Classes inheritance and something that confuses me

nstk

The book is about casting and the author says that the above is legal instead the following is not:

class Employee { }

class ContractEmployee : Employee { }

class CastExample2
{

public static void Main ()
{
    ContractEmployee ce = new Employee(); // Won't compile.
}

}

And this is ok for me, but still I do not understand why is such a declaration useful and what it actually means. If I declare Employee e = new ContractEmployee() and e has the properties of the Employee class, why then not write Employee e = new Employee() as we "normally" do? What is the difference between those two?

Lukasz Nowakowski

You have Employee instance and you assign an instance of the ContractEmployee to it. So you can do something like this (I'll take example from Wayne Gaylard):

class Employee
{
public void Talk()
{
MessageBox.Show("I am an Employee.");
}
}

class ContractEmployee: Employee
{
public void ContractTalk()
{
MessageBox.Show("I am a Contract Employee.");
}
}

Then you can do:

static void CreateEmployee()
{
Employee e = new ContractEmployee();
e.Talk();
((ContractEmployee)e).ContractTalk();
}

But you can't do:

static void CreateEmployee()
{
Employee e = new Employee();
e.Talk();
((ContractEmployee)e).ContractTalk(); // Here you will get an exception at runtime. Compliler won't find any errors.
}

On this example it seems useless, but it is sometimes useful. For example you can have a class structure of different user type, all inheriting from User, you can store them on in a collection of Users

Don't forget to rate answer, that helped you. It will allow other people find their answers faster.

Lost User

As I said before, you need to go back to the book and see where the author is going with this. There are often cases where you would do this when passing an object through some code that knows about Employees but does not know about ContractEmployees. Some common code could adjust the properties of e without needing to know what type of employee the object refers to. At some later point you could upcast e to a ContractEmployee and process as necessary.

The best things in life are not things.

MicroVirus

nstk wrote:

And this is ok for me, but still I do not understand why is such a declaration useful and what it actually means. If I declare
Employee e = new ContractEmployee()
and e has the properties of the Employee class, why then not write
Employee e = new Employee()
as we "normally" do? What is the difference between those two?

The first code line means you create a new instance of class ContractEmployee. This object in memory *is* a ContractEmployee. You then have a variable e which is an Employee. Since a ContractEmployee is specifically also an Employee (ContractEmployee derives from Employee) this is a valid assigment. The difference between the two lines is that in the first, e behaves like a ContractEmployee and in the second e behaves like an Employee. However, through e you can only access members of ContractEmployee which are also part of Employee. The most important thing to distinguish in these examples is the difference between the actual object created (and in memory), which is specified by new, and the interface you are using to access it. The actual object you create is different in both examples, but the interface you use to access them is the same. To properly explain this sample, we'd need to explain almost everything about classes and inheritance, and it seems you're not there yet. So probably, it's best to keep those questions in the back of your head until you get to that subject and ask them again; probably, you'll find you know the answers

Lost User

Nice explanation.

The best things in life are not things.

_Erik_

Let's make a better example for this:

class Employee
{
public string Name { get; set; }
}

class ContractEmployee : Employee
{
public object Contract { get; set; }
}

class Program
{
static void Main()
{
Employee e = new ContractEmployee();
e.Name = "Ricky"; // No problem with this
e.Contract = new object(); // Problem: this does not compile
}
}

nstk wrote:

which properties and methods does the object e has?

e object, at runtime, has the properties and methods defined within ContractEmployee class, because it is the object you have created with new, so it has Name (inherited from Employee class), and Contract (defined within ContractEmployee). However, without any casting, from e object you will only be able to access those members defined within Employee class (Name property in this case), because it is the type you have used to declare the object. Why? Because e object is declared as Employee, and Employee does not define a member named Contract. Yes, e is a ContractEmployee instance and, yes, it has a Contract property, but the compiler does not know it. All the compiler knows is that e object is declared as Employee, so it allows you to access only the members defined in Employee class. So, if you want to access the Contract property of e object, in this case, you would need a previous casting operation:

((ContractEmployee)e).Contract = new object();

// or

ContractEmployee ce = (ContractEmployee)e;
ce.Contract = new object();

nstk wrote:

how should I read this?

