Assume the following:
class A {
public:
A();
protected:
int a;
};
class B : public A {
public:
B();
};
My questions:
Can I access (directly) the protected data members of A and maybe modify them without having to use public member functions?
Can I treat A inside B as an "existing" object so that I can return A by reference and be able to modify its data member?
I am new to C++. I tried couple things to treat "A" as an object but I keep getting error messages. Here is one thing I have tried:
A & B::getA() {
return A; //error: "A does not refer to a value"
}
Can I access (directly) the protected data members of A and maybe
modify them without having to use public member functions?
A protected member variable or function is very similar to a private member but it provided one additional benefit that they can be accessed in child classes which are called derived classes.
Can I treat A inside B as an "existing" object so that I can return A
by reference and be able to modify its data member?
In B you can directly access to A (public and protected) method and attributes.
Read here for more.
To return a reference to your base class you can write
A & B::getA() {
return *this;
}
to access A's protected data member you can simply write
a = 5;
Related
I have an abstract class base with private member variable base_var. I want to create a derived class, which also has base_var as a private member.
To me, this seems like an obvious thing you would want to do. base is abstract so it will never be instantiated. The only time I will create a base-object is if it is actually a derived object, so obviously when I give ´base´ a private member variable, what I am really trying to do is give that variable to all of its derived objects.
However, the below diagram seems to suggest that this is not doable with inheritance?
Why not? What would then even be the point of having private stuff in an abstract class? That class will never be instantiated, so all that private stuff is essentially useless?
However, the below diagram seems to suggest that this is not doable with inheritance?
Correct, private members of a class can not be accessed by derived classes. If You want a member of a class to be accessible by its derived classes but not by the outside, then You have to make it protected.
Why not? What would then even be the point of having private stuff in an abstract class? That class will never be instantiated, so all that private stuff is essentially useless?
Even an abstract class can have member functions which act on a (private) member variable. Consider (somewhat silly example, but well):
class MaxCached
{
private:
int cache = std::numeric_limits<int>::min();
public:
bool put(int value)
{
if (value > cache)
{
cache = value;
return true;
}
return false;
}
int get() const
{
return cache;
}
virtual void someInterface() const = 0;
};
Deriving from this class gives You the functionality of the base class (put and get) without the danger of breaking it (by for example writing a wrong value to cache).
Side note: Above is a purely made up example! You shouldn't add such a cache (which is independent of Your interface) into the abstract base class. As it stands the example breaks with the "Single Responsibility Principle"!
Just because a class is abstract doesn't mean there cannot be code implemented in that class that might access that variable. When you declare an item in a class to be private, the compiler assumes you had a good reason and will not change the access just because it there is a pure virtual function in the class.
If you want your derived classes to have access to a base class member declare the member as protected.
I have an abstract class base with private member variable base_var
class foo {
public:
virtual void a_pure_virtual_method() = 0;
int get_var() { base_var; }
virtual ~foo(){}
private:
int base_var;
};
Note that a class is said to be abstract when it has at least one pure virtual (aka abstract) method. There is nothing that forbids an abstract class to have non-pure virtual or even non-virtual methods.
I want to create a derived class, which also has base_var as a private member.
class derived : public foo {};
To me, this seems like an obvious thing you would want to do.
Sure, no problem so far.
The only time I will create a base-object is if it is actually a derived object, so obviously when I give ´base´ a private member variable, what I am really trying to do is give that variable to all of its derived objects.
Still fine.
Why not?
You are confusing access rights that are display in the image you included with the mere presence of the members in the derived. The derived class has no access to members that are private in the base class. Period. This is just according to the definition of what is private.
What would then even be the point of having private stuff in an abstract class? That class will never be instantiated, so all that private stuff is essentially useless?
It is not useless at all. Derived classes inherit all members, they just cannot access all of them. The private stuff is there you just cannot access it directly. Thats the whole point of encapsulation. Consider this example:
class bar : public foo {
void test() {
std::cout << base_var; // error base_var is private in foo
std::cout << get_var(); // fine
}
};
Why doesnt the subclasses share the same private membervariable in the superclass using polymoprhism? There are only one instance of the baseclass and if SubA is setting the private member through a mutator - why then cannot SubB access this value. How would it look like if I want the subclasses to share the same private member?
