Ce cours introduit la programmation orientée objet (encapsulation, abstration, héritage, polymorphisme) en l'illustrant en langage Java. Il présuppose connues les bases de la programmation (variables, types, boucles, fonctions, ...). Il est conçu comme la suite du cours « Initiation à la programmation (en Java) ». Comme son prédécesseur, ce cours s'appuie sur de nombreux éléments pédagogiques : vidéos sous-titrées, quizz dans et hors vidéos, exercices, devoirs notés automatiquement, notes de cours.
Héritage et polymorphisme : compléments
Loading...
來自 École Polytechnique Fédérale de Lausanne 的課程
Introduction à la programmation orientée objet (en Java)
148 個評分
École Polytechnique Fédérale de Lausanne
148 個評分
從本節課中
Polymorphisme (modificateur abstract) ; le modificateur final
Cette semaine nous étudions la notion de classe abstraite et approfondissons les notions d'héritage et de polymorphisme. Nous abordons aussi une notion complémentaire qui n'est pas directement liée à ces thèmes, le modificateur final.
與講師見面
Jamila SamDr
School of Computer and Communication Sciences
Jean-Cédric ChappelierDr.
School of Computer and Communication Sciences
The objective of this episode is to present various small complements
related to polymorphism and inheritance.
The first subject we will cover concerns polymorphism
and how it applies to the construction of objects.
Constructors are somewhat special methods
in the sense that they are specifically dedicated
to constructing the current instance of a class.
They are not expected to have a polymorphic behavior.
Imagining a polymorphic constructor would mean that it
would be dedicated to initializing objects other than the current instance;
objects of subclass type,
for example, which doesn't really make sense.
However, nothing stops us from invoking a polymorphic method
in the body of a constructor.
But we advise against this.
Indeed, as we will see in the next example,
the method then acts on an object
which may be only partially initialized.
In this example, we have a superclass <i>A</i>
which contains a default constructor as well as an abstract method,
that it, a method without a body, <i>m</i>.
It so happens that the constructor invokes this method <i>m</i>.
Remember that an abstract class can never be instantiated
meaning that we can never call the constructor
of the class <i>A</i> like this,
since the class is abstract.
Invoking an abstract method in <i>A</i>'s constructor,
that is, invoking a method with no body, poses no problem
since we will never call this constructor
to instantiate an object of type <i>A</i>,
that is, an object for which <i>m</i> has no concrete definition.
Thus, this is completely legal.
Essentially, the method <i>m</i> will be invoked only if the constructor of a subclass
invokes the constructor of the superclass
and if this subclass has a concrete definition of the method <i>m</i>.
We have an example of this here, in the class <i>B</i>.
Class <i>B</i> inherits from <i>A</i> and overrides the method <i>m</i>.
The overriding simply consists in displaying a message
containing the value of the attribute <i>b</i> of class B. Class <i>B</i> can be instantiated,
since it overrides all the abstract methods inherited,
in this case, the method <i>m</i> only.
Now, let's see what happens when we create an instance of <i>B</i>.
The instance is created here using the default constructor for the class <i>B</i>.
We know that any constructor from a subclass must invoke
a constructor from the superclass when there is no explicit call
to the constructor via the <i>super</i> syntax,
so we know that there is an implicit call to the default constructor.
This call is made to initialize a current instance
which is a <i>B</i>.
At the moment <i>m</i> is called in <i>A</i>'s constructor,
since there is necessarily
dynamic binding in Java, the method will be chosen
based on the real nature of the instance
and so it is this method <i>m</i> which will be called.
Remember that at this stage,
we have not yet evaluated the instruction
which initializes the attribute <i>b</i> of <i>B</i> with a specific value.
As such, the attribute <i>b</i> has the value which is given to it by default
before any explicit initialization, which is the value zero.
This means that when the method <i>m</i> is executed, the attribute <i>b</i> has a value of zero
and the construction of a <i>B</i> will result in displaying the message
"b vaut: 0".