Any ContractEmployee object is a Employee, always, so you can declare an Employee and instantiate a ContractEmployee, but not in the other direction, I mean, not all of the Employee objects have to be ContractEmployee.

BobJanova

In short, such a declaration is useful because of polymorphism. Code which does not know about ContractEmployee can still call methods on it if they are overrides of virtual (or abstract or interface-defined) methods of a base class. For example consider the class hierarchy (another common OO tutorial one):

abstract class Animal {
public abstract string Noise { get; }
public virtual string Feed() { return "Mm, tasty."; }
}

class Dog : Animal {
public override string Noise { get { return "Woof."} }
public void GiveBone() { Console.WriteLine("Chomp."); }
}

class Cow : Animal {
public override string Noise { get { return "Moo."; } }
public override string Feed() { return "I'll only eat it if it's green."; }
}

Now, because the methods are virtual, polymorphism will ensure that the appropriate one is called when they are dispatched on an instance of Animal, and 'down-casting' to feed instances to a more generic (in the normal English sense of the word) method is appropriate:

void FeedAndListen(Animal animal){
Console.WriteLine(animal.Feed());
Console.WriteLine(animal.Noise);
// You can't do this, because the animal type is not known
// animal.GiveBone();
}

static void Main(){
List animals = new List<Animal>();
animals.Add(new Dog());
animals.Add(new Cow());
foreach(Animal animal in animals) FeedAndListen(animal);
}

Note how FeedAndListen produces the output from the derived methods, even though the instances are 'downcasted' into a List<Animal> and passed to it as Animal, not their actual declaration class. Also note how you can put the different subclasses into the one list. Finally, see how Dog inherits the Feed behaviour because it is not overridden. And you can't call methods which are not defined on the base class. I agree with the other comment which says that as you complete your OO training you will naturally see how these concepts work.

Ravi Bhavnani

The object refered to by e is a ContractEmployee and therefore contains all the trimmings of a ContractEmployee. However... because e has been declared as an Employee, it exposes only those properties and methods defined by the base class. In other words, the following test will succeed:

Employee e = new ContractEmployee();
Debug.Assert (e is Employee);
Debug.Assert (e is ContractEmployee());

For e to expose its ContractEmployee properties and methods, you'd need to first cast it to a ContractEmployee.

Employee e = new ContractEmployee();
ContractEmployee ce = e as ContractEmployee; // unsafe casting; better to use the cast operator
// (ContractEmployee) which will throw on failure
Debug.Assert (ce != null);

/ravi

My new year resolution: 2048 x 1536 Home | Articles | My .NET bits | Freeware ravib(at)ravib(dot)com

nstk

I thought of it, that maybe it is one of the questions that is going to be answered by itself with time. Nevertheless, thank you for your descriptive answer, it helps a lot. There is however one point regarding the following:

MicroVirus wrote:

The difference between the two lines is that in the first, e behaves like a ContractEmployee and in the second e behaves like an Employee.

In that case shouldn't have been the same if we wrote: ContractEmployee e = new ContractEmployee()

nstk

BobJanova wrote:

Code which does not know about ContractEmployee can still call methods on it if they are overrides of virtual (or abstract or interface-defined) methods of a base class.

This is also an answer that explains a lot. Thank you.

Abhinav S

nstk wrote:

Or maybe it's the opposite that is true?

Yes. It is indeed the opposite that is true. The sub class has access to all the protected members of the parent class but not the other way round. Whenever we say we have a object deriving from something, we are always assuming that this object has properties of the parent and adds on some of its own.

The funniest thing about this particular signature is that by the time you realise it doesn't say anything it's too late to stop reading it.