#include <iostream>
class Super {
private:
int cnt;
public:
int getCnt() {
return cnt;
}
void setCnt(int cnt) {
this->cnt = cnt;
}
};
class SubA: public Super {
};
class SubB: public Super {
};
int main() {
Super *super;
SubA a;
SubB b;
super = &a;
super->setCnt(10);
super = &b;
std::cout << super->getCnt() << std::endl;
super = &a;
std::cout << super->getCnt() << std::endl;
return 0;
}
produces:
-8589546555 (garbage)
10
There are only one instance of the baseclass and if SubA
That is wrong. a and b are different objects. They each have an instance of an A sub object. You have not set cnt in b, so it is no surprise that looking at it gives you a garbage value, because reading from an uninitialized object is undefined behaviour.
How would it look like if I want the subclasses to share the same private member?
You could give the base class a static data member. That means all instances of A would share the same member.
Why doesnt the subclasses share the same private membervariable in the superclass using polymoprhism?
Polymorphism has nothing to do with how data is embedded in classes; (run-time) polymorphism is only relevant to virtual dispatch and Run-Time Type Information (RTTI) for dynamic_cast and type_info.
If you imagine your a and b objects on the stack, their memory layout can be illustrated like this:
A:[[Super: int cnt;]A-specific fields (if there were any)]
B:[[Super: int cnt;]B-specific fields (if there were any)]
In effect, class A : public Super is saying "I may want to extend 'Super', optionally appending my own data members to those in Super, possibly adding further functions / overriding virtual ones".
There are only one instance of the baseclass and if SubA is setting the private member through a mutator - why then cannot SubB access this value.
This is wrong... each sub class embeds its own instance of the super class.
How would it look like if I want the subclasses to share the same private member?
Well, there are lots of ways in which you could orchestrate that:
You could have the subclasses hold pointers to a Data object into which you move the superclass's data you want shared, then you'd need a way to initialise the Data object and make it known to all the subclass instances you want to share it.
You could make the superclass data static, which means a single copy of each static variable will be shared by all instances of the superclass, whether embedded in subclasses or not.
Given a class method in C++ that has to be private there, how to make C++/CLI access it. Example: C++ class has a private constructor that shouldn't be exposed to the external user, C++/CLI managed code has to have similar private constructor, but internally it must access private C++ constructor to instantiate a member pointer referring to C++ native class.
Keep in mind, first of all, the goal of making things inaccessible (i.e. private) generally contradicts the goal of making things accessible. If you're able to change the class declaration, then you have some options. I don't know the specifics of your C++/CLI requirements, but perhaps one of these options will give you what you need.
Protected and Inheritance
To an "outsider", private and protected members are equally inaccessible. But to a subclass, private is still private while protected members are accessible. You can gain access to protected members by subclassing. This can be one way to unit test "private" members of a class without fully exposing it to the world.
class A {
protected: // was private
int sum(int a, int b);
};
class TestA : public A {
public:
bool testSum(int expected, int a, int b) {
return expected == sum(a, b);
}
};
So access to protected method sum in class A becomes accessible by making it protected and subclassing. The private member will remain inaccessible to everything else except subclasses.
Static Factory Method
You mentioned a private constructor; if you are needing to manage construction, you might accomplish that with a static factory method. For instance:
class Foobar {
private:
Foobar() { } // Don't let anyone just make one.
public:
static Foobar* Create() { return new Foobar; }
};
// To create a Foobar instance:
Foobar* pfoo = Foobar::Create();
This is not the best interface (better to use a shared_ptr, for instance), but it demonstrates my meaning. Note that since the factory function Create is public, anyone can create one. However, you can change the body of Create to something more complex: include bookkeeping, initialization, etc. The factory method is a way to manage creation, not restrict it. Granted, you may not want this, but it's an option if you need to manage construction.
Friends
C++ permits giving access of the private parts of classes to other classes/functions via the friend keyword.
class A {
int x; //private
friend class B;
};
class B {
void foobar(A& a) {
a.x = 42; // permissible, because A declared class B a friend
}
};
Friendship can be granted to another class or to a function. This can easily break encapsulation and make code difficult to follow or keep safe, but it also gives you the means to expose private members to exactly just those who need it.
Basically as far as I know, when you create a base class with a public, protected, and private section and variables/functions in each the public and protected sections will get inherited into the appropriate section of the sub-class (defined by class subclass : private base, which will take all public and protected members of base and put them into public, changing the word private to public puts them all in public and changing it to protected puts them all into protected).
So, when you create a sub-class you never receive anything from the private section of the previous class (the base class in this case), if this is true then an object of the sub-class should never have it's own version of a private variable or function from the base class correct?
Let's run over an example:
#include <iostream>
class myClass // Creates a class titled myClass with a public section and a private section.
{
public:
void setMyVariable();
int getMyVariable();
private:
int myVariable; // This private member variable should never be inherited.
