MORE INHERITANCE
In the last chapter we developed a model using modes of transportation to illustrate the concept of inheritance. In this chapter we will use that model to illustrate some of the finer points of inheritance and what it can be used for. If it has been a while since you read and studied chapter 7, it would be good for you to return to that material and review it in preparation for a more detailed study of the topic of inheritance.
REORGANIZED FILE STRUCTURE
Example program ------> INHERIT1.CPP
A close examination of the file named INHERIT1.CPP will reveal that it is identical to the program developed in chapter 7 named ALLVEHIC.CPP except that the program text is rearranged. The biggest difference is that some of the simpler methods in the classes have been changed to inline code to shorten the file considerably. In a practical programming situation, methods that are this short should be programmed inline since the actual code to return a simple value is shorter than the code required to send a message to a non-inline method.
The only other change is the reordering of the classes and associated methods with the classes all defined first, followed by the main program. This puts all class interface definitions on a single page to make the code easier to study. The implementations for the methods are deferred until the end of the file where they are available for quick reference but are not cluttering up the class definitions which we wish to study carefully in this chapter. This should be an indication to you that there is considerable flexibility in the way the classes and methods can be arranged in C++. Of course you realize that this violates the spirit of C++ and its use of separate compilation, but is only done here for convenience. The best way to package all of the example programs in this chapter is like the packaging illustrated in chapter 7.
As mentioned before, the two derived classes, car and truck, each have a variable named passenger_load which is perfectly legal. The car class has a method of the same name, initialize(), as one declared in the super-class named vehicle. The rearrangement of the files in no way voids this allowable repeating of names.
After you have convinced yourself that this program is truly identical to the program named ALLVEHIC.CPP from chapter 7, compile and execute it with your compiler to assure yourself that this arrangement is legal. Due to this means of code packaging, you will not need a "make" file or a "project" capability to compile and execute this code. This code arrangement is designed to make it easy to compile and execute the example programs in this chapter.
THE SCOPE OPERATOR
Because the method initialize() is declared in the derived car class, it hides the method of the same name which is part of the base class, and there may be times you wish to send a message to the method in the base class for use in the derived class object. This can be done by using the scope operator in the following manner in the main program;
sedan.vehicle::initialize(4, 3500.0);
As you might guess, the number and types of parameters must agree with those of the method in the base class because it will respond to the message.
WHAT IS PROTECTED DATA?
If the data within a base class were totally available in all classes inheriting that base class, it would be a simple matter for a programmer to inherit the base class into a derived class and have free access to all data in the parent class. This would completely override the protection afforded by the use of information hiding. For this reason, the data in a class are not automatically available to the methods of an inheriting class. There are times when you may wish to automatically inherit all variables directly into the subclasses and have them act just as though they were declared as a part of those classes also. For this reason, the designer of C++ has provided the keyword protected.
In the present example program, the keyword protected is given in line 6 so that all of the data of the vehicle class can be directly imported into any derived classes but are not available outside of the base class or derived classes. As we have mentioned before, all data are automatically defaulted to private at the beginning of a class if no specifier is given.
You will notice that the variables named wheels and weight are available to use in the method named initialize() in lines 83 through 88 just as if they were declared as a part of the car class itself. They are available because they were declared protected in the base class. Of course, they would be available here if they had been decared public in the base class, but then they would be available outside of both classes and we would lose our protection. Note that the two variables are available for use in the base class as illustrated in lines 77 and 78. We can now state the rules for the three means of defining variables and methods.
You will note that these three means of definition can also be used in a struct type. The only difference with a struct is that everything defaults to public until one of the other keywords is used.
Be sure to compile and execute this program before continuing on to the next example program.
WHAT IS PRIVATE DATA?
Example program ------> INHERIT2.CPP
Examine the file named INHERIT2.CPP where the data in the base class is permitted to use the private default because line 6 is commented out. In this program, the data is not available for use in the derived classes, so the only way the data in the base class can be used is by sending messages to methods in the base class, even within the derived class.
It seems a little silly to have to call methods in the base class to access the data which is actually a part of the derived class, but that is the way C++ is defined to work. This would indicate to you that you should spend some time thinking about how any class you define will be used. If you think somebody may wish to inherit your class into a new class and expand it, you should make the data members protected so they can be easily used in the new class. Lines 86 and 87 are invalid now since the members are not visible, but line 88 now does the job they did before they were hidden by calling the public method of the base class. Line 104 is also changed because of the hidden data. You will notice that the data is still available in lines 77 and 78 just as they were before because the member variables are protected in the vehicle class. Be sure to compile and execute this program.
HIDDEN METHODS
Example program ------> INHERIT3.CPP
Examine the file named INHERIT3.CPP carefully and you will see that it is a repeat of the first example program in this chapter with a few minor changes.
You will notice that the derived classes named car and truck do not have the keyword public prior to the name of the base class in the first line of each class declaration. The keyword public, when included prior to the base class name, makes all of the methods defined in the base class available for use in the derived class at the same security level as they were defined at in the base class. Therefore, in the previous program, we were permitted to call the methods defined as part of the base class from the main program even though we were working with an object of one of the derived classes.