If for us, the object <i>B</i> is properly initialized only if its <i>b</i> attribute is worth 1,
then we can clearly see that this method <i>m</i> is working on an object
which is partially initialized.
Hence the original piece of advice: do not invoke polymorphic methods
in the body of constructors.
Let's move on to the second subject, completely independent from the first.
You no doubt remember the <i>toString</i> and <i>equals</i> methods that you learned
to write for classes.
But where exactly do these methods come from?
This is what we will see.
Remember, for example, the <i>toString</i> method.
It was imposed to you with a specific header.
It allowed, when it was present in a class, to define
a display format for the object in the form of a String.
This <i>toString</i> method allows us to produce more explicit displays
for objects.
Yet, <i>toString</i> is simply the overriding of a method
existing higher up in the class hierarchy.
But what does higher up mean since here, for example,
the <i>Rectangle</i> class inherits from nobody.
True, it inherits from nobody <i>explicitly</i>,
but in Java, every class that you write
inherits from a universal superclass which is the <i>Object</i> superclass.
And this happens without you needing to explicitly write an inheritance link.
Writing this is unnecessary,
but the inheritance link does indeed exist.
So in Java, every class that you define,
if it inherits from no class explicitly, inherits from <i>Object</i>.
Thus, we can affect an instance of any class
to a variable of type <i>Object</i>.
The <i>Object</i> superclass contains, among other things, the default definitions
for a certain number of methods useful to all possible objects,
such as <i>toString</i>, <i>equals</i> or <i>clone</i>.
For example, the default definition of <i>toString</i> allows one
to display objects in the form of a representation of their references,
that is, something a little strange that we have already
encountered previously.
The default definition of <i>equals</i>, located in the <i>Object</i> superclass,
simply compares references by comparing 2 objects using <i>==</i>.
In most cases, these default definitions
do not satisfy the needs for your own classes
and you are thus led to redefine them,
to allow a more explicit display, for example,
or a comparison which compares the contents, or a correct copy of your objects.
When we wrote the <i>toString</i> method
in the Rectangle class of our previous example,
we were actually overriding
the <i>toString</i> method inherited from <i>Object</i>.
Many predefined classes override these methods
and typically, the <i>String</i> class overrides the <i>toString</i> method,
as well as the <i>equals</i> method.
Let's go back to the case of the <i>equals</i> method as we wrote it up until now,
in the previous episodes.
In your opinion, does this header for the <i>equals</i> method
correspond to a redefinition of the one inherited from <i>Object</i>?
The answer is, of course, no. The <i>Object</i> class cannot
know of all the subclasses
which will be derived from it.
The header for the <i>equals</i> method as it exists
in the <i>Object</i> superclass is as follows:
this means, essentially, that the argument expected
by <i>equals</i> is of type <i>Object</i>,
and not of type <i>UneClasse</i> (TN: means "A Class") as we have seen up until now.
Thus, in the method <i>equals</i> of the Rectangle class, we had
a parameter of type Rectangle.
The <i>equals</i> methods that we have written
up until now are thus not overrides
of the <i>equals</i> method inherited from <i>Object</i>; rather, they are overloaded methods.
We speak of redefining, or overriding,
when a subclass's method
defines a method already present in the superclass,
a method with exactly the same name, list of parameters
and identical types, and with compatible return types.
What we mean by compatible is: for basic types,
they must be identical; for types defined using classes,
these types will be compatible if there is an inheritance link.
For example, if I have a superclass <i>A</i>
defining a method <i>m</i>,
returning an object of type <i>A</i>
If there exists a class <i>B</i> inheriting from <i>A</i>,
which defines a method <i>m</i>, but which returns a <i>B</i>,
since <i>B</i> inherits from <i>A</i>, we consider that the types are compatible in this case,
and so this is also considered as an override.
If these conditions are not met and we only have
the name of the method which is the same, we speak of overloading.
Defining the method <i>equals</i> by overloading, as we did
in the previous episodes, can work flawlessly,
however in Java, it is usually recommended to use overriding
because some methods of the Java API,
typically methods which work
on so-called collections,
for example methods that will retrieve
a value from a collection, will implicitly use
the <i>equals</i> method, which has this exact header.