};
class yourClass : public myClass {}; // Creates a sub-class of myClass that inherits all the public/protected members into the
// public section of yourClass. This should only inherit setMyVariable()
// and getMyVariable() since myVariable is private. This class does not over-ride any
// functions so it should be using the myClass version upon each call using a yourClass
// object. Correct?
int main()
{
myClass myObject; // Creates a myClass object called myObject.
yourClass yourObject; // Creates a yourClass object called yourObject
yourObject.setMyVariable(); // Calls setMyVariable() through yourObject. This in turn calls the myClass version of it because
// there is no function definition for a yourClass version of this function. This means that this
// can indeed access myVariable, but only the myClass version of it (there isn't a yourClass
// version because myVariable is never inherited).
std::cout << yourObject.getMyVariable() << std::endl; // Uses the yourClass version of getMyVariable() which in turn
// calls the myClass version, thus it returns the myClass myVariable
// value. yourClass never has a version of myVariable Correct?
std::cout << myObject.getMyVariable() << std::endl; // Calls the myClass version of getMyVariable() and prints myVariable.
return 0;
}
void myClass::setMyVariable()
{
myVariable = 15; // Sets myVariable in myClass to 15.
}
int myClass::getMyVariable()
{
return myVariable; // Returns myVariable from myClass.
}
Now, in theory based on what I think, this should print:
15
15
Due to it simply always using the myClass version of the functions (thus using the myClass myVariable). But, strangely, this is not the case. The result of running this program prints:
15
0
This makes me wonder, are we actually not only inheriting myVariable, but do we also have the ability to mess around with it? Clearly this is creating an alternate version of myVariable somehow otherwise there wouldn't be a 0 for the myClass version. We are indeed editing a second copy of myVariable by doing all this.
Can someone please explain this all to me, this has torn apart my understanding of inheritance.
Basically as far as I know, when you create a base class with a public, protected, and private section and variables/functions in each the public and protected sections will get inherited into the appropriate section of the sub-class (defined by class subclass : private base, which will take all public and private members of base and put them into public, changing the word private to public puts them all in public and changing it to protected puts them all into protected).
There's a bit of confusion in this statement.
Recall that inheritance is defined for classes and structs in C++. Individual objects (ie. instances) do not inherit from other objects. Constructing an object using other objects is called composition.
When a class inherits from another class, it gets everything from that class, but the access level of the inherited fields may inhibit their use within the inheritor.
Furthermore, there are 3 kinds of inheritance for classes: private (which is the default), protected, and public. Each of them changes the access level of a class properties and methods when inherited by a subclass.
If we order the access levels in this manner: public, protected, private, from the least protected to the most protected, then we can define the inheritance modifiers as raising the access levels of the inherited class fields to at least the level they designate, in the derived class (ie. the class inheriting).
For instance, if class B inherits from class A with the protected inheritance modifier:
class B : protected A { /* ... */ };
then all the fields from A will have at least the protected level in B:
public fields become protected (public level is raised to protected),
protected fields stay protected (same access level, so no modification here),
private fields stay private (the access level is already above the modifier)
"When you create a sub-class you never receive anything from the private section of the [base class]. If this is true then an object of the sub-class should never have it's own version of a private variable or function from the base class, correct?"
No. The derived class inherits all the members of the base class, including the private ones. An object of the inherited class has those private members, but does not have direct access to them. It has access to public members of the base class that may have access to those members, but it (the derived class) may not have new member functions with such access:
class yourClass : public myClass
{
public:
void playByTheRules()
{
setMyVariable(); // perfectly legal, since setMyVariable() is public
}
void tamperWithMyVariable()
{
myVariable = 20; // this is illegal and will cause a compile-time error
}
};
myObject and yourObject are two different objects! Why should they share anything?
Think about it that way: Forget about inheritance and suppose you have a class Person with private int age; and public void setAge (int age) {...}. You then instantiate two objects:
Person bob;
Person bill;
bob.setAge(35);
Would you expect Bill to be 35 now, too? You wouldn't, right? Similarly, your myObject doesn't share its data with yourObject.
In response to your comment:
The class yourClass inherits from myClass. That means that both yourObject and myObject have their own myVariable, the latter obviously by definition, the former inherited from myClass.
Physically, every single member( including member functions) of base class goes into the subclass. Doesn't matter if they are private. Doesn't matter if you inherit them publically/protected-ly/privately. So in your example, yourClass contains all three of getMyVariable(), setMyVariable() and myVariable. All this is pretty simple, okay?