In this program, all entities are inherited as private due to the use of the keyword private prior to the name of the base class. They are therefore unavailable to any code outside of the derived class. The general rule is that all elements are inherited into the derived class at the most restrictive of the two restrictions placed on them, one being the definition in the base class and the other being the restriction on inheritance. This defines the way the elements are viewed outside of the derived class.
The elements are all inherited into the derived class such that they have the same level of protection they had in the base class, as far as their visibility restrictions within the derived class. We have returned to the use of protected data instead of private in the base class, therefore the member variables are available for use within the derived class.
In the present program, the only methods available for objects of the car class, are those that are defined as part of the class itself, and therefore we only have the methods named initialize() and passengers() available for use with objects of class car.
When we declare an object of type car, according to the definition of the C++ language, it contains three variables. It contains the one defined as part of its class named passenger_load and the two that are part of its parent class, wheels and weight. All are available for direct use within its methods because of the use of the keyword protected in the base class. The variables are a part of an object of class car when it is declared and are stored as part of the object.
The observant student will notice that several of the output statements have been commented out of the main program since they are no longer legal or meaningful operations. Lines 57 through 59 have been commented out because the methods named get_weight() and wheel_loading() are not available as members of the car class because we are using private inheritance. You will notice that initialize() is still available but this is the one in the car class, not the method of the same name in the vehicle class.
USING THE TRUCK CLASS
Moving on to the use of the truck class in the main program, we find that lines 63 and 65 are commented out for the same reason as given above, but lines 66 and 67 are commented out for an entirely different reason. Even though the method named efficiency() is available and can be called as a part of the truck class, it cannot be used because we have no way to initialize the wheels or weight of the truck object. We can get the weight of the truck object, as we have done in line 102, but since the weight has no way to be initialized, the result is meaningless and lines 66 and 67 are commented out.
Private inheritance is very similar to using an embedded object and, in fact, is very rarely used. Until you gain a lot of experience with C++ and the proper use of Object Oriented Programming, you should use public inheritance exclusively. There is probably no reason to ever use private or protected inheritance. They were probably added to the language for completeness.
Be sure to compile and execute this example program to see that your compiler gives the same result. It would be a good exercise for you to reintroduce some of the commented out lines to see what sort of an error message your compiler issues for these errors.
INITIALIZING ALL DATA
Example program ------> INHERIT4.CPP
If you will examine the example program named INHERIT4.CPP, you will find that we have fixed the initialization problem that we left dangling in the last example program. We also added default constructors to each of the classes so we can study how they are used when we use inheritance and we have returned to the use of public inheritance.
When we create an object of the base class vehicle, there is no problem since inheritance is not a factor. The constructor for the base class operates in exactly the same manner that all constructors have in previous chapters. You will notice that we create the unicycle object in line 47 using the default constructor and the object is initialized to the values contained in the constructor. Line 49 is commented out because we no longer need the initialization code for the object.
When we define an object of one of the derived classes in line 57, the sedan, it is a little different because not only do we need to call a constructor for the derived class, we have to worry about how we get the base class initialized through its constructor also. Actually, it is no problem because the compiler will automatically call the default constructor for the base class unless the derived class explicitly calls another constructor for the base class. We will explicitly call another constructor in the next example program, but for now we will only be concerned about the default constructor for the base class that is called automatically.
ORDER OF CONSTRUCTION
The next problem we need to be concerned about is the order of construction, and it is easy to remember if you remember the following statement, "C++ classes honor their parents by calling their parents constructor before they call their own." The base class constructor will be called before the derived class constructor. This makes sense because it guarantees that the base class is properly constructed when the constructor for the derived class is executed. This allows you to use some of the data from the base class during construction of the derived class. That was a mouthful, but if you spend a little time with this concept, it will make a lot of sense, and you will not easily forget it. In this case, the vehicle part of the sedan object is constructed, then the local portions of the sedan object will be constructed, so that all member variables are properly initialized. This is why we can comment out the initialize method in line 59. It is not needed.
When we define a semi object in line 66, it will also be constructed in the same manner. The constructor for the base class is executed, then the constructor for the derived class will be executed. The object is now fully defined and useable with default data in each member. Lines 68 and 69 are therefore not needed.
The remainder of this program should be no problem for you to understand except for the order of destruction of the various objects.
HOW ARE THE DESTRUCTORS EXECUTED?
As the objects go out of scope, they must have their destructors executed also, and since we didn't define any, the default destructors will be executed. Once again, the destruction of the base class object named unicycle is no problem, it's destructor is executed and the object is gone. The sedan object however, must have two destructors executed to destroy each of its parts, the base class part and the derived class part. It should not be too much of a surprise that the destructors for this object are executed in reverse order from the order in which they were constructed. In other words, the object is dismantled in the opposite order from the order in which it was assembled. The derived class destructor is executed first, then the base class destructor is executed and the object is removed from they system.
