited by lride, 28 October 2012 - 09:12 AM.

[Brother Bob]

We don't know what you need to do in your constructor either. But I can tell you one thing: you cannot allocate a new menu unconditionally in each menu, because each menu will contain a menu, which will contain a menu, which will contain a menu, which will contain a menu, which will... ad infinitum. Whatever you do, you must have some logic that controls when the chain breaks; that is, which menus don't contain a submenu.

[lride]

thanks, I see

[Narf the Mouse]

We don't know what you need to do in your constructor either. But I can tell you one thing: you cannot allocate a new menu unconditionally in each menu, because each menu will contain a menu, which will contain a menu, which will contain a menu, which will contain a menu, which will... ad infinitum. Whatever you do, you must have some logic that controls when the chain breaks; that is, which menus don't contain a submenu.

This is a form of endless loop. Any loop that does not have a 100% valid, viable and usable exit condition is very bad.

[_Sauce_]

This is a form of endless loop. Any loop that does not have a 100% valid, viable and usable exit condition is very bad.

Technically this would be recursion, and you would overflow the stack long before you run out of memory allocating Menu objects, unless Menu allocates hundreds of megabytes of data in addition to each child Menu.

[ATC]

I of course tried this, but I asked this because I didn't know what to put in the constructor
[source lang="cpp"]Menu::Menu(){subMenu=new Menu()//??}[/source]

If you define an object type which invokes its own constructor within its constructor (like you've done above) it creates an infinite recursion, which is similar to an endless loop. Example ::

[source lang="cpp"]class MyClass{public: MyClass() { myClass = new MyClass(); } MyClass *myClass;};void main() { MyClass* mc = new MyClass;}[/source]

As soon as the "main" function of this program starts it creates a new MyClass instance and invokes its constructor with the new operator... and then what happens? Control jumps to the MyClass constructor, MyClass::MyClass()... but the code in the constructor AGAIN invokes the constructor with the new operator, in an attempt to instantiate the myClass instance... but that myClass instance will have its own myClass instance too lol... So the constructor is invoked again, and again and again and never stops... at least not until you terminate the program! ;-)

A better way to handle this dilemma is allow the constructor in your program to take a Menu* parameter for its sub-menu... let outside code create the sub-menu (if the menu has one) and pass it to the new, primary menu instance. Alternatively you could offer a static (or even an instance) method in the Menu class, like [static] Menu::AddSubMenu(Menu* primary, <creation params> ) or [instance] Menu::AddSubMenu(...) and let that method handle adding the sub-menu to the primary menu...

Edited by ATC, 29 October 2012 - 11:56 PM.

[rip-off]

Did you try to compile this? Simple question like this could just be tested. Learn to be more independent.

There is no harm in asking the question. Just because something can be compiled and appear to work doesn't mean it is correct.

[lride]

Did you try to compile this? Simple question like this could just be tested. Learn to be more independent.

There is no harm in asking the question. Just because something can be compiled and appear to work doesn't mean it is correct.

This is one of the reason I asked the question. In C++, there are many dangerous things that the compiler allows but never gives a warning about.

Edited by lride, 30 October 2012 - 01:20 PM.

[Aardvajk]

You probably want something like:

class Menu

{

public:

    Menu() : subMenu(0) { }

    ~Menu(){ delete subMenu; }



    void

    setSubMenu(Menu *m){ subMenu = m; }



private:

    Menu *subMenu;

};



void f()

{

    Menu *mainMenu = new Menu();

    mainMenu->setSubMenu(new Menu());

}

In other words, set the submenu pointer to 0 in the constructor and assign it elsewhere.

[Felix Ungman]

In theory, you could actually recursively include objects using some funky C++ template programming.

template <int level>
struct Menu
{
	Menu<level + 1> subMenu;
    ...
};

template<>
struct Menu<3> // hierarchy ends here
{
    ...
};

static void foo()
{
	Menu<0> menu;
	menu.subMenu.subMenu.subMenu; // ok
	menu.subMenu.subMenu.subMenu.subMenu; // error, no such member
}

Not that this is the way to do it, but it illustrates that C++ is flexible enough to do this kind of perverted solution.

[achild]

An alternative type of having an instance in itself (I realize this isn't exactly what you're trying to do)

class Profiler {
public:
Profiler();
~Profiler() { }
// This gives us a global version without limiting us to only 1.  We can make other Profiler
// objects if we want.
static Profiler &GetGlobalInstance() {
  static Profiler prof;
  return prof;
}
[...] stuff
};

Then, you know, modify it accordingly if you need an instance per thread, or if you need to ensure synchronization is handled, or whatever.

Edited by achild, 30 October 2012 - 11:53 AM.