3.4 – Simple Inheritance
Exercise 1: Colon Syntax
The colon syntax can improve the performance of constructors. To test this, make sure that you print some text in the point’s constructors, destructor and also the assignment operator.
Now, execute the following code in the test program and count the number of point constructor, destructor and assignment calls:
Line l;
Now, change the constructors in the Line class to use the colon syntax to set the start- and end-point data members and run the test program again. Is the number of point constructor, destructor and assignment calls less than before?
Apply the colon syntax also for the Point class constructors and if applicable also for the Circle class.

Exercise 2: Creating Shape Base Class
It can be useful to create a hierarchy of related classes using base- and derived classes.
• Classes are related (same family)
• Common data and functionality can be put in a base class.
• You can work with derived classes as if it is the base class.
In this exercise we are going to transform the Point and Line class into a Shape hierarchy as shown in Figure 1.
Shape
-m_id
+ID() +ToString()
Point
+ToString()
Line
Figure 1: Shape Hierarchy
First create a Shape base class.
• Add a source- and header file for a Shape class.
• Add a data member for an id number of type int.
• Add a default constructor that initializes the id using a random number. You can use
the rand() function from the “stdlib.h” header file.
• Add a copy constructor that copies the id member.
• Add an assignment operator that copies the id member.
• Add a ToString() function that returns the id as string e.g. “ID: 123”.
• Add an ID() function the retrieve the id of the shape.
Next the Point and Line classes (and the Circle class if applicable) must derive from Shape.
• Add theShapeclass in the inheritance list of thePoint,Lineand optionally the Circle class.
• The constructors of the Point, Line and optionally the Circle class should call the appropriate constructor in the Shapebase class.
• The assignment operator should call the assignment operator of theShapebase class. Otherwise the shape data will not be copied.
• Finally add code to the main program to test inheritance:
Circle
+ToString()
+ToString()

Shape s;
Point p(10, 20);
Line l(Point(1,2), Point(3, 4));
cout<<s.ToString()<<endl;
cout<<p.ToString()<<endl;
cout<<l.ToString()<<endl;
cout<<“Shape ID: “<<s.ID()<<endl;
cout<<“Point ID: “<<p.ID()<<endl;
cout<<“Line ID: “<<l.ID()<<endl;
Shape* sp;
sp=&p;
cout<<sp->ToString()<<endl;
// Create shape.
// Create point.
// Create line.
// Print shape.
// Print point.
// Print line
// ID of the shape.
// ID of the point. Does this work?
// ID of the line. Does this work?
// Create pointer to a shape variable.
// Point in a shape variable. Possible?
// What is printed?
// Create and copy Point p to new point.
Point p2;
p2=p;
cout<<p2<<“, “<<p2.ID()<<endl; // Is the ID copied if you do not call
// the base class assignment in point?
Answer the questions in the comments of the code above.

3.5 – Polymorphism
Exercise 1: Polymorphic ToString() Function
The ToString() functions of Point and Line override the ToString() from the Shape base class. We saw that we could put aPoint in a Shape* variable. But when calling
the ToString() method on the Shape* variable, the function in Shape was called instead the one in Point. To make the compiler generate the required code to find out what type of object the Shape*variable is actually pointing to so it can call the right version; we need to declare the function as virtual.
Thus declare the ToString() function in the Shape class as virtual and test the program again. Is the ToString() function of Point called when you use a Shape* that contains a Point now?
Exercise 2: Calling Base Class Functionality
The ToString() function of the Shape class is overridden in the derived classes. But for the derived class it is still possible to use the base class functionality. In the ToString() function of Point and Line we also want to incorporate the ID from theShape base class.
• In the ToString() method of Point, call the ToString() method of the Shape base class: std::string s=Shape::ToString();
• Append the shape description string to the point description string before returning.
• Do this also for the ToString() function in the Line class (and Circle class).
• Test the application again. Is now the ID printed when printing a point or line?
Exercise 3: Virtual Destructors
When objects are removed from memory, the destructor is called. When a derived class destructor is called, it will automatically call the base class destructor. But when you have pointers to a base class, deleting objects might not be done correctly.
If not done already, print some text in the destructors of the Shape, Point and Line classes. Then test the following code:
Shape* shapes[3];
shapes[0]=new Shape;
shapes[1]=new Point;
shapes[2]=new Line;
for (int i=0; i!=3; i++) delete shapes[i];
Will the proper destructors (including the destructor of the Shape base class) be called?
In this case, the derived class destructor will only be called when the destructor is
declared virtual in the base class. Do this in the Shape class and run the code again. Are the proper destructors called now?