If it is not present in one of your classes
and you use this method,
then it is the default method, inherited from <i>Object</i>, that will be used,
which is not satisfactory in most cases.
So, how do we proceed if we really want to override the <i>equals</i> method
inherited from <i>Object</i> rather than overload it?
We present to you a possible way of overriding the <i>equals</i> method,
a fairly common solution that you can find in the litterature.
Note that there are other ways of writing this <i>equals</i> method,
this is just one solution among others.
The main difficulty we face
when we want to override the <i>equals</i> method
is that now, it can take as parameter
any type of object. When we used overloading,
the parameter had the same type
as the class in which we defined the overload, so this here,
and invoking the <i>equals</i> method on something other than a rectangle
for example a string, caused the compiler to react.
If there was just an overload:
the compiler would tell you:
"I am expecting a String, and you are giving me a Rectangle"
However, with an override, this form becomes legal,
the compiler won't say anything, it will accept this without a hitch.
Why? Because an object of type <i>String</i> is an <i>Object</i> by inheritance
and so this is totally legal, I can have a <i>String</i> in an <i>Object</i>.
Note that if we do not redefine the <i>equals</i> method in the Rectangle class,
at the moment we made such a call, we would use the <i>equals</i> method
as it is defined in the <i>Object</i> class, which only compares references
and doesn't necessarily apply the procedures
we would like to use
when we compare a rectangle with another object.
So it is up to the programmer of the <i>equals</i> method
to correctly define the body of their method
in order for the comparison with objects of types other than
rectangles to be donne correctly, for example by returning <i>false</i>.
We want a rectangle to be comparable to another rectangle,
but not to an object of another type.
To guarantee that a rectangle can be equal only
to another rectangle,
we have to test if the object passed as parameter is of the same class
as the Rectangle class.
This can be done using the <i>getClass</i> method
inherited once again from <i>Object</i>.
With this method, we can test if the class of the other object
is the same as the class of the current instance.
And if not, we return <i>false</i>.
To summarize, when we want to override the <i>equals</i> method,
we start, as we did for overloading,
by testing if the parameter has a <i>null</i> value:
if it is <i>null</i>, we return <i>false</i>. If not, and this is the new part,
we test if the other object is of the same class as the current instance.
If not, we return <i>false</i> here too.
And if so, if we know that <i>autreObjet</i> (TN: means "otherObject") is not <i>null</i>,
that the rectangle passed as parameter is indeed of class Rectangle,
then we can proceed to an attribute-by-attribute comparison,
as we did in the case of overloading.
But here, we have another little difficulty to overcome:
indeed, if I want to compare attribute-by-attribute,
I must test if the width of the current object
is the same as the width as the other object, <i>autreObjet</i>,
and if the height of the current object
is the same as the height of the other object.
Yet <i>autreObjet</i> is of type <i>Object</i>, which does not guarantee the presence
of the height and width attributes, <i>largeur</i> and <i>hauteur</i>.
We know that <i>autreObjet</i> does indeed contain a Rectangle,
we made sure of that with the precautions we took.
However, if we write something like this,
so if I try to compare the width of the current object
with the width of the other object,
by using this statement here,
the compiler will throw an error message
telling us that <i>autreObjet</i> is of type <i>Object</i>
which does not contain a <i>largeur</i> field.
Here, we convert <i>autreObjet</i> to a Rectangle. This is known as casting.
This guarantees to the compileer that it is possible to access the <i>largeur</i> field.
This will work perfectly here, since we have guaranteed
that at execution time, <i>autreObjet</i> will contain an object of type Rectangle.
If we try to do a cast
when <i>autreObjet</i> does not contain a Rectangle,
we will get an error message at execution time.
But this is not the case here.
So, basically, this is how we can proceed
to override the <i>equals</i> method
with the new dangers that arise
when we use overriding rather than overloading.
This ends our short episode on complements.