What matters is how we can access them. It is like when a file is deleted on your system. So, you should first understand the difference between a member being not there and a member being there but inaccessible. Assume for now that all inheritance takes place publically. Then, all public members of base class are public in derived class, protected members are protected and private members are inaccessible. They are inaccessible and not non-existent because there can be some member functions in protected and public sections in base class which access the private members of base class. Thus, we need all those private members of base which are accessed by public and protected member functions of base, for their functionality. Since there is no way that we can determine which member is needed by which member function in a simple manner, we include all private members of the base class in derived class. All this simply means that in a derived class, a private member can be modified by only through the base class' member functions.
Note: every private member has to be accessed, directly or indirectly [through another private member function which in turn is called by a public/protected member function] by a public/protected meber function, else it has no use.
So, we know till now that a private member variable of base class has its use in derived class i.e. for the functionality of its public/protected member functions. But they can't be accessed directly in base class.
Now, we turn our attention to private/public inheritance. For public inheritance, it means that all the accessible members of base class (that is, the public and protected members) can not be at a level more permissive than public. Since, public is the most permissive level, public and protected members remain public. But at protected and private inheritance, both become protected and private in the derived class, respectively. Inthe latter case, since all these members are private, they can't be accessed further in the hierarchy chain, but can be accessed by the given derived class all the same.
Thus, the level of each base class member in derived class is the lesser of their level in derived class () and the type of inheritance (public/protected/private).
Same concept applies to the functions outside the class. For them private and protected members are inaccessible but they do exist and can be accessed by the public member functions.
And taking your case as a final example, setMyvariable() and getMyVariable() can access myVariable in the derived class. But no function specified in derived class can access myVariable. Modifying your class:
class myClass
{
public:
void setMyVariable();
int getMyVariable();
private:
int myVariable;
};
class yourClass : public myClass
{
public:
// void yourFunction() { myVariable = 1; }
/*Removing comment creates error; derived class functions can't access myVariable*/
};
Further: you can add exceptions to the type of inheritance too e.g. a private inheritance except a member made public in derived class. But that is another question altogether.
You never call myObject.setMyVariable(), so myObject.getMyVariable() will not return 15.
private does not imply static.
After:
class yourClass : public myClass {};
there is still only one member variable. But there are two ways of accessing it by name: myClass::myVariable, and yourClass::myVariable.
In these expressions, the class name is known as the naming class. The second key thing to understand is that access rights apply to the combination of naming class and member name; not just to the member name and not to the variable itself.
If a member is mentioned without explicitly having the naming class present, then the naming class is inferred from the type of the expression on the left of the . or -> that named the member (with this-> being implied if there is no such expression).
Furthermore, there are really four possible types of access: public, protected, private, and no access. You can't declare a member as having no access, but that situation arises when a private member is inherited.
Applying all this theory to your example:
The name myClass::myVariable is private.
The name yourClass::myVariable is no access.
To reiterate, there is only actually one variable, but it may be named in two different ways, and the access rights differ depending on which name is used.
Finally, back to your original example. myObject and yourObject are different objects. I think what you intended to write, or what you are mentally imagining is actually this situation:
yourClass yourObject;
myClass& myObject = yourObject;
// ^^^
which means myObject names the base class part of yourObject. Then after:
yourObject.setMyVariable();
the variable is set to 15, and so
std::cout << myObject.getMyVariable() << std::endl;
would output 15 because there is indeed only one variable.
This may help
#include<iostream>
using namespace std;
class A
{
int b;
};
class B : private A
{
};
int main()
{
C obj;
cout<<sizeof(obj);
return 0;
}
I have a simple object, of type "ObjectX", with a simple method called "doSomething()". I'd like to make doSomething ONLY accessable by other ObjectX's. In other words, if something that is either static, or not an object of type "ObjectX" tries to call doSomething, it will be unable to. However, if an object of type ObjectX tries to call the method, it WILL be able to.
This differs from a private method, in that, a private method can only be called from the same object it is in. If there were another object of the same type calling that method on a different object, it would be locked out.
private does almost exactly what you want. Only objects of the same type can use private methods, and other objects can call those functions on other objects (that is, private functions are not restricted to the invoking object).
The only thing that is not as you described is that that static functions in the same class can also use private functions. There is no language feature that lets you restrict a function to the object only (excluding static functions).
I don't know where you get your facts. A private function A::foo can be called by any object of type A. Be it on itself or on another instance.
class A
{
public:
void foo(const A& other) { other.priv(); }
private:
void priv() const {}
};
int main()
{
A a1, a2;
a1.foo(a2);
return 0;
}
In C++
Class A{
A a;
doSomething(){
a.doSomething();
}
}
The above code would work.And this is the reason copy constructor work.
Class B{
A a;
doSomethingElse(){
a.doSomething();
}
}
Above would not anyways work.
If you want to provide access to B as well make A a friend of B by specifying friend B; in class A.