Exercise 4: Abstract Functions
Sometimes functions in the base class are only there to be overridden in the derived class. Assume that you want to draw all the shapes using the following code:
Shape* shapes[10];
shapes[0]=new Line;
shapes[1]=new Point;
…
shapes[9]=new Line(Point(1.0, 2.5), Point(3.4, 5.2));
for (int i=0; i!=10; i++) shapes[i]->Draw(); for (int i=0; i!=10; i++) delete shapes[i];
Create the Draw() function in the Shape base class and override it in the derived classes (point, line and if present the circle class). Simulate drawing by just printing some text.
What implementation did you give the Draw() function in Shape? Shape is just an abstraction to work with various kinds of shapes like lines and circles. Shapes don’t have a physical appearance. Therefore its Draw() function will have an empty implementation. But better is to give it no implementation at all by making it a pure virtual member function:
virtual void Draw()=0;
Do this in your code. Try to create an instance of the Shape class. Is this possible?
Now the Shape class is really an abstraction. You don’t make shape instances but you can still create shape pointers that point to concrete shapes like point and line. The Shape class is now an abstract base class.
Shape
-m_id
+ID() +ToString() +Draw()
Point
+ToString() +Draw()
Line
Figure 2: Abstract Shape base class with Draw() pure virtual member function
Circle
+ToString() +Draw()
+ToString() +Draw()

Exercise 5: Template Method Pattern
In this exercise we are going to create a Print() function that is printing the shape information to the cout object. ThePrint() function can use the ToString() to obtain the string to print.
You see that the implementation of Print() is in each derived class largely the same; calling ToString() and sending the result to cout. Since the ToString() function is polymorphic, we can use the polymorphic behavior in the Print() function of Shape.
Thus:
• Add a Print() function to the Shape class.
• In this function, call the ToString() function and send the result to the cout object.
• In the test program, create a point and line object and call the Print() function. Does it
print the right information even when point and line do not have the Print() function?
You have now created a function for the base class (Print()) that does all the functionality common to all derived classes. Only the part of that function that is different for each derived class is delegated to a polymorphic function (ToString()). This mechanism is an often used design pattern called the “Template Method Pattern”.

3.6 – Exception Handling
Exercise 1: Bounds Checking Array
In the array class we created previously, the bounds checking was very basic. There was no error generated, but setting an element was ignored or the first element was returned. Obviously you want to know if there was an out of bounds error. This is possible by exception handling.
Change the Array class to throw exceptions:
• In the GetElement(), SetElement() and index operator throw -1 if the index was too small or too big.
• In the main program, create an Array object and access an element that does not exist. Run the program. What happens?
• The exception must be caught, so place the code inside a try … catch block that catches an int.
• In the catch handler, print an error message.

Exercise 2: Exception Objects
Throwing an int is an easy solution. But exception handling is also object oriented and allows us to throw an object.
In this exercise we will create an exception hierarchy with an ArrayException base class and an OutOfBoundsExceptionderived class as shown in Figure 3:
• You can implement both exception classes in the header file for simplicity.
• Give the ArrayException an abstract GetMessage() function that returns a std::string.
• Give the OutOfBoundsException class a constructor with an int as argument that
indicates the erroneous array index and store it in a data member.
• Override the GetMessage() function and let the implementation return a message
string saying the given index is out of bounds.
• In the Array class, throw now a OutOfBoundsException object instead of an integer.
• Change the main program so that it catches the ArrayException base class and uses
the GetMessage() function to display an error message. catches
uses throws
Figure 3: Array Exception
MainProgram
ArrayException
+main()
+GetMessage()
-m_data -m_size
Array
T
OutOfBoundsException
-m_index
+SetElement() +GetElement() +operator []()
+GetMessage()