Remember that every time an object is defined or created, every portion of it must have a constructor executed on it. Every object must also have a destructor executed on each of its parts when it is destroyed in order to properly dismantle the object.
Be sure to compile and execute this program following your detailed study of it.
INHERITANCE WHEN CONSTRUCTORS ARE USED
Example program ------> INHERIT5.CPP
Examine the example program named INHERIT5.CPP for yet another variation to our basic program, this time using constructors that are more than just the default constructors. You will notice that each class has another constructor declared within it. The additional constructor added to the vehicle class in lines 12 through 14 is nothing special, it is just like some of the constructors we have studied earlier in this tutorial. It is used in line 59 of the main program where we define unicycle with two values passed in to be used when executing this constructor.
The constructor for the car class which is declared in lines 28 through 31 is a bit different, because we pass in three values. One of the values, the one named people, is used within the derived class itself to initialize the member variable named passenger_load. The other two literal values however, must be passed to the base class somehow in order to initialize the number of wheels and the weight. This is done by using a member initializer, and is illustrated in this constructor. The colon near the end of line 28 indicates that a member initializer list follows, and all entities between the colon and the opening brace of the constructor body are member initializers. The first member initializer is given in line 29 and looks like a constructor call to the vehicle class that requires two input parameters. That is exactly what it is, and it calls the constructor for the vehicle class and initializes that part of the sedan object that is inherited from the vehicle class. We can therefore control which base class initializer gets called when we construct an object of the derived class.
The next member initializer, in line 30, acts kind of like a constructor for a simple variable. By mentioning the name of the variable and including a value of the correct type within the parentheses, that value is assigned to that variable even though the variable is not a class, but a simple predefined type. This technique can be used to initialize all members of the derived class or any portion of them. When all of the members of the member initializer list are executed, the code within the braces is executed. In this case, there is no code within the executable block of the constructor. The code within the braces would be written in a normal manner for a constructor.
WHAT ABOUT THE ORDER OF EXECUTION?
This may seem to be very strange, but the elements of the member initializer list are not executed in the order in which they appear in the list. The constructors for the inherited classes are executed first, in the order of their declaration in the class header. When using multiple inheritance, several classes can be listed in the header line, but in this program, only one is used. The member variables are then initialized, but not in the order as given in the list, but in the order in which they are declared in the class. Finally, the code within the constructor block is executed, if there is any code in the block.
There is a good reason for this seemingly strange order. The destructors must be executed in reverse order from the construction order, but if there are two constructors with different construction order defined, which should define the destruction order? The correct answer is neither. The system uses the declaration order for construction order and reverses it for the destruction order.
You will notice that the truck class uses one initializer for the base class constructor and two member initializers, one to initialize the passenger_load, and one to initialize the payload. The body of the constuctor, much like the car class, is empty.
The two constructors in the car class and the truck class are called to construct objects in lines 69 and 78 for a car object and a truck object as illustrations in this example program.
The remainder of this program should be easy for you to folow. Be sure to compile and execute this program before moving on.
POINTERS TO AN OBJECT AND AN ARRAY OF OBJECTS
Example program ------> INHERIT6.CPP
Examine the example program named INHERIT6.CPP for examples of the use of an array of objects and a pointer to an object. In this program, the objects are instantiated from an inherited class and the intent of this program is to illustrate that there is nothing magic about a derived class. A class acts the same whether it is a base class or a derived class.
This program is identical to the first program in this chapter until we get to the main() program where we find an array of 3 objects of class car declared in line 53. It should be obvious that any operation that is legal for a simple object is legal for an object that is part of an array, but we must be sure to tell the system which object of the array we are interested in by adding the array subscript as we do in lines 58 through 64. The operation of this portion of the program should be very easy for you to follow, so we will go on to the next construct of interest.
You will notice, in line 68, that we do not declare an object of type truck but a pointer to an object of type truck. In order to use the pointer, we must give it something to point at which we do in line 70 by dynamically allocating an object. Once the pointer has an object to point to, we can use the object in the same way we would use any object, but we must use the pointer notation to access any of the methods of the object. This is illustrated for you in lines 76 through 80, and will be further illustrated in the example program of chapter 12 in this tutorial.
Finally, we deallocate the object in line 81. You should spend enough time with this program to thoroughly understand the new material presented here, then compile and execute it.
THE NEW TIME CLASS
We began a series of nontrivial classes in chapter 5 where we developed a date class, then a time class, and finally a newdate class in the last chapter. Now it is your turn to add to this series. Your assignment is to develop the newtime class which inherits the time class and adds a new member variable named seconds_today and a method to calculate the value of seconds since midnight to fill the variable.
A complete solution to this problem will be found in cppans.zip available for download. The files named NEWTIME.H, NEWTIME.CPP, and TRYNTIME.CPP are the solution files. It would be a good exercise for you to attempt to write this new class before you look at the example solution.
PROGRAMMING EXERCISES
Return to the Table of